Automated digital knowledge formation

Abstract

A method of electronic communication assistance involves receiving a partial electronic communication at an Al assistant computing facility from a first electronic identifier linked to a first user. This communication includes content associated with both the first user and a second user. The method extracts the communication context and encodes the partial communication for processing, creating an encoded version. It retrieves the first user's communication profile from a database using their identifier, containing user attributes, and retrieves the second user's profile similarly. The encoded communication is then processed by a processor to generate a compositional change using at least one of the communication context, the first user communication attribute, or the second user communication attribute. Finally, a revised electronic communication is generated from the partial communication and the compositional change.

Description

AUTOMATED DIGITAL KNOWLEDGE FORMATIONCOPYRIGHT NOTICE

[0001] A portion of the disclosure of this patent document contains material that is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the Patent and Trademark Office patent file or records, but otherwise reserves all copyright or rights. © 2023 Grammarly, Inc.BENEFIT CLAIM

[0002] This application claims the benefit under 35 U.S.C. § 119(e) of provisional application 63 / 519,120, filed August 11, 2023, the entire contents of which are hereby incorporated by reference for all purposes as if fully set forth herein.TECHNICAL FIELD

[0003] One technical field of the present disclosure is computer-supported natural language processing, including the automatic composition of natural language text. Another technical field is software engineering as applied to large language models in artificial intelligence systems, particularly techniques for engineering prompts and context to submit during the execution of the inference stage of an LLM. Other fields are computer-implemented digital assistants, methods, and systems using artificial intelligence assistants to increase the effectiveness of communications.BACKGROUND

[0004] The approaches described in this section are approaches that could be pursued but not necessarily approaches that have been previously conceived or pursued. Therefore, unless otherwise indicated, it should not be assumed that any of the approaches described in this section qualify as prior art merely by virtue of their inclusion in this section.

[0005] Current digital writing assistants provide simple aids to sending communications, such as illustrated in FIG. 1 , where the text of an electronically composed communication is evaluated by a ‘writing assistant’ to correct content errors, such as spelling or grammatical errors. However, these simple aids are very limited, generally only addressing corrections tocontent. What is needed is an intelligent digital assistant system that can incorporate knowledge and context surrounding the communication exchanges to make a communication exchange more effective.

[0006] Large language models (LLMs), such as GPT-class LLMs, are widely used to generate a text response to a prompt or query. Public LLMs, accessible via application programming interfaces (APIs), like GOOGLE BARD and OPENAI GPT, have been trained on a corpus of data that the user does not access, see, or control. Consequently, in some application areas, the responses that public LLMs provide to prompts are inadequate or incomplete. The responses suffer from the inability of the LLM or its provider to access application- specific training data. One example arises in large business enterprises that have historically created and stored large numbers of internal documents relating to business operations, technology, product plans, and other enterprise-specific domains or topics.

[0007] Based on the foregoing, the referenced technical fields have developed an acute need for better ways to efficiently and effectively access the vast knowledge captured in enterprise-specific documents for use in enterprise-specific prompts and responses.SUMMARY

[0008] The appended claims may serve as a summary of the invention.BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In the drawings:

[0010] FIG. 1 depicts a prior art example of a writing assistant.

[0011] FIG. 2 illustrates a functional flow diagram of an artificial intelligence assistant according to exemplary and non-limiting embodiments.

[0012] FIG. 3 illustrates aspects of an artificial intelligence assistant according to exemplary and non-limiting embodiments.

[0013] FIG. 4 illustrates an overview of a comprehensive communication assistance workflow according to exemplary and non-limiting embodiments.

[0014] FIG. 5 illustrates an overview of a composition workflow according to exemplary and non-limiting embodiments.

[0015] FIG. 6 illustrates an overview of an editing workflow according to exemplary and non-limiting embodiments.

[0016] FIG. 7 illustrates a reaction prediction workflow according to exemplary and nonlimiting embodiments.

[0017] FIG. 8 illustrates a reaction prediction workflow with ranking according to exemplary and non-limiting embodiments.

[0018] FIG. 9 illustrates a reaction capture workflow according to exemplary and nonlimiting embodiments.

[0019] FIG. 10 illustrates a workflow for optimizing an incoming communication according to exemplary and non-limiting embodiments.

[0020] FIG. 11 illustrates a workflow for optimizing an incoming communication and feedback generation according to exemplary and non-limiting embodiments.

[0021] FIG. 12 illustrates a receiver interface workflow according to exemplary and nonlimiting embodiments.

[0022] FIG. 13 illustrates an audio- visual communication workflow according to exemplary and non-limiting embodiments.

[0023] FIG. 14, FIG. 15, FIG. 16, FIG. 17, FIG. 18, FIG. 19, and FIG. 20 illustrate user interface views of an artificial intelligence assistant according to exemplary and non-limiting embodiments.

[0024] FIG. 21, FIG. 22, FIG. 23, FIG. 24, FIG. 25, FIG. 26, and FIG. 27 illustrate an example augmented reality glasses user interface view of an artificial intelligence assistant according to exemplary and non-limiting embodiments.

[0025] FIG. 28 illustrates a distributed computer system showing the context of use and principal functional elements with which one embodiment could be implemented.

[0026] FIG. 29 illustrates an example of structural architecture and data flow for the connector administration system of FIG. 28.

[0027] FIG. 30 illustrates an example of structural architecture and data flow for the crawler system of FIG. 28.

[0028] FIG. 31 illustrates an example of structural architecture and data flow for the search system of FIG. 28.

[0029] FIG. 32 illustrates an example of structural architecture and data flow for the storage system of FIG. 28.

[0030] FIG. 33 illustrates an example method for generating a prompt and using it with a large language model (LLM) to obtain a response.

[0031] FIG. 34A, FIG. 34B, FIG. 34C, and FIG. 34D depict example processes for generating a prompt to be provided to an LLM.

[0032] FIG. 35 presents an example diagram demonstrating the incorporation of actions for prompt generation.

[0033] FIG. 36 shows an illustrative system designed to facilitate prompt generation.

[0034] FIG. 37 provides an illustrative diagram for determining the usefulness of context in providing knowledge to generate an appropriate response from the LLM.

[0035] FIG. 38 illustrates a computer system with which one embodiment could be implemented.DETAILED DESCRIPTION

[0036] In the following description, for the purposes of explanation, numerous specific details are set forth to provide a thorough understanding of the present invention. It will be apparent, however, that the present invention may be practiced without these specific details. In other instances, well-known structures and devices are shown in block diagram form to avoid unnecessarily obscuring the present invention.

[0037] The text of this disclosure, in combination with the drawing figures, is intended to state in prose the algorithms that are necessary to program the computer to implement the claimed inventions at the same level of detail that is used by people of skill in the arts to which this disclosure pertains to communicate with one another concerning functions to be programmed, inputs, transformations, outputs and other aspects of programming. That is, the level of detail set forth in this disclosure is the same level of detail that persons of skill in the art normally use to communicate with one another to express algorithms to be programmed or the structure and function of programs to implement the inventions claimed herein.

[0038] This disclosure may describe one or more different inventions, with alternative embodiments to illustrate examples. Other embodiments may be utilized, and structural, logical, software, electrical, and other changes may be made without departing from the scope of the invention. Various modifications and alterations are possible and expected. Some features of one or more of the inventions may be described with reference to one or more embodiments or drawing figures, but such features are not limited to usage in the one or more embodiments or figures with reference to which they are described. Thus, the present disclosure is neither a literal description of all embodiments of one or more inventions nor a listing of features of one or more inventions that must be present in all embodiments.

[0039] Headings of sections and the title are provided for convenience but are not intended to limit the disclosure in any way or as a basis for interpreting the claims. Devices described as in communication with each other need not be in continuous communication with each other, unless expressly specified otherwise. In addition, devices that communicate with each other may communicate directly or indirectly through one or more intermediaries, logical or physical.

[0040] A description of an embodiment with components in communication with one other does not imply that all such components are required. Optional components may be described to illustrate a variety of embodiments and to illustrate one or more aspects of the inventions fully. Similarly, although process steps, method steps, algorithms, or the like may be described in sequential order, such processes, methods, and algorithms may generally be configured to work in different orders unless specifically stated to the contrary. Any sequence or order of steps described in this disclosure is not a required sequence or order. The steps of described processes may be performed in any order practical. Further, some steps may be performed simultaneously. The illustration of a process in a drawing does not exclude variations and modifications, does not imply that the process or any of its steps are necessary to one or more of the invention(s), and does not imply that the illustrated process is preferred. The steps may be described once per embodiment but need not occur only once. Some steps may be omitted in some embodiments or occurrences, or some steps may be executed more than once in each embodiment or occurrence. When a single device or article is described, more than one device or article may be used in place of a single device or article. Where more than one device or article is described, a single device or article may be used instead of more than one device or article.

[0041] The functionality or features of a device may be alternatively embodied by one or more other devices that are not explicitly described as having such functionality or features. Thus, other embodiments of one or more inventions need not include the device itself. Techniques and mechanisms described or referenced herein will sometimes be described in singular form for clarity. However, it should be noted that embodiments include multiple iterations of a technique or manifestations of a mechanism unless noted otherwise. Process descriptions or blocks in figures should be understood as representing modules, segments, or portions of code, including one or more executable instructions for implementing specific logical functions or steps in the process. Alternate implementations are included within the scope of embodiments of the present invention in which, for example, functions may beexecuted out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved.

[0042] 1. GENERAL OVERVIEW

[0043] A computer-implemented client process can programmatically call functions of a common application programming interface (CAPI) using a call that conveys a request for a response to a user prompt or user query. In response, the CAPI programmatically calls a search system and provides the user prompt or user query with a request to obtain a list of documents that are relevant to the context of the user prompt or user query. The search system, having previously crawled and indexed a large corpus of documents, and having accessed definitions or other resources relating to terms in the documents, returns a list of documents that are relevant to the context of the user prompt or user query. The CAPI forwards the user interaction context, organizational context, and user prompt or user query to a prompt-based text composition service. The prompt-based text composition service communicates programmatically with prompt-building instructions, for example, by retrieving or reading action templates and decorators for an LLM prompt. The prompt-based text composition service calls a public API of an open large language model (LLM), such as a GPT-type LLM, and sends an engineered prompt, the user query, and context data. In response to receiving a response from the LLM, the prompt-based text composition service formats, filters, and / or returns the response to the CAPI, which can further format or filter the response and transmit a final response to the client process.

[0044] In an embodiment, a method of electronic communication assistance may include receiving a partial electronic communication at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the partial electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user; extracting a communication context from the partial electronic communication; encoding the partial electronic communication for processing creating an encoded partial electronic communication; retrieving from a communication profile database a first communication profile for the first user using the first electronic identifier associated with the first user, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user using the second electronic identifier associated with the second user, wherein the second communication profile comprises a second usercommunication attribute; processing the encoded partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using at least one of the communication context, the first user communication attribute, or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0045] In an embodiment, a method of electronic communication assistance may include receiving a partial electronic communication at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the partial electronic communication comprising a communication content and comprising or associated with the first electronic identifier and a second electronic identifier associated with a second user; retrieving a first communication profile, wherein the first communication profile comprises a first user communication attribute; retrieving a second communication profile, wherein the second communication profile comprises a second user communication attribute; processing the partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using at least one of the first user communication attribute or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0046] In an embodiment, the method may further include transmitting the changed electronic communication to the first electronic identifier associated with the first user, and / or transmitting the changed electronic communication to the second electronic identifier associated with the second user. The compositional change may be derived from representations of previous content and context from a plurality of user profiles stored in the communication profile database which are like at least one of the first communication profile or the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning model, deep learning model, or other statistical learning algorithm for creating the compositional change. The compositional change may be an auto-generated textual completion; the auto-generated textual completion may be a phrasal completion, and the processor may generate the compositional change by optimizing generated language as determined by the processor from the second user communication attribute. The processor may generate the compositional change by replicating acommunication style of the first user as determined by the processor from the first user communication attribute. The partial electronic communication may include a communication goal, and the processor may generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion. The processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find a most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automatically generated template generated by the processor. The processor may select the communication template by using a machine learning model, deep learning model, or statistical learning algorithm to score the plurality of communication templates based at least in part on the communication content, first user communication attribute, or second user communication attribute, or communication context.

[0047] In an embodiment, a method of electronic communication assistance may include receiving a partial electronic communication at an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the partial electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user; extracting a communication context from the partial electronic communication; encoding the partial electronic communication for processing creating an encoded partial electronic communication; retrieving from a communication profile database a communication profile for the user using the electronic identifier associated with the user, wherein the communication profile comprises a user communication attribute; processing the encoded partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using at least one of the communication context or the user communication attribute; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0048] In an embodiment, the method may include transmitting the changed electronic communication to the electronic identifier associated with the user or transmitting the changed electronic communication to a second electronic identifier associated with a second user. The compositional change may be derived from representations of previous content and contextfrom a plurality of user profiles stored in the communication profile database which are like the communication. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning model, deep learning model, or other statistical learning algorithm for creating the compositional change. The compositional change may be an auto-generated textual completion. The auto-generated textual completion may be a phrasal completion. The processor may generate the compositional change by optimizing generated language of a user as determined by the processor from a user communication attribute, may generate the compositional change by replicating a communication style of the user as determined by the processor from the user communication attribute. The partial electronic communication may include a communication goal, and the processor may generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion, the processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find a most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automatically generated template generated by the processor. The processor may select the communication template by using a machine learning model to score the plurality of communication templates based at least in part on the communication content, communication context, first user communication attribute, or second user communication attribute.

[0049] In an embodiment, a method of electronic communication assistance may include receiving a partial electronic communication at an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the partial electronic communication comprising a communication content and comprising or associated with the electronic identifier; retrieving a communication profile, wherein the communication profile comprises a user communication attribute; processing the partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using the user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0050] In an embodiment, the method may include transmitting the changed electronic communication to the electronic identifier associated with the user or transmitting the changed electronic communication to a second electronic identifier associated with a second user. The compositional change may be derived from representations of previous content and context from a plurality of user profiles which are like at least one of the communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles, the processor may use at least one of a machine learning model, deep learning model, or other statistical learning algorithm for creating the compositional change. The compositional change may be an auto-generated textual completion. The auto-generated textual completion may be a phrasal completion. The processor may generate the compositional change by optimizing generated language of a user as determined by the processor from a user communication attribute. The processor may generate the compositional change by replicating a communication style of the user as determined by the processor from the user communication attribute. The partial electronic communication may include a communication goal, and the processor may generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion. The processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find a most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automatically generated template generated by the processor. The processor may select the communication template by using a machine learning model to score the plurality of communication templates based at least in part on the communication content, communication context, first user communication attribute, or second user communication attribute.

[0051] In an embodiment, a method of electronic communication assistance may include receiving a partial electronic communication at an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the partial electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user and a second electronic identifier associated with a second user; extracting a communication context from the partial electronic communication;encoding the partial electronic communication for processing creating an encoded partial electronic communication; retrieving from a communication profile database a communication profile for the second user using the second electronic identifier associated with the second user, wherein the communication profile comprises a second user communication attribute; processing the encoded partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using at least one of the communication context or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0052] In an embodiment, the method may include transmitting the changed electronic communication to the electronic identifier associated with the user or transmitting the changed electronic communication to a second electronic identifier associated with a second user. The compositional change may be derived from representations of previous content and context from a plurality of user profiles stored in a communication profile database which are like at least one of the communication profiles of the first or second user. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning model, deep learning model, or other statistical learning algorithm for creating the compositional change. The compositional change may be an auto-generated textual completion. The auto-generated textual completion may be a phrasal completion. The processor may generate the compositional change by optimizing generated language as determined by the processor from the second user communication attribute. The processor may generate the compositional change by replicating a communication style of the user. The partial electronic communication may include a communication goal, and the processor may generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion. The processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find a most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automatically generated template generated by the processor. The processor may select the communicationtemplate by using a machine learning model to score the plurality of communication templates based at least in part on the communication content, communication context, first user communication attribute, or second user communication attribute.

[0053] In an embodiment, a method of electronic communication assistance may include: receiving a partial electronic communication at an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the partial electronic communication comprising a communication content and comprising or associated with the electronic identifier and a second electronic identifier associated with a second user;_retrieving a second communication profile, wherein the second communication profile comprises a second user communication attribute; processing the partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0054] In an embodiment, the method may include transmitting the changed electronic communication to the electronic identifier associated with the user or transmitting the changed electronic communication to a second electronic identifier associated with a second user. The compositional change may be derived from representations of previous content and context from a plurality of user profiles which are like at least one of the communication profiles._The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles._The processor may use at least one of a machine learning model, deep learning model, or other statistical learning algorithm for creating the compositional change..The compositional change may be an auto-generated textual completion. The auto-generated textual completion may be a phrasal completion. The processor may generate the compositional change by optimizing generated language as determined by the processor from the second user communication attribute. The processor may generate the compositional change by replicating a communication style of the user as determined by the processor from the user communication attribute. The partial electronic communication may include a communication goal, and the processor may generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structuralelements for user completion. The processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find the most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automatically- generated template generated by the processor. The processor may select the communication template by using a machine learning model to score the plurality of communication templates based at least in part on the communication content, communication context, first-user communication attribute, or second-user communication attribute.

[0055] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving a partial electronic communication at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the partial electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user; extracting a communication context from the partial electronic communication; encoding the partial electronic communication for processing creating an encoded partial electronic communication; retrieving from a communication profile database a first communication profile for the first user using the first electronic identifier associated with the first user, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user using the second electronic identifier associated with the second user, wherein the second communication profile comprises a second user communication attribute; processing the encoded partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using at least one of the communication context, the first user communication attribute, or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0056] In an embodiment, the system computer may be enabled to perform operations, including transmitting the changed electronic communication to the electronic identifier associated with the user and / or transmitting the changed electronic communication to a second electronic identifier associated with a second user. The compositional change may be derivedfrom representations of previous content and context from a plurality of user profiles stored in the communication profile database, which are like at least one of the first communication profiles or the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use a machine learning model, deep learning model, or other statistical learning algorithm for creating the compositional change. The compositional change may be an auto-generated textual completion. The auto-generated textual completion may be a phrasal completion. The processor may generate the compositional change by optimizing the generated language as determined by the processor from the second user communication attribute. The processor may generate the compositional change by replicating the communication style of the first user as determined by the processor from the first user communication attribute. The partial electronic communication may include a communication goal, and the processor may generate the compositional change by optimizing the impact and effectiveness of the generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of the prepared text or placeholder locations for defining structural elements for user completion. The processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find the most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automatically-generated template generated by the processor. The processor may select the communication template by using a machine learning model to score the plurality of communication templates based at least in part on the communication content, communication context, first user communication attribute, or second user communication attribute.

[0057] In an embodiment, a system may include: a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving a partial electronic communication at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the partial electronic communication comprising a communication content and comprising or associated with the first electronic identifier and a second electronic identifier associated with a second user; retrieving a first communication profile, wherein the first communication profile comprises a first user communication attribute;retrieving a second communication profile, wherein the second communication profile comprises a second user communication attribute; processing the partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using at least one of the first user communication attribute or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0058] In an embodiment, the system computer may be enabled to perform operations, including transmitting the changed electronic communication to the first electronic identifier associated with the first user or transmitting the changed electronic communication to a second electronic identifier associated with a second user. The compositional change may be derived from representations of previous content and context from a plurality of user profiles which are like at least one of the first communication profile or the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use a machine learning model, deep learning model, or other statistical learning algorithm for creating the compositional change. The compositional change may be an auto-generated textual completion. The auto-generated textual completion may be a phrasal completion. The processor may generate the compositional change by optimizing the generated language as determined by the processor from the second user communication attribute. The processor may generate the compositional change by replicating the communication style of the first user as determined by the processor from the first user communication attribute. The partial electronic communication may include a communication goal, and the processor may generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of the prepared text or placeholder locations for defining structural elements for user completion. The processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find a most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automatically generated template generated by the processor. The processor may select the communication template by using a machine learning model to score the plurality of communication templatesbased at least in part on the communication content, communication context, first user communication attribute, or second user communication attribute.

[0059] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving a partial electronic communication at an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the partial electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user; extracting a communication context from the partial electronic communication; encoding the partial electronic communication for processing creating an encoded partial electronic communication; retrieving from a communication profile database a communication profile for the user using the electronic identifier associated with the user, wherein the communication profile comprises a user communication attribute; processing the encoded partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using at least one of the communication context or the user communication attribute; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0060] In an embodiment, the system computer may be enabled to perform operations including transmitting the changed electronic communication to the electronic identifier associated with the user or transmitting the changed electronic communication to a second electronic identifier associated with a second user. The compositional change may be derived from representations of previous content and context from a plurality of user profiles stored in the communication profile database which are like the communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning model, deep learning model, or other statistical learning algorithm for creating the compositional change. The compositional change may be an auto-generated textual completion. The auto-generated textual completion may be a phrasal completion. The processor may generate the compositional change by optimizing generated language of a user as determined by the processor from a user communication attribute. The processor may generate the compositional change by replicating a communication style of the user as determined by the processor from the user communication attribute. The partial electroniccommunication further comprises a communication goal, and the processor may generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion. The processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find a most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automatically generated template generated by the processor. The processor may select the communication template by using a machine learning model to score the plurality of communication templates based at least in part on the communication content, communication context, first user communication attribute, or second user communication attribute.

[0061] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving a partial electronic communication at an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the partial electronic communication comprising a communication content and comprising or associated with the electronic identifier; retrieving a communication profile, wherein the communication profile comprises a user communication attribute; processing the partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using the user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0062] In an embodiment, the system computer may be enabled to perform operations including transmitting the changed electronic communication to the electronic identifier associated with the user or transmitting the changed electronic communication to a second electronic identifier associated with a second user. The compositional change may be derived from representations of previous content and context from a plurality of user profiles which are like at least one of the communication profiles. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning model, deep learning model,or other statistical learning algorithm for creating the compositional change. The compositional change may be an auto-generated textual completion. The auto-generated textual completion may be a phrasal completion. The processor may generate the compositional change by optimizing generated language of a user as determined by the processor from a user communication attribute. The processor may generate the compositional change by replicating a communication style of the user as determined by the processor from the user communication attribute. The partial electronic communication further comprises a communication goal, and the processor may generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion. The processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find a most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automatically generated template generated by the processor. The processor may select the communication template by using a machine learning model to score the plurality of communication templates based at least in part on the communication content, communication context, first user communication attribute, or second user communication attribute.

[0063] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving a partial electronic communication at an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the partial electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user and a second electronic identifier associated with a second user; extracting a communication context from the partial electronic communication; encoding the partial electronic communication for processing creating an encoded partial electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier associated with the second user, wherein the second communication profile comprises a second user communication attribute; processing the encoded partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communicationusing at least one of the communication context or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0064] In an embodiment, the system computer may be enabled to perform operations including transmitting the changed electronic communication to the electronic identifier associated with the user or transmitting the changed electronic communication to a second electronic identifier associated with a second user. The compositional change may be derived from representations of previous content and context from a plurality of user profiles stored in a communication profile database which are like the communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning model, deep learning model, or other statistical learning algorithm for creating the compositional change. The compositional change may be an auto-generated textual completion. The auto-generated textual completion may be a phrasal completion. The processor may generate the compositional change by optimizing generated language as determined by the processor from the second user communication attribute. The processor may generate the compositional change by replicating a communication style of the user as determined by the processor from the user communication attribute. The partial electronic communication further comprises a communication goal, and the processor may generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion. The processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find a most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automatically generated template generated by the processor. The processor may select the communication template by using a machine learning model to score the plurality of communication templates based at least in part on the communication content, communication context, first user communication attribute, or second user communication attribute.

[0065] In an embodiment, a system of electronic communication assistance may perform the following: receiving a partial electronic communication at an artificial intelligence assistantcomputing facility from an electronic identifier associated with a user, the partial electronic communication comprising a communication content and comprising or associated with the electronic identifier and a second electronic identifier associated with a second user; retrieving a second communication profile, wherein the second communication profile comprises a second user communication attribute; processing the partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

[0066] In an embodiment, the system computer may be enabled to perform operations including transmitting the changed electronic communication to the electronic identifier associated with the user or transmitting the changed electronic communication to a second electronic identifier associated with a second user. The compositional change may be derived from representations of previous content and context from a plurality of user profiles which are like at least one of the communication profile or the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning model, deep learning model, or other statistical learning algorithm for creating the compositional change. The compositional change may be an auto-generated textual completion. The auto-generated textual completion may be a phrasal completion. The processor may generate the compositional change by optimizing generated language as determined by the processor from the second user communication attribute. The processor may generate the compositional change by replicating a communication style of the user as determined by the processor from the user communication attribute. The partial electronic communication further comprises a communication goal, and the processor may generate the compositional change by optimizing for the impact and effectiveness of generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion. The processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find a most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automaticallygenerated template generated by the processor. The processor may select the communication template by using a machine learning model to score the plurality of communication templates based at least in part on the communication content, communication context, first user communication attribute, or second user communication attribute.

[0067] In an embodiment, a method of electronic communication assistance may include intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a first communication profile for the first user using the first user identifier, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute; processing the encoded electronic communication with a processor to generate a compositional change for the communication content of the electronic communication using at least one of the communication context, the first user communication attribute, or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0068] In an embodiment, the method may include transmitting the changed electronic communication to the first electronic identifier associated with the first user or transmitting the changed electronic communication to the second electronic identifier associated with the second user. The compositional change may be derived from representations of previous content and context from a plurality of user profiles stored in the communication profile database which are like at least one of the first communication profile or the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning language model or a statistical algorithm for creatingthe compositional change. The processor may generate the compositional change by optimizing generated language of the second user as determined by the processor from the second user communication attribute. The processor may generate the compositional change by replicating a communication style of the first user as determined by the processor from the first user communication attribute. The electronic communication further comprises a communication goal, and the processor may generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal.

[0069] In an embodiment, a method of electronic communication assistance may include: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a first communication profile for the first user using the first user identifier, wherein the first communication profile comprises a first user communication attribute; retrieving a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute; processing the electronic communication with a processor to generate a compositional change for the communication content of the electronic communication using at least one of the first user communication attribute or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0070] In an embodiment, a method of electronic communication assistance may include: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from an electronic identifier associated with a user, the electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a communication profile for the user using the user identifier, wherein the communication profile comprises a user communication attribute; processing the encoded electronic communication with a processorto generate a compositional change for the communication content of the electronic communication using at least one of the communication context or the user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0071] In an embodiment, a method of electronic communication assistance may include: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from an electronic identifier associated with a user, the electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user; retrieving a communication profile for the user using the user identifier, wherein the communication profile comprises a user communication attribute; processing the electronic communication with a processor to generate a compositional change for the communication content of the electronic communication using the user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0072] In an embodiment, a method of electronic communication assistance may include: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a communication profile for the second user using the second electronic identifier, wherein the communication profile comprises a second user communication attribute; processing the encoded electronic communication with a processor to generate a compositional change for the communication content of the electronic communication using at least one of the communication context or the second user communication attribute to generate the compositional change; and generating a changedelectronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0073] In an embodiment, a method of electronic communication assistance may include: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving from a communication profile database a communication profile for the second user using the second electronic identifier, wherein the communication profile comprises a second user communication attribute; processing the electronic communication with a processor to generate a compositional change for the communication content of the electronic communication using the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0074] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a first communication profile for the first user using the first user identifier, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute; processing the encoded electronic communication with a processor to generate a compositional change for the communicationcontent of the electronic communication using at least one of the communication context, the first user communication attribute, or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0075] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a first communication profile for the first user using the first user identifier, wherein the first communication profile comprises a first user communication attribute; retrieving a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute; processing the electronic communication with a processor to generate a compositional change for the communication content of the electronic communication using at least one of the first user communication attribute or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0076] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from an electronic identifier associated with a user, the electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a communication profile for the user using the user identifier,wherein the communication profile comprises a user communication attribute; processing the encoded electronic communication with a processor to generate a compositional change for the communication content of the electronic communication using at least one of the communication context or the user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0077] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from an electronic identifier associated with a user, the electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user; retrieving a communication profile for the user using the user identifier, wherein the communication profile comprises a user communication attribute; processing the electronic communication with a processor to generate a compositional change for the communication content of the electronic communication using the user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0078] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a communication profile for the second user using the second electronic identifier, wherein the communication profile comprises a second user communicationattribute; processing the encoded electronic communication with a processor to generate a compositional change for the communication content of the electronic communication using at least one of the communication context or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0079] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted to second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving from a communication profile database a communication profile for the second user using the second electronic identifier, wherein the communication profile comprises a second user communication attribute; processing the electronic communication with a processor to generate a compositional change for the communication content of the electronic communication using the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

[0080] In an embodiment, a method of electronic communication assistance may include: receiving electronic communication information as an electronic communication is composed at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the electronic communication information comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user; extracting a communication context from the electronic communication information; retrieving from a communication profile database a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user using the second electronic identifier, wherein thesecond communication profile comprises a second user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to the electronic communication information using at least one of the communication context, the first user communication attribute, or the second user communication attribute to generate the compositional recommendation; and transmitting the compositional recommendation to the first electronic identifier.

[0081] In an embodiment, the method may further include intercepting an electronic communication from the first electronic identifier, the electronic communication relating to the electronic communication information and the compositional recommendation; encoding the electronic communication to create an encoded electronic communication; processing the encoded electronic communication with the processor to extract an additional communication attribute from the electronic communication; and updating the first communication profile for the first user based on the additional communication attribute. The compositional recommendation may be derived from representations of previous content and context from a plurality of user profiles stored in the communication profile database which are like at least one of the first communication profile or the second communication profile. The processor may calculate weighted scores for predicted reaction outcome and expected reaction outcome, and derives the compositional recommendation based on a comparison of the calculated weighted scores. The processor may use at least one of a machine learning mode, deep learning model, or statistical learning algorithm for creating the compositional recommendation. The processor may generate the compositional recommendation by evaluating a communication style of the first user as determined by the processor from the first user communication attribute. The processor may generate the compositional recommendation further using a communication recommendation template selected from a plurality of communication recommendation templates. The processor may select the communication recommendation template using at least one of a machine learning model, deep learning model, or statistical learning model to find a most effective communication recommendation template based at least in part on the electronic communication information. The plurality of communication recommendation templates may include at least one automatically generated template generated by the processor. The processor may select the communication recommendation template by using a machine learning model to score the plurality of communication recommendation templates based at least in part on the electronic communication information.The electronic communication information may include a communication goal. The communication goal may be used in generating the compositional recommendation.

[0082] In an embodiment, a method of electronic communication assistance may include: receiving electronic communication information as an electronic communication is composed at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the electronic communication information comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user; retrieving a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to the electronic communication information using at least one of the first user communication attribute or the second user communication attribute to generate the compositional recommendation; and transmitting the compositional recommendation to the first electronic identifier.

[0083] In an embodiment, a method of electronic communication assistance may include: receiving electronic communication information as an electronic communication is composed at an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the electronic communication information comprising a communication content and comprising or associated with the electronic identifier associated with the user; extracting a communication context from the electronic communication information; retrieving from a communication profile database a communication profile for the user using the electronic identifier, wherein the communication profile comprises a user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to the electronic communication information using at least one of the communication context or the user communication attribute to generate the compositional recommendation; and transmitting the compositional recommendation to the electronic identifier.

[0084] In an embodiment, a method of electronic communication assistance may include: receiving electronic communication information as an electronic communication is composedat an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the electronic communication information comprising a communication content and comprising or associated with the electronic identifier associated with the user; retrieving a communication profile for the user using the electronic identifier, wherein the communication profile comprises a user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to the electronic communication information using the user communication attribute; and transmitting the compositional recommendation to the electronic identifier.

[0085] In an embodiment, a method of electronic communication assistance may include: receiving electronic communication information as an electronic communication is composed at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the electronic communication information comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user; extracting a communication context from the electronic communication information; retrieving a communication profile for the second user using the second electronic identifier, wherein the communication profile comprises a second user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to the electronic communication information using at least one of the communication context or the second user communication attribute to generate the compositional recommendation; and transmitting the compositional recommendation to the first electronic identifier.

[0086] In an embodiment, a method of electronic communication assistance may include receiving electronic communication information as an electronic communication is composed at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the electronic communication information comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user; retrieving a communication profile for the second user using the second electronic identifier, wherein the communication profile comprises a second user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to theelectronic communication information using the second user communication attribute to generate the compositional recommendation; and transmitting the compositional recommendation to the first electronic identifier.

[0087] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving electronic communication information as an electronic communication is composed at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the electronic communication information comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user; extracting a communication context from the electronic communication information; retrieving from a communication profile database a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to the electronic communication information using at least one of the communication context, the first user communication attribute, or the second user communication attribute to generate the compositional recommendation; and transmitting the compositional recommendation to the first electronic identifier.

[0088] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving electronic communication information as an electronic communication is composed at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the electronic communication information comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user; retrieving a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving a second communication profile for the second user using the second electronic identifier, wherein the secondcommunication profile comprises a second user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to the electronic communication information using at least one of the first user communication attribute or the second user communication attribute to generate the compositional recommendation; and transmitting the compositional recommendation to the first electronic identifier.

[0089] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving electronic communication information as an electronic communication is composed at an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the electronic communication information comprising a communication content and comprising or associated with the electronic identifier; extracting a communication context from the electronic communication information; retrieving from a communication profile database a communication profile for the user using the electronic identifier, wherein the communication profile comprises a user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to the electronic communication information using at least one of the communication context or the user communication attribute to generate the compositional recommendation; and transmitting the compositional recommendation to the electronic identifier.

[0090] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving electronic communication information as an electronic communication is composed at an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the electronic communication information comprising a communication content and comprising or associated with the electronic identifier associated with the user; retrieving a communication profile for the user using the electronic identifier, wherein the communication profile comprises a user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to the electronic communication information using the user communicationattribute to generate the compositional recommendation; and transmitting the compositional recommendation to the electronic identifier.

[0091] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving electronic communication information as an electronic communication is composed at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the electronic communication information comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user; extracting a communication context from the electronic communication information; retrieving from a communication profile database a communication profile for the second user using the second electronic identifier, wherein the communication profile comprises a second user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to the electronic communication information using at least one of the communication context or the second user communication attribute to generate the compositional recommendation; and transmitting the compositional recommendation to the first electronic identifier.

[0092] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving electronic communication information as an electronic communication is composed at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the electronic communication information comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user; retrieving a communication profile for the second user using the second electronic identifier, wherein the communication profile comprises a second user communication attribute; processing the electronic communication information with a processor to generate a compositional recommendation for developing a communication content related to the electronic communication information using the second user communication attribute to generate the compositional recommendation; and transmitting the compositional recommendation to the first electronic identifier.

[0093] In an embodiment, a method of electronic communication assistance may include: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and the second electronic identifier associated with the second user; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, wherein the processor uses at least one of the communication content, the first user communication attribute, or the second user communication attribute to process the encoded electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0094] In an embodiment, the processor may generate the modified electronic communication by removing or replacing language from the electronic communication based at least in part on the second user communication attribute. The removed or replaced language may be offensive or abusive language. The processor may generate the modified electronic communication by summarizing language from the electronic communication based at least in part on the second user communication attribute. The processor may generate the modified electronic communication by reformatting the electronic communication based at least in part on the second user communication attribute. The processor may generate the modified electronic communication by recomposing language from the electronic communication based at least in part on the second user communication attribute. The processor may generate the modified electronic communication by incorporating explanatory text associated with phases based at least in part on the second user communication attribute. The processor may generate the modified electronic communication derived at least in part from representations of previous electronic communications from a plurality of user profiles stored in the communication profiledatabase that are like at least one of the first communication profile or the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning model, deep learning model, or statistical learning model for generating the modified electronic communication. The modified electronic communication may be used to generate at least one of an updated first communication profile or an updated second communication profile. Prior to transmitting the modified electronic communication to the second electronic identifier, receiving a request by the second user may be received to provide the modified electronic communication.

[0095] In an embodiment, a method of electronic communication assistance may include: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, wherein the processor uses at least one of the communication content, the first user communication attribute, or the second user communication attribute to process the electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0096] In an embodiment, a method of electronic communication assistance may include: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a first communication profile for the first user using the firstelectronic identifier, wherein the first communication profile comprises a first user communication attribute; processing the encoded electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, wherein the processor uses at least one of the communication content or the first user communication attribute to process the encoded electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0097] In an embodiment, a method of electronic communication assistance may include: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; processing the electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, wherein the processor uses at least one of the communication content or the first user communication attribute to process the electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0098] In an embodiment, a method of electronic communication assistance may include: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, wherein the processor uses at least one of the communication content or the second user communicationattribute to process the encoded electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0099] In an embodiment, a method of electronic communication assistance may include: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, wherein the processor uses at least one of the communication content or the second user communication attribute to process the electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0100] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and the second electronic identifier associated with the second user; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, wherein the processor uses at least one of the communication content, the first user communicationattribute, or the second user communication attribute to process the encoded electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0101] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, wherein the processor uses at least one of the communication content, the first user communication attribute, or the second user communication attribute to process the electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0102] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; processing the encoded electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, whereinthe processor uses at least one of the communication content or the first user communication attribute to process the encoded electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0103] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; processing the electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, wherein the processor uses at least one of the communication content or the first user communication attribute to process the electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0104] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, wherein the processor uses at least one of the communication content or the second user communication attribute to process theencoded electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0105] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: intercepting a electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the electronic communication with a processor to generate a modified electronic communication that is a modified version of the electronic communication, wherein the processor uses at least one of the communication content or the second user communication attribute to process the electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0106] In an embodiment, a method of electronic communication assistance may include: receiving an audio-visual electronic communication at an artificial intelligence assistant computing platform from a first user, the first user identified by a first electronic identifier, the audio-visual electronic communication comprising an audio communication content and a video communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user, wherein the intended recipient of the audio- visual electronic communication is the second user; encoding the audio communication content for processing creating an encoded audio communication; extracting an audio communication information from the encoded audio communication content; encoding the video communication content for processing creating an encoded video communication; extracting a video communication information from the encoded video communication content; processing the audio-visual electronic communication with a processor to generate an audio-visual compositional change for the communication content of the audio-visual electronic communication using at least one of the audio communication content, the video communication content, audio communication information, video communication information, a sensor input from a wearable user device, a first usercommunication attribute retrieved from a communication profile for the first user using the first electronic identifier or a second user communication attribute retrieved from a communication profile for the second user using the second electronic identifier to generate the compositional change; generating a changed electronic communication from the audio- visual electronic communication and the compositional change; and providing the changed electronic communication.

[0107] In an embodiment, the compositional change may be derived from representations of previous content and context from a plurality of user profiles stored in a communication profile database which are like at least one of the first communication profile or the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from a plurality of user profiles. The processor may use at least one of a machine learning model, deep learning model, or statistical learning algorithm for creating the compositional change. The processor may generate the compositional change by optimizing generated language of the first user as determined by the processor from the first user communication attribute. The processor may generate the compositional change by replicating a communication style of the first user as determined by the processor from the first user communication attribute. The audio-visual electronic communication may include a communication goal, and the processor generates the compositional change by optimizing in respect of the communication goal. The method may further include providing a textual representation of the audio- visual electronic communication within a graphical user interface on a screen of a computing device worn by the first user and providing the changed electronic communication on the screen.

[0108] In an embodiment, a method of electronic communication assistance may include: receiving an audio-visual electronic communication at an artificial intelligence assistant computing platform from a first user, the audio-visual electronic communication comprising an audio communication content and a video communication content, wherein the intended recipient of the audio- visual electronic communication is a second user; extracting an audio communication information from the audio communication content; extracting a video communication information from the video communication content; processing the audiovisual electronic communication with a processor to generate an audio-visual compositional change for the communication content of the audio-visual electronic communication using at least one of the audio communication content, the video communication content, audio communication information, video communication information, a sensor input from a wearableuser device, a first user communication attribute retrieved from a communication profile for the first user or a second user communication attribute retrieved from a communication profile for the second user to generate the compositional change; generating a changed electronic communication from the audio-visual electronic communication and the compositional change; and providing the changed electronic communication.

[0109] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving an audio-visual electronic communication at an artificial intelligence assistant computing platform from a first user, the first user identified by a first electronic identifier, the audio-visual electronic communication comprising an audio communication content and a video communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user, wherein the intended recipient of the audio- visual electronic communication is the second user; encoding the audio communication content for processing creating an encoded audio communication; extracting an audio communication information from the encoded audio communication content; encoding the video communication content for processing creating an encoded video communication; extracting a video communication information from the encoded video communication content; processing the audio-visual electronic communication with a processor to generate an audio-visual compositional change for the communication content of the audio-visual electronic communication using at least one of the audio communication content, the video communication content, audio communication information, video communication information, a sensor input from a wearable user device, a first user communication attribute retrieved from a communication profile for the first user using the first electronic identifier or a second user communication attribute retrieved from a communication profile for the second user using the second electronic identifier to generate the compositional change; generating a changed electronic communication from the audio- visual electronic communication and the compositional change; and providing the changed electronic communication.

[0110] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: receiving an audio-visual electronic communication at an artificial intelligence assistant computing platform from a firstuser, the audio- visual electronic communication comprising an audio communication content and a video communication content, wherein the intended recipient of the audio-visual electronic communication is a second user; extracting an audio communication information from the audio communication content; extracting a video communication information from the video communication content; processing the audio-visual electronic communication with a processor to generate an audio- visual compositional change for the communication content of the audio-visual electronic communication using at least one of the audio communication content, the video communication content, audio communication information, video communication information, a sensor input from a wearable user device, a first user communication attribute retrieved from a communication profile for the first user or a second user communication attribute retrieved from a communication profile for the second user to generate the compositional change; generating a changed electronic communication from the audio-visual electronic communication and the compositional change; and providing the changed electronic communication.

[0111] In an embodiment, a method of electronic communication assistance may include: obtaining a copy of an in-transit audio-visual electronic communication at an artificial intelligence assistant computing platform from a first user, the first user identified by a first electronic identifier, the audio-visual electronic communication comprising an audio communication content and a video communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user, wherein the intended recipient of the audio-visual electronic communication is the second user; encoding the audio communication content for processing creating an encoded audio communication; extracting an audio communication context from the encoded audio communication content; encoding the video communication content for processing creating an encoded video communication; extracting a video communication context from the encoded video communication content; processing the audio- visual electronic communication with a processor to generate a communication feedback related to the audiovisual electronic communication using at least one of the audio communication content, the video communication content, audio communication context, video communication context, a sensor input from a wearable device, a first user communication attribute retrieved from a communication profile for the first user using the first electronic identifier or a second user communication attribute retrieved from a communication profile for the second user using the second electronic identifier to generate the communication feedback, wherein thecommunication feedback provides communication information directed to the first user concerning the audio-visual communication; and providing the audio communication content and the communication feedback. In an embodiment, the communication feedback may be derived from representations of previous content and context from a plurality of user profiles stored in a communication profile database which are like at least one of the first communication profile or the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from a plurality of user profiles. The processor may use at least one of a machine learning model, deep learning model, or statistical learning algorithm for creating the communication feedback. The processor may generate the communication feedback by optimizing generated language of the first user as determined by the processor from the first user communication attribute. The processor may generate the communication feedback by optimizing a next electronic communication with respect to the second user communication attribute. The audio-visual electronic communication may include a communication goal, and the processor may generate the communication feedback by optimizing in respect of the communication goal. The communication feedback may provide communication information directed to the first user concerning the audio-visual communication related to generating a next audio-visual communication. A textual representation of the audio-visual electronic communication may be provided within a graphical user interface on a screen of a wearable computing device worn by the first user and providing the audio communication content and the communication feedback on the screen.

[0112] In an embodiment, a method of electronic communication assistance may include: obtaining a copy of an in-transit audio-visual electronic communication at an artificial intelligence assistant computing platform from a first user, the audio-visual electronic communication comprising an audio communication content and a video communication content, wherein the intended recipient of the audio-visual electronic communication is a second user; extracting an audio communication context from the audio communication content; extracting a video communication context from the video communication content; processing the audio-visual electronic communication with a processor to generate a communication feedback related to the audio-visual electronic communication using at least one of the audio communication content, the video communication content, audio communication context, video communication context, a sensor input from a wearable device, a first user communication attribute retrieved from a communication profile for the first useror a second user communication attribute retrieved from a communication profile for the second user to generate the communication feedback, wherein the communication feedback provides communication information directed to the first user concerning the audio-visual communication; and providing the audio communication content and the communication feedback.

[0113] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: obtaining a copy of an in-transit audio-visual electronic communication at an artificial intelligence assistant computing platform from a first user, the first user identified by a first electronic identifier, the audiovisual electronic communication comprising an audio communication content and a video communication content and comprising or associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user, wherein the intended recipient of the audio- visual electronic communication is the second user; encoding the audio communication content for processing creating an encoded audio communication; extracting an audio communication context from the encoded audio communication content; encoding the video communication content for processing creating an encoded video communication; extracting a video communication context from the encoded video communication content; processing the audio-visual electronic communication with a processor to generate a communication feedback related to the audio-visual electronic communication using at least one of the audio communication content, the video communication content, audio communication context, video communication context, a sensor input from a wearable device, a first user communication attribute retrieved from a communication profile for the first user using the first electronic identifier or a second user communication attribute retrieved from a communication profile for the second user using the second electronic identifier to generate the communication feedback, wherein the communication feedback provides communication information directed to the first user concerning the audio-visual communication; and providing the audio communication content and the communication feedback.

[0114] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: obtaining a copy of an in-transit audio-visual electronic communication at an artificial intelligence assistant computingplatform from a first user, the audio-visual electronic communication comprising an audio communication content and a video communication content, wherein the intended recipient of the audio- visual electronic communication is a second user; extracting an audio communication context from the audio communication content; extracting a video communication context from the video communication content; processing the audio-visual electronic communication with a processor to generate a communication feedback related to the audio-visual electronic communication using at least one of the audio communication content, the video communication content, audio communication context, video communication context, a sensor input from a wearable device, a first user communication attribute retrieved from a communication profile for the first user or a second user communication attribute retrieved from a communication profile for the second user to generate the communication feedback, wherein the communication feedback provides communication information directed to the first user concerning the audio-visual communication; and providing the audio communication content and the communication feedback.

[0115] In an embodiment, a method of electronic communication assistance may include: obtaining a copy of an in-transit electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and the second electronic identifier associated with the second user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a first communication profile for the first user utilizing the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user utilizing the second electronic identifier, wherein the second communication profile comprises a second user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user, wherein the electronic reaction data is generated in response to the second user receiving the electronic communication; encoding the electronic reaction data for processing creating an encoded electronic reaction data; processing the encoded electronic communication and the encoded electronic reaction data to extract a reaction context with a processor to generate a responsefeedback communication using at least one of the communication content, the extracted communication context, the first user communication attribute, or the second user communication attribute; and sending the response feedback communication to the first electronic identifier.

[0116] In an embodiment, the response feedback communication may be derived at least in part from representations of previous electronic communications from a plurality of user profiles stored in the communication profile database which are like at least one of the first communication profile or the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may utilize at least one of a machine learning language model or a statistical algorithm for creating the response feedback communication. The electronic communication may include a communication goal, and the processor generates the response feedback communication at least in part by computing and utilizing a reaction difference attribute derived from determined differences between the communication goal and the electronic reaction data. The electronic reaction data may include at least one of location information for the second user, response time between the second user receiving the electronic communication and reading the electronic communication, or an action taken by the second user. The reaction context may include an emotional state of the second user. The electronic reaction data may be used to generate at least one of an updated first communication profile or an updated second communication profile. At least one of the updated first communication profile or the updated second communication profile may be used to predict a most likely reaction outcome in a second electronic communication. The electronic reaction data may be used to train a machine learning model that is configured to at least one of generate a communication content or modify a communication content.

[0117] In an embodiment, a method of electronic communication assistance may include: obtaining a copy of an in-transit electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a first communication profile for the first user utilizing the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving a second communication profile for the second user utilizing the second electronicidentifier, wherein the second communication profile comprises a second user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user, wherein the electronic reaction data is generated in response to the second user receiving the electronic communication; processing the electronic communication and the electronic reaction data to extract a reaction context with a processor to generate a response feedback communication using at least one of the communication content, the first user communication attribute, or the second user communication attribute; and sending the response feedback communication to the first electronic identifier.

[0118] In an embodiment, a method of electronic communication assistance may include: obtaining a copy of an in-transit electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from an electronic identifier associated with a user, the electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a communication profile for the user utilizing the electronic identifier, wherein the communication profile comprises a user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user, wherein the electronic reaction data is generated in response to the second user receiving the electronic communication; encoding the electronic reaction data for processing creating an encoded electronic reaction data; processing the encoded electronic communication and the encoded electronic reaction data to extract a reaction context with a processor to generate a response feedback communication using at least one of the communication content, the extracted communication context, or the user communication attribute; and sending the response feedback communication to the electronic identifier.

[0119] In an embodiment, a method of electronic communication assistance may include: obtaining a copy of an in-transit electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from an electronic identifier associated with a user, the electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user; retrieving a communication profile for the user utilizing the electronic identifier, wherein the communication profile comprises a user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user, wherein the electronic reaction datais generated in response to the second user receiving the electronic communication; processing the electronic communication and the electronic reaction data to extract a reaction context with a processor to generate a response feedback communication using at least one of the communication content or the user communication attribute; and sending the response feedback communication to the electronic identifier.

[0120] In an embodiment, a method of electronic communication assistance may include: obtaining a copy of an in-transit electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a communication profile for the second user utilizing the second electronic identifier, wherein the communication profile comprises a second user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user, wherein the electronic reaction data is generated in response to the second user receiving the electronic communication; encoding the electronic reaction data for processing creating an encoded electronic reaction data; processing the encoded electronic communication and the encoded electronic reaction data to extract a reaction context with a processor to generate a response feedback communication using at least one of the communication content, the extracted communication context, or the second user communication attribute; and sending the response feedback communication to the first electronic identifier.

[0121] In an embodiment, a method of electronic communication assistance may include: obtaining a copy of an in-transit electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a communication profile for the second user utilizing the second electronic identifier, wherein the communication profile comprises a second user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user,wherein the electronic reaction data is generated in response to the second user receiving the electronic communication; processing the electronic communication and the electronic reaction data to extract a reaction context with a processor to generate a response feedback communication using at least one of the communication content or the second user communication attribute; and sending the response feedback communication to the first electronic identifier.

[0122] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: obtaining a copy of an in-transit electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user and the second electronic identifier associated with the second user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a first communication profile for the first user utilizing the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user utilizing the second electronic identifier, wherein the second communication profile comprises a second user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user, wherein the electronic reaction data is generated in response to the second user receiving the electronic communication; encoding the electronic reaction data for processing creating an encoded electronic reaction data; processing the encoded electronic communication and the encoded electronic reaction data to extract a reaction context with a processor to generate a response feedback communication using at least one of the communication content, the extracted communication context, the first user communication attribute, or the second user communication attribute; and sending the response feedback communication to the first electronic identifier.

[0123] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: obtaining a copy of an in-transitelectronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a first communication profile for the first user utilizing the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving a second communication profile for the second user utilizing the second electronic identifier, wherein the second communication profile comprises a second user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user, wherein the electronic reaction data is generated in response to the second user receiving the electronic communication; processing the electronic communication and the electronic reaction data to extract a reaction context with a processor to generate a response feedback communication using at least one of the communication content, the first user communication attribute, or the second user communication attribute; and sending the response feedback communication to the first electronic identifier.

[0124] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: obtaining a copy of an in-transit electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from an electronic identifier associated with a user, the electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a communication profile for the user utilizing the electronic identifier, wherein the communication profile comprises a user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user, wherein the electronic reaction data is generated in response to the second user receiving the electronic communication; encoding the electronic reaction data for processing creating an encoded electronic reaction data; processing the encoded electronic communication and the encoded electronic reaction data to extract a reaction context with a processor to generate a response feedback communication using at least one of the communication content, the extractedcommunication context or the user communication attribute; and sending the response feedback communication to the electronic identifier.

[0125] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: obtaining a copy of an in-transit electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from an electronic identifier associated with a user, the electronic communication comprising a communication content and comprising or associated with the electronic identifier associated with the user; retrieving from a communication profile database a communication profile for the user utilizing the electronic identifier, wherein the communication profile comprises a user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user, wherein the electronic reaction data is generated in response to the second user receiving the electronic communication; processing the electronic communication and the electronic reaction data to extract a reaction context with a processor to generate a response feedback communication using at least one of the communication content, the extracted communication context or the user communication attribute; and sending the response feedback communication to the electronic identifier.

[0126] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: obtaining a copy of an in-transit electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a communication profile for the second user utilizing the second electronic identifier, wherein the communication profile comprises a second user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user, wherein the electronic reaction data is generated in response to the second user receiving the electronic communication; encoding the electronic reaction data for processing creating anencoded electronic reaction data; processing the encoded electronic communication and the encoded electronic reaction data to extract a reaction context with a processor to generate a response feedback communication using at least one of the communication content, the extracted communication context or the second user communication attribute; and sending the response feedback communication to the first electronic identifier.

[0127] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: obtaining a copy of an in-transit electronic communication at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a communication profile for the second user utilizing the second electronic identifier, wherein the communication profile comprises a second user communication attribute; receiving electronic reaction data from a second electronic identifier of the second user, wherein the electronic reaction data is generated in response to the second user receiving the electronic communication; processing the electronic communication and the electronic reaction data to extract a reaction context with a processor to generate a response feedback communication using at least one of the communication content or the second user communication attribute; and sending the response feedback communication to the first electronic identifier.

[0128] In an embodiment, a computer-implemented method for modifying an incoming communication through a graphical user interface may include: intercepting a first electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the first electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the first electronic communication for processing creating an encoded first electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded first electronic communication with a processor; displayingthe first electronic communication within a graphical user interface on a screen of a computing device of the second user, wherein the displaying the first electronic communication comprises displaying the communication content and displaying a communication transformation query directed to the second user to determine if the first electronic communication should be transformed; receiving a communication transformation indication from the computing device that directs the processor to transform the first electronic communication to a modified electronic communication that is a modified version of the first electronic communication, wherein the processor uses the communication content and the second user communication attribute to transform the first electronic communication; and displaying the modified electronic communication within the graphical user interface on the screen of the computing device.

[0129] In an embodiment, the processor may transform the first electronic communication by removing language from the first electronic communication based at least in part on the second user communication attribute. The removed language may be offensive language. The processor may transform the first electronic communication by summarizing language from the first electronic communication based at least in part on the second user communication attribute. The processor may transform the first electronic communication by reformatting the first electronic communication based at least in part on the second user communication attribute. The processor may transform the first electronic communication by recomposing language from the first electronic communication based at least in part on the second user communication attribute. The processor may transform the first electronic communication by incorporating explanatory text associated with phases based at least in part on the second user communication attribute. The processor may transform the first electronic communication derived at least in part from representations of previous electronic communications from a plurality of user profiles stored in the communication profile database which are like the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning language model or a statistical algorithm to transform the first electronic communication. The transformed first electronic communication may be used to generate an updated second communication profile. The updated second communication profile may be used to predict a most likely modification outcome in a second electronic communication. Transformation data may be extracted from the transformation tothe first electronic communication, wherein the transformation data is used to train a language model used by the processor.

[0130] In an embodiment, a computer-implemented method for modifying a transmitted communication through a graphical user interface may include: intercepting a first electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the first electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the first electronic communication for processing creating an encoded first electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded first electronic communication with a processor; displaying the first electronic communication within a graphical user interface on a screen of a computing device of the first user, wherein the displaying the first electronic communication comprises displaying the communication content and displaying a communication transformation query directed to the first user to determine if the first electronic communication should be transformed and resent; receiving a communication transformation indication from the computing device that directs the processor to transform the first electronic communication to a modified electronic communication that is a modified version of the first electronic communication, wherein the processor uses the communication content and the second user communication attribute to transform the first electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0131] In an embodiment, the processor may transform the first electronic communication derived at least in part from representations of previous electronic communications from a plurality of user profiles stored in the communication profile database which are like the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning language model or a statistical algorithm to transform the first electronic communication. The transformed first electronic communication may be used to generate an updated second communication profile. The updated second communication profile may be used to predict a most likely modification outcome in a secondelectronic communication. Transformation data may be extracted from the transformation to the first electronic communication, wherein the transformation data is used to train a language model used by the processor.

[0132] In an embodiment, a computer-implemented method for modifying a transmitted communication through a graphical user interface may include: intercepting a first electronic communication from further transmission at an artificial intelligence assistant computing facility wherein, the first electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a communication profile for a user using the second electronic identifier, wherein the communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the first electronic communication with a processor; displaying the first electronic communication within a graphical user interface on a screen of a second computing device of the second user; displaying a communication transformation query directed to the second user; receiving a communication transformation indication from a first computing device of the first user that directs the processor to transform the electronic communication to a modified electronic communication; and displaying the modified electronic communication within the graphical user interface on the screen of the second computing device.

[0133] In an embodiment, a computer-implemented method for modifying a transmitted communication through a graphical user interface may include: intercepting a first electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the first electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a communication profile for the second user using the second electronic identifier, wherein the communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the first electronic communication with a processor; displaying the first electronic communication within a graphical user interface on a screen of a computing device of the first user; displaying a communication transformation query directed to the first user to determine if the first electronic communication should betransformed and resent; receiving a communication transformation indication from the computing device that directs the processor to transform the first electronic communication to a modified electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0134] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: intercepting a first electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the first electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the first electronic communication for processing creating an encoded first electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded first electronic communication with a processor; displaying the first electronic communication within a graphical user interface on a screen of a computing device of the second user, wherein the displaying the first electronic communication comprises displaying the communication content and displaying a communication transformation query directed to the second user to determine if the first electronic communication should be transformed; receiving a communication transformation indication from the computing device that directs the processor to transform the first electronic communication to a modified electronic communication that is a modified version of the first electronic communication, wherein the processor uses the communication content and the second user communication attribute to transform the first electronic communication; and displaying the modified electronic communication within the graphical user interface on the screen of the computing device.

[0135] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: intercepting a first electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the first electronic communication was transmitted from a first electronicidentifier associated with a first user to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the first electronic communication for processing creating an encoded first electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded first electronic communication with a processor; displaying the first electronic communication within a graphical user interface on a screen of a computing device of the first user, wherein the displaying the first electronic communication comprises displaying the communication content and displaying a communication transformation query directed to the first user to determine if the first electronic communication should be transformed and resent; receiving a communication transformation indication from the computing device that directs the processor to transform the first electronic communication to a modified electronic communication that is a modified version of the first electronic communication, wherein the processor uses the communication content and the second user communication attribute to transform the first electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0136] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: intercepting a first electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the first electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving from a communication profile database a communication profile for the second user using the second electronic identifier, wherein the communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the first electronic communication with a processor; displaying the first electronic communication within a graphical user interface on a screen of a second computing device of the second user; displaying a communication transformation query directed to the second user; receiving a communication transformation indication from a first computing device of the first user thatdirects the processor to transform the first electronic communication to a modified electronic communication; and displaying the modified electronic communication within the graphical user interface on the screen of the second computing device.

[0137] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: intercepting a first electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the first electronic communication was transmitted from a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a communication profile for the second user using the second electronic identifier, wherein the communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the first electronic communication with a processor; displaying the first electronic communication within a graphical user interface on a screen of a computing device of the first user; displaying a communication transformation query directed to the first user to determine if the first electronic communication should be transformed and resent; receiving a communication transformation indication from the computing device that directs the processor to transform the first electronic communication to a modified electronic communication; and transmitting the modified electronic communication to the second electronic identifier.

[0138] In an embodiment, a computer-implemented method for modifying an incoming communication through a graphical user interface may include: receiving a first electronic communication at an artificial intelligence assistant computing facility, wherein the first electronic communication is associated with a first electronic identifier associated with a first user and directed to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the first electronic communication for processing creating an encoded first electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded first electronic communication with a processor; presenting the firstelectronic communication through an electronic communication link within a graphical user interface on a screen of a computing device of the second user, wherein the first electronic communication comprises displaying the communication content and displaying a communication transformation query directed to the second user to determine if the first electronic communication should be transformed; receiving a communication transformation indication from the computing device that directs the processor to transform the first electronic communication to a modified electronic communication that is a modified version of the first electronic communication, wherein the processor uses the communication content and the second user communication attribute to transform the first electronic communication; and presenting the modified electronic communication through the electronic communication link within the graphical user interface on the screen of the computing device of the second user.

[0139] In an embodiment, a computer-implemented method for modifying an incoming communication through a graphical user interface may include: receiving a first electronic communication at an artificial intelligence assistant computing facility, wherein the first electronic communication is associated with a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the first electronic communication for processing creating an encoded first electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded first electronic communication with a processor; presenting the first electronic communication through an electronic communication link within a graphical user interface on a screen of a computing device of the first user, wherein the first electronic communication comprises displaying the communication content and displaying a communication transformation query directed to the first user to determine if the first electronic communication should be transformed and resent; receiving a communication transformation indication from the computing device of the first user that directs the processor to transform the first electronic communication to a modified electronic communication that is a modified version of the first electronic communication, wherein the processor uses at least one of input from the first user, the communication content, or the second user communication attribute totransform the first electronic communication; and presenting the modified electronic communication through the electronic communication link to the second electronic identifier.

[0140] In an embodiment, a computer-implemented method for modifying an incoming communication through a graphical user interface may include: receiving a first electronic communication at an artificial intelligence assistant computing facility, wherein the first electronic communication is associated with a first electronic identifier associated with a first user and directed to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the first electronic communication for processing creating an encoded first electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier; processing the encoded first electronic communication with a processor; presenting the first electronic communication through an electronic communication link within a graphical user interface on a screen of a computing device of the second user; receiving a communication transformation indication from the computing device that directs the processor to transform the first electronic communication to a modified electronic communication that is a modified version of the first electronic communication; and presenting the modified electronic communication through the electronic communication link within the graphical user interface on the screen of the computing device of the second user.

[0141] In an embodiment, a computer-implemented method for modifying an incoming communication through a graphical user interface may include: receiving a first electronic communication at an artificial intelligence assistant computing facility, wherein the first electronic communication is associated with a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the first electronic communication with a processor; presenting the first electronic communication through an electronic communication link within a graphical user interface on a screen of a computing device of the first user; receiving a communication transformation indication from the computing device of the first user that directs the processor to transform the first electronic communication to a modifiedelectronic communication that is a modified version of the first electronic communication; and presenting the modified electronic communication through the electronic communication link to the second electronic identifier.

[0142] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: receiving a first electronic communication at an artificial intelligence assistant computing facility, wherein the first electronic communication is associated with a first electronic identifier associated with a first user and directed to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the first electronic communication for processing creating an encoded first electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded first electronic communication with a processor; presenting the first electronic communication through an electronic communication link within a graphical user interface on a screen of a computing device of the second user, wherein the first electronic communication comprises displaying the communication content and displaying a communication transformation query directed to the second user to determine if the first electronic communication should be transformed; receiving a communication transformation indication from the computing device that directs the processor to transform the first electronic communication to a modified electronic communication that is a modified version of the first electronic communication, wherein the processor uses the communication content and the second user communication attribute to transform the first electronic communication; and presenting the modified electronic communication through the electronic communication link within the graphical user interface on the screen of the computing device of the second user.

[0143] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: receiving a first electronic communication at an artificial intelligence assistant computing facility, wherein the first electronic communication is associated with a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the first electroniccommunication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the first electronic communication for processing creating an encoded first electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the encoded first electronic communication with a processor; presenting the first electronic communication through an electronic communication link within a graphical user interface on a screen of a computing device of the first user, wherein the first electronic communication comprises displaying the communication content and displaying a communication transformation query directed to the first user to determine if the first electronic communication should be transformed and resent; receiving a communication transformation indication from the computing device of the first user that directs the processor to transform the first electronic communication to a modified electronic communication that is a modified version of the first electronic communication, wherein the processor uses at least one of input from the first user, the communication content, or the second user communication attribute to transform the first electronic communication; and presenting the modified electronic communication through the electronic communication link to the second electronic identifier.

[0144] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: receiving a first electronic communication at an artificial intelligence assistant computing facility, wherein the first electronic communication is associated with a first electronic identifier associated with a first user and directed to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; encoding the first electronic communication for processing creating an encoded first electronic communication; retrieving from a communication profile database a second communication profile for the second user using the second electronic identifier; processing the encoded first electronic communication with a processor; presenting the first electronic communication through an electronic communication link within a graphical user interface on a screen of a computing device of the second user; receiving a communication transformation indication from the computing device that directs the processor to transform the first electronic communication to a modifiedelectronic communication that is a modified version of the first electronic communication; and presenting the modified electronic communication through the electronic communication link within the graphical user interface on the screen of the computing device of the second user.

[0145] In an embodiment, a system may include a server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations including: receiving a first electronic communication at an artificial intelligence assistant computing facility, wherein the first electronic communication is associated with a first electronic identifier associated with a first user to a second electronic identifier associated with a second user, the first electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; retrieving a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute that identifies a receiving communication preference; processing the first electronic communication with a processor; presenting the first electronic communication through an electronic communication link within a graphical user interface on a screen of a computing device of the first user; receiving a communication transformation indication from the computing device of the first user that directs the processor to transform the first electronic communication to a modified electronic communication that is a modified version of the first electronic communication; and presenting the modified electronic communication through the electronic communication link to the second electronic identifier.

[0146] Further, various embodiments discussed herein encompass the subject matter of the following numbered clauses:

[0147] 1. A computer-implemented method includes: determining embeddings for tokens of request data using a machine learning model (MML); accessing text embeddings associated with texts of digital documents, the text embeddings generated by the MML; using K-nearest neighbor method (KNN) performing a search over the text embeddings by comparing the text embeddings with the embeddings for the tokens to identify target text embeddings; obtaining request-related documents corresponding to the target text embedding; generating a prompt comprising the request data and prompt-associated context obtained from the request-related documents; communicating programmatically the prompt to a large language model (LLM); and receiving an output from the LLM.

[0148] 2. The computer-implemented method of clause 1 , further comprising formatting the received output from the LLM based on a template.

[0149] 3. The computer-implemented method of clause 1 , further comprising: parsing the request data into the tokens; identifying a knowledge-shared term corresponding to at least one of the tokens by comparing the tokens with a list of knowledge shared terms; and retrieving from a data storage at least one of documents corresponding to the knowledge-shared term.

[0150] 4. The computer-implemented method of clause 1, wherein the request data comprises: a query from a client device; and a request-related context, the request-related context comprising at least one of: data from an email thread, text data, or session information.

[0151] 5. The computer-implemented method of clause 4, wherein the request-related context is included as a part of the prompt-associated context.

[0152] 6. The computer-implemented method of clause 4, further comprising: receiving an action to perform based on information available in the prompt; in response to receiving the action generating a new context; and merging the prompt-associated context and the new context to obtain a new prompt, the new prompt comprising the request data and new prompt- associated context.

[0153] 7. The computer-implemented method of clause 6, wherein the action comprises summarizing a text within the prompt-associated context.

[0154] 8. The computer-implemented method of clause 6, wherein the LLM comprises a first LLM, the action comprises: transmitting a request to a second LLM to rewrite text data from the prompt- associated context based on a decorator, and wherein the generating the new context includes receiving an output from the second LLM.

[0155] 9. The computer-implemented method of clause 8, wherein the decorator comprises elements or structures added to the prompt to enhance, modify, or contextualize the prompt.

[0156] 10. The computer- implemented method of clause 9, wherein the decorator is an adjective describing a tone or a style of a text.

[0157] 11. The computer-implemented method of clause 6, wherein the action comprises requesting to provide examples for a set of features described in the prompt.

[0158] 12. The computer-implemented method of clause 1, further comprising: generating a report based on the prompt- associated context; receiving an action, in response to the generated report; and retrieving new request-related documents based on the received action.

[0159] 13. The computer-implemented method of clause 12, wherein the received action comprises a click onto a link presented within the generated report.

[0160] 14. The computer-implemented method of clause 1 , wherein the text embeddings are generated from table of contents of the digital documents.

[0161] 15. The computer- implemented method of clause 1, further comprising: parsing the request data into the tokens; and performing a keyword-based search to identify the digital documents.

[0162] 16. The computer-implemented method of clause 1, wherein the prompt-associated context obtained from the request-related documents comprises entire texts from the request-related documents.

[0163] 17. The computer-implemented method of clause 1, wherein obtaining the prompt- associated context comprises for a document of the request-related documents: using the KNN method performing a search within document’s text embeddings by comparing the document’s text embeddings with the embeddings for the tokens to identify target document’s text embeddings; obtaining sections of the document corresponding to associated target document’s text embeddings; and appending the obtained sections to the prompt-associated context.

[0164] 18. The computer-implemented method of clause 1, further comprising: determining a measure of relevance of a document from the request-related documents to the request data; and preventing generation of the prompt-associated context for the document when the measure of relevance is below a target threshold.

[0165] 19. The computer-implemented method of clause 18, wherein the measure of relevance for the document is determined based on a KNN distance between a first vector representing embeddings for the tokens and a second vector representing text embeddings for that document.

[0166] 20. The computer-implemented method of clause 1, further comprising: receiving a template prompt having at least one placeholder; and replacing the at least one placeholder with at least one of the request data or the prompt-associated context.

[0167] 21. The computer- implemented method of clause 20, wherein the template prompt comprises a plurality of placeholders, wherein one of the plurality of placeholders is replaced with a query associated with the request data, and another one of the plurality of placeholders is replaced with the prompt-associated context.

[0168] 22. The computer-implemented method of clause 1, further comprising: providing an interface for receiving actions related to one or more terms; and receiving an action via theinterface causing the search over the text embeddings related to the one or more terms to not be performed.

[0169] 23. The computer-implemented method of clause 22, wherein the interface comprises graphical user interface, and wherein the action includes at least clicking on the one or more terms.

[0170] 24. The computer-implemented method of clause 23 wherein the interface is configured to display: a list of request related documents; and the output from the LLM.

[0171] 25. The computer-implemented method of clause 1, wherein the determining embeddings for a token of the request data corresponding to a knowledge shared term comprises: accessing embeddings for tokens associated with at least one of documents related to the knowledge sharing term; and determining an aggregated embedding for the accessed embeddings.

[0172] 26. The computer-implemented method of clause 25, wherein the determining the aggregated embedding for the accessed embeddings comprises averaging the accessed embeddings.

[0173] In any system or method embodiments described herein, the system or method may further include any of the described features, such as, but not limited to, transmitting the changed electronic communication to the first electronic identifier associated with the first user, and / or transmitting the changed electronic communication to the second electronic identifier associated with the second user. The compositional change may be derived from representations of previous content and context from a plurality of user profiles stored in the communication profile database which are like at least one of the first communication profile or the second communication profile. The processor may be trained on large-scale data mixed with prior communication and effective communications from the plurality of user profiles. The processor may use at least one of a machine learning model, deep learning model, or other statistical learning algorithm for creating the compositional change. The compositional change may be an auto-generated textual completion; the auto-generated textual completion may be a phrasal completion, and the processor may generate the compositional change by optimizing generated language as determined by the processor from the second user communication attribute. The processor may generate the compositional change by replicating a communication style of the first user as determined by the processor from the first user communication attribute. The partial electronic communication may include a communication goal, and the processor may generate the compositional change by optimizing for impact andeffectiveness of generated language with respect to the communication goal. The processor may generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion. The processor may select the communication template using at least one of a machine learning model, deep learning model, or statistical learning model to find a most effective communication template based at least in part on the communication content. The plurality of communication templates may include at least one automatically generated template generated by the processor. The processor may select the communication template by using a machine learning model, deep learning model, or statistical learning algorithm to score the plurality of communication templates based at least in part on the communication content, communication context, first user communication attribute, or second user communication attribute, communication context, first user communication attribute, or second user communication attribute.

[0174] 2. STRUCTURAL & FUNCTIONAL OVERVIEW

[0175] 2.1 ARTIFICIAL INTELLIGENCE ASSISTANT

[0176] Effective communication requires much more than corrections to spelling and grammar. It requires attention to the audience of the communication as well as to context surrounding the conversation. Therefore, to increase the effectiveness of communications, a writing assistant needs to incorporate an intelligent facility. The present disclosure describes an artificial intelligence assistant (AIA), which can utilize information about the participants and the surrounding context and other environments to generate more effective communications. The AIA thus provides a valuable facility to users in communication amongst a diverse population.

[0177] A communication may be understood to be a communicative act between entities, such as between a user as a sender and a user as a receiver, where there is content and context associated with the communication. Although embodiments herein describe communications between two users, such as from a user as a sender and a second user as a receiver, communications described herein should also be understood to encompass embodiments where communications are sent from a user or group or audience, published from a user to a group or audience (for example, where something is not sent directly but published or posted online, such as including when it may visible to anybody but is intended and / or optimized for a specific audience), between a group of user senders (for example, sharing the creation of a communication) and a user receiver, and between multiple users and a group or audience.Users may also be administrators or other individuals that provide services for another user (for example, administrating or aiding in communications of individuals, helping to create and maintain user profiles, and the like). Users may also be a computing entity, such as computing platform, a software agent, a bot, an online page, a webpage, and the like. Users have certain goals, intentions and meaning that they try to convey through content. A user’s ability to express those in content accurately and in a way that a specific receiver or audience would understand is limited, and receivers perceive the content with distortions and can react to it in different ways (for example, actions, inactions, responses, emotions, conclusions, and the like). Receivers’ ability to recover the meaning intended by the user is limited. Content can be classified into communication types and can take different formats and forms. A larger context may include the use case, user intents and goals, the environment (for example, communication channel, platform, conditions, and the like), the current state of the parties (for example, availability, emotional state, or physical state, such as whether they’re a hurry, in commute, on vacation, under stress, tired, happy, upset, and the like), implicit or unconscious signals (for example, tone, pauses, and the like), prior communication history (for example, previous communications in the same thread, previous threads involving the same users, and the like), knowledge of the parties (for example, who they are, their schedules, and the like), general world knowledge (for example, recent news, public facts, and the like), and the like.

[0178] The AIA may have access to user information through a user communication profile (for example, knowledge of the user), receiver communication profile (for example knowledge of the receiver(s), group profiles, and / or target audience profiles, where profiles may include information pertaining to professional vs. informal communication styles, communication preferences, native language, emotional characteristics, and the like), goals (for example, knowledge of what the user wants to accomplish, and the like), communication history (for example, prior communication (for example, email thread, conversation, and the like), shared documents, current state of user (for example, emotions, focused or distracted, and the like), relationship between them, environment (for example, home, office, or commute, and the like), past communications, shared context (for example, current news, pop culture, and the like), individual context (for example, vocabulary, emotions, and the like), modes of communication (for example, email history, digital history, or chats), survey information, and the like), reactions to previous or similar communications, world knowledge (for example, concepts and relations, logic, facts, current news and events, pop culture, and the like), modality (for example, communications channels available, such as email, texting, voice, video, augmentedreality, virtual reality, and the like), may be utilized to modify a user’s communication, either directly (for example, making or providing explicit options for changes to a communication) or indirectly (for example, prompting the user for ways to improve the communication). Although the present disclosure applies the AIA to improvements of a user’ s communication, the AIA may similarly be applied to improving a receiver’s communications, such as re-writing incoming communications for them in a way that they prefer or understand.

[0179] The AIA may help to optimize a communication for a specific receiver or group of receivers, assist in a real-time conversation, support multiple modalities of language and communication, be adaptive to individual users, learn over time based on how participants communicate, and the like. The AIA may be continuously updated, have access and insights about individuals and groups that a user has never communicated with, and can perform in realtime. The AIA, with access to a wide range of continuously updated data and modeled communication behavior, may be enabled with an automated means of determining the perception and predicting the reaction of a receiver, where the system can learn the patterns it observes in the interactions between participants to predict how a specific communication will be perceived and reacted to. In an embodiment, prediction of the perception and reaction of a receiver may be enabled through models of language effectiveness and / or impact, as described herein, including general, group-domain specific, and receiver-specific modeling, such as trained on a corpus of previously observed communications and reactions.

[0180] The AIA may suggest changes related to communication structure, language, style, and the like, that relate to a specific receiver, group of receivers, or broad audience. The AIA may suggest recomposing or rewriting a communication based on a goal, context, communication style of a user, prior communication that achieved similar goals and effects, and the like, such as stored in a user communication profile. The AIA may assist during a realtime dialog, such as on a display of a computing device while a conversation is ongoing (for example, displaying feedback to the user during a texting exchange, providing audio prompting (for example, via a speaker in the user’s ear), providing haptic feedback during a conversation on a smartphone or augmented reality glasses, and the like). In an embodiment, the AIA may be able to provide iterative real-time feedback that is visible to the user as a sender but not to the receiver, taking the reaction to the previous communication as additional input.

[0181] FIG. 2 illustrates an AIA communication system model, where a user generates a communication as an input to the AIA, such as a written electronic text (for example, email, text message, document, and the like), voice communication (for example, voice input to atelecommunications system), and the like. The AIA receives and processes the communication input with respect to a user communication profile, receiver communication profile, goals, context, world knowledge, and modality, and generates an output in the form of a modified communication or feedback to the user that is directed at optimizing the effectiveness, clarity, and correctness of the communication. Input and training data to the system may also consider the current communication (for example, a draft communication), revisions of communications, targets for a communication (for example, goal, intended impact, audience, desired style, and the like), ‘lookalike’ communications (for example, similar communications between similar parties in the past that are predictive to the current communication, and the like), context (for example, domain, topic, type of communication, prior communication history, user emotional state, and the like), user’s communication profile (for example, user’s preferred style, user’s vocabulary, user’s proficiency, and the like), receiver’s communication profile, relationship between users, communication constraints (for example, time, length, medium, and the like), world knowledge, domain knowledge, shared documents between users, reaction to previous communication (for example, textual response, verbal response, video recorded body language (for example, posture), and the like), geography with respect to language (for example, different spellings, different idioms, and the like) and usage combined with geography to determine a language variation (for example, U.S. vs. Australia or dialect (for example, different regions within a country)), geography with respect to context (for example, service representative providing feedback to a customer with respect to geography to incorporate ‘relatedness’ into a communication), information from wearable devices (for example, biosensors, cameras, microphones, and the like), and the like.

[0182] The AIA may utilize user communication profiles in the process of optimizing a communication, where communication profiles contain properties and communication preferences of users. A communication profile may be described as a set of characteristics, traits, features, and the like that can be used to characterize and represent an individual either independently or to cluster them as part of a larger cohort. Communication profiles may be records consisting of at least one property expressing characteristics related to communication, such as language proficiency, vocabulary, style, topics, preferences, reactions, behavior, and the like. Communication profiles may change over time (for example, with respect to vocabulary, language proficiency, triggers, interaction preferences, and the like). A communication profile of a person can be learned and updated by analyzing their outgoing communication and reactions to incoming communication, asking them questions, having themfill in some of the profile properties, using communication profiles of similar users as an approximation, and the like. Communication profiles of groups of users or larger audiences can be formed by aggregating individual profiles or getting statistics on communication with a certain group / audience.

[0183] Profiles may be associated with certain communication parties (for example, entities that can be senders or receivers). These may be individuals, clusters of individuals with similar socio-demographic or psychological properties, clusters of individuals with similar communication profile properties (for example, similar vocabulary, topics, style, communication preferences, behavior, and the like), groups of connected individuals (for example, coworkers within one team sharing a certain context), larger populations (for example, children audience, non-native English speakers, clusters of individuals with similar socio-demographic or psychological properties, clusters of individuals with similar communication profiles, groups of connected individuals (for example, teams), audiences / audience types, and the like). Communication profiles may contain records. Records of communication parties may include their socio-demographic and psychological properties, descriptions and properties that define or unify a group, and the like. Individuals may belong to multiple clusters, groups, and audiences, such as native English speakers, IT professionals, teenagers, certain geographies, and the like. Individuals may be connected as a graph with links denoting familiarity level and types of relationships. Familiarity may be initiated and updated based on the frequency of observed communication and its tone, among other signals. Relationship may be specified by users or inferred by the AIA from observed communication between them.

[0184] Profiles may include properties such as language proficiency level, native language, characteristics of the party’s communication such as vocabulary (for example, word lists with frequencies) and frequently used phrases (for example, word and phrase frequency counters updated on each communication session processed by the system by extracting most frequent collocations and language patterns), other characteristics of writing / speech, such as style, structure (for example, including vector representation of a sample of writing, individual language models, frequency counters on different aspects of style, structure, tone, and the like), topics (for example, list of topics with frequency counters updated on each communication session processed and / or classified by the system), negative and positive triggers (for example, specified by the user or inferred from observed reactions), communication behaviors and preferences (for example, activity hours, time to respond, communication channels, low-context vs. high-context, verbose vs. concise, emotions vs. facts, format (for example, plainrich, short-long, bullets-prose, and the like), and the like. Triggers may be communication elements or characteristics (for example, words, phrases, errors, topics, formatting, and the like) that an individual or members of a group consistently like or dislike. Triggers may cause strong emotions. Negative triggers carry the risk of tension or conflicts in communication.

[0185] The AIA may learn communication properties directly from a user (for example, through surveys, prompts to fill in properties, or state preferences), by observing and analyzing their communication, and by analyzing / processing their reactions to incoming communications. For instance, communication parties may inherit the properties of groups or audiences to which they belong. Properties of more granular parties may override the equivalent properties of (for example, larger) groups or audiences. Communication profiles may include reaction data, such as in a corpus of communications and documents (for example, original or vector representations) and the corresponding reactions from different communication parties (for example, used to train AIA to predict reactions of a receiver). Profiles may be initialized with user input through surveys, iterative questions, analyzing records of prior communication, and the like. In an embodiment, communication profiles of the receivers who do not have an existing communication profile in the system (for example, not a registered user of AIA) may have a communication profile generated in runtime and not stored persistently or stored in de-identified clusters.

[0186] Data sources for communication profiles may include results of researching characteristics of certain populations, user surveys, explicitly stated user preference (for example, filled-in part of the profile), observed outgoing communication (for example, sampling aggregated representation, data extracted through analysis / inference, observed communication behavior (for example, schedule, channels, and the like), observed reactions to incoming communications (for example, responses, actions / inactions, and the like), data from biometric sensors (for example, to track the current state, emotional reactions to various communication, and the like), integrations with other systems holding relevant data, and the like.

[0187] Communication profiles may be taken from public and private sources, such as databases, academic datasets, demographic information (which can be inferred by joining the different DBs), psychological profiles, author profiling, usage guidelines, relevant data sets (for example, psychological profiles, demographic data, social media status updates), and the like. Communication profile collections and usage may utilize various methods such asauthorship attribution (for example, word distribution from social media), clustering, deep learning (for example, language modeling, author vectors, comparison, and the like), text classification (for example, with respect to gender), authorship score calculation, author profiling, user profiles (for example, from e-commerce, social media, articles, and the like), user embeddings (for example, factor adaption, user content, user type embeddings, and the like), personalized natural language processing, personality assessment, linguistic homophily, linguistic style, and the like.

[0188] Communication profiles may include information about the user of communications, such as their age, gender, race and ethnicity, professional concentration, occupation, past employment, current employment, workplace hierarchy and dynamics, residential geographic location, professional / work geographic location, current geographic location, geographic location of origin (for example, home country, city, nationhood), primary and secondary communication languages, religious heritage, religious views, political views, personal preferences, social hierarchy dynamics, personal psychological type (for example, neuroticism, extraversion, openness, agreeableness, conscientiousness, and the like), personal human values (for example, self-transcendence, self-enhancement, conservation, openness-to- change, hedonism, and the like), Meyers-Briggs Type Indicator (MBTI), cognitive style(s), contextual emotional states (for example, under what conditions the individual experiences general positive emotions, optimism, general negative emotions, depression, anxiety, anger, and the like), and the like.

[0189] Communication profiles may include an individual’s lexical writing features, such as lexical features at the character-level (for example, character n-grams (for example, labeled char-n-grams), percent of characters per document (for example, ratios between upper case characters, emoticons, periods, ellipses, return characters, commas, parentheses, exclamations, colons, digits, semicolons, hyphens and quotation marks and the total number of characters in a communication, and the like), at the character-level (for example, number of words, mean number of characters per word, mean number of characters per word, and the like), at the sentence-level (for example, mean number of words per sentence, standard deviation of words per sentence, difference between the maximum and minimum number of words per sentence), emoticons (for example, number of emoticons, number of emotions per word, sentence, document, and the like), special characters (for example, ‘$’, ‘%’, ‘@’, and the like), vocabulary richness (for example, percentage of distinct words, type-token ratio, hapax legomena, Yule’s K, dis legomena, and the like), readability (for example, metrics (forexample, Flesch- Kincaid metric), word length distribution (for example, word-lengths per sentence, email selection, document, and the like), sentence length distribution (for example, sentence-length per email selection, or document)), and the like.

[0190] Communication profiles may include an individual’s syntactic features, such as parts of speech (POS) features (for example, POS n-grams (for example, fine-grained and coarse-grained), relative frequency of comparative and superlative adjectives and adverbs, relative frequency of the present and past tenses, phrase structure of general grammatical categories (for example, noun-phrase, verb-phrase, and the like)), dependency features (for example, frequency of each individual dependency relation per sentence, percent of modifier relations used per tree, frequency of adverbial dependencies (for example, providing information on manner, direction, purpose, and the like), ratio of modal verbs with respect to the total number of verbs, percent of verbs that appear in complex tenses referred to as “verb chains” (VCs)), tree features (for example, tree width, tree depth, branching factor (for example, mean number of children per level)), function words, filler words, stop-words, punctuation, formality, and the like.

[0191] Communication profiles may include an individual’s discourse features. For instance, to obtain a discourse structure, the system may use a discourse parser, which receives as input a raw text, divides it into elementary discourse units (EDUs) and links them via discourse relations that follow a rhetorical structure. The system may then compute the frequency of each discourse relation per EDU (for example, dividing the number of occurrences of each discourse relation by the number of EDUs per text). The system may also determine the shape of the discourse trees (for example, with respect to depth, width, branching factor, and the like), ratios of discourse markers, inteij ections, abbreviations, curse words, polar words (for example, positive and negative words in polarity dictionaries with respect to the total number of words in a text), and the like.

[0192] Communication profiles may include an individual’s idiosyncratic features, such as spelling errors, syntactically classified punctuation (for example, end-of-sentence period, comma separating main and dependent clauses, comma in list, and the like), grammatical errors (for example, sentence fragments, run-on sentences, subject-verb mismatch, repeated words, missing word errors, all-caps words, abbreviated words, letter inversions, only one of doubled letters, repeated letters, missing hyphen, wrong singular / plural, wrong tense, wrong verb form, wrong article, wrong preposition, repeated non-letter / non-numeric characters (for example, ???, or ! ! !), and the like.

[0193] Communication profiles may include an individual’s communicative structural features, such as at the communication-level (for example, has greeting, signature, URL, quoted content, body length, subject length, and the like), paragraph-level (for example, mean number of paragraphs per email, paragraph lengths, mean number of sentences per paragraph, mean number of words per paragraph, and the like), in technical structure (for example, attachments, fonts, use of images, and the like), and the like.

[0194] Communication profiles may include information about what a user may write, such as the results of statistical analysis of the user’s content (for example, word n-grams, word and document embeddings, term frequency-inverse document frequency (tf-idf), acronyms, first- person pronouns, slangs, leetspeak (for example, 133t, pwn3d, and the like), emoji, dictionaries , use of stock phrases (for example, idiomatic expressions, metaphors, support verb constructions, names of persons, locations, and other entities, dates, domain-based technical terms (for example, academic, legal, medical, scientific, administrative, and the like).

[0195] Communication profiles may be constantly updated, such as with respect to communications between participants, documents that the user generates or edits, writing style, errors encountered, previously stated goals, business associations, preferences, and the like. User communication profiles may then be used by the AIA to help modify new communications based on the latest information available for the user(s).

[0196] Communication profiles may include representations of typical reactions and interactions extracted from communications, such as information extracted from previous responses (for example, using sentiment analysis and other natural language tools), extracted non-verbal reactions from speech and / or video (for example, tone, facial expressions, posture, and the like), received interactions with other systems (for example, views, clicks, reaction time, and the like), inputs explicitly reported by a user (for example, in the setup or update of a communication profile, through a survey, use of emoji, and the like), inputs reported by a user (for example, through a product survey), lookalike communications, and the like.

[0197] Communication profile data may be used in training data for transformation processes. In addition to data associated with the content of communications, other factors may be considered. For example, a video stream of a receiver may be evaluated for body language during a video-call, capturing the gesture and posture of the receiver. For instance, this data may be correlated to geographic location to account for different body language for different countries / cultures, where the information about geography is mapped to the posture in a reaction (for example, posture mapping to anger or acknowledgment).

[0198] In an embodiment, when a communication is addressed to a plurality of receivers, the communication profiles for those targeted receivers may be used individually to generate unique communications by the AIA for each receiver, to generate a single audience targeted communication, or a combination of individual and group targeted communications. In an embodiment, the user may be able to specify how a plurality of individuals in a targeted audience population is to be communicated with, such as individually or as a group. For instance, the user may generate a single communication, but intends it to be transmitted to an audience of ten individuals, where three may be targeted for individual communications and the rest targeted for a group communication. The individual communications may utilize the receiver’s individual communication profiles to process the user’s communication into unique communications, and the rest of the communication profiles may be used to generate some ‘average’ response (for example, a weighted averaging of receiver characteristics from the group’s communication profiles) from the user’s communication. The AIA may enable the user to generate a single communication that is customized per the target audience.

[0199] The AIA may utilize communication profiles to transfer meaning to a user communication. As illustrated in FIG. 3, a user, generating a communication, may interact with the AIA, and the AIA may in turn access the user’s communication profile in addition to the receiver’ s communication profile. Communication profiles may comprise characteristics such as writing style, speaking style, communicative emotions employed in past communications, vocabulary usage, the user’s native language, contextual information, and the like. This information from users may then be used by the AIA to predict perception related to how the communication could be received; provide editing suggestions, composition suggestions, and coaching; assess how goals may be met; how lookalike communications (for example, templates) may be used to compose a new communication; and the like. The AIA may assist the user in composing communications that transfer meaning to the receiver as intended. The process may begin with the AIA processing and determining the user’s goals, communication draft and key points, processing and incorporating the user’s communication profiles, structuring a communication to achieve a goal, determining the best structure and tone to meet the goals, and the like. For instance, the AIA may add a clarifying point to a communication based on email history and context of the receiver, use language and style that relates with the receiver, change the tone of an emotionally charged communication so it’s less likely to offend the receiver, use a different word the receiver will understand, make the text written by a non-native speaker less idiomatic, and the like.

[0200] In an embodiment, the AIA may operate on a plurality of computer-based platforms, such as on laptop / desktop computers (for example, native applications, editor integrations, and the like), Web (for example, standalone, browser extension, and the like), mobile computing devices (for example, smart phone, tablet, and the like), augmented or virtual reality glasses, home and office interactive assistant augmentation, video conferencing, voice assistance, wearables, and the like. The AIA’s ability to imbibe information, context, and global knowledge into a communication may be applied to a great diversity of applications, including personal and professional communications (for example adapting to different genres, styles, formalities, and the like), individual and group communications (for example, targeting individuals, groups, or a combination of individuals and groups, and the like), real-time conversational assistance (for example, aiding a user during a live conversation), interactive communication generation (for example, aiding the user in drafting an email or text prior to sending), entertainment (for example, applying a humor filter to a communication), business services (for example, modifying a communication with a client), and the like. For example, the AIA may help a user generate a communication through an interactive exchange, where the AIA interjects feedback associated with different target audience characteristics (for example, composing an email to your mother, versus to a friend or professional associate). In another example, the AIA may aid a user during a conversation, such as during a voice conversation or a text exchange (for example, where the AIA provides feedback during the conversation on a smart phone through a displayed dialog box, on a pair of augmented reality glasses, and the like). The AIA may also be applied to entertainment. For example, a user may want to add a personal anecdote (for example, to lighten the tone of the communication) or turn a benign communication into a funny one (for example, applying a humor filter, such as in a similar fashion as people applying a humor filter to a photo).

[0201] The AIA may be applied to a plurality of use-cases, such as for grammar and fluency correction, clarity, and effectiveness in combination with context and communication profiles, improving human-computer interaction effectiveness, augmenting outgoing voice and video messages, augmenting conversations, augmenting incoming communications, providing communication analytics and coaching, providing conversational agent modality, and the like.

[0202] The AIA may be able to predict the reaction of a receiver, such as alerting the user if a communication doesn’t match the target receiver’s preferences and suggest revisions. Given a communication and a goal or desired reaction (for example, where the goal is to entertain, where the target reaction is to produce a smile or a laugh, and the like) the systemmay predict whether the communication is likely to cause this reaction, based on the corresponding component of the receiver’s communication profile (for example, what he or she tends to find entertaining). Then the AIA may generate revisions that make the communication more likely to spark the desired reaction. For example, the AIA can be programmed to generate revisions that make the communication more like other communications that the system has seen that triggered a similar reaction from this specific receiver or group. Once the communication has been sent and a response has been received, the AIA may help the user understand a particular reaction, such as detecting tone and other non-verbal signals in the incoming communication, generate advice on how to reinforce desirable reaction and mitigate an undesirable reaction, and the like. The AIA may utilize or generate a communication template given the goal and context of a communication, such as based on patterns learned from other communications with the same goal and similar context that the system has seen or suggest a communication format and modality based on the goals, context, and communication profile of the receiver, such as converting the communication into a different modality, picking a delivery channel, and the like.

[0203] In an embodiment, the AIA may improve human-computer interaction effectiveness, such as with an API for a third-party system that allows rewriting an automatically generated communication to look more like it was drafted by a human or appealing to a given target audience or receiver, and effective for a given goal. AIA may customize the communication for each receiver based on their communication profiles and select an appropriate delivery channel, such as based on an automatically generated communication with context, goals, a list of receivers and the like.

[0204] In an embodiment, the AIA may provide communication analytics and coaching. For instance, the AIA may provide coaching based on observed communication, where the system generates personalized advice on what the user could do to improve communication and be more effective or productive (for example, develop active listening or non-violent communication skills, expand vocabulary, brush up on grammar, be more concise, improve pronunciation, speak slower, use questions instead of directives, limit social media use, and the like), avoid common mistakes the user typically makes (for example, to gradually reduce repetitive errors or behavior), and the like. The AIA may capture not only outgoing communication or conversations but also consumed information, such as using a camera or microphone feed from a wearable device, such as on smart glasses, analyze the receiver’s reactions to various content and conversations using data from sensors, such as wearable andbiosensors (for example, temporal data, and the like), and use this to build or augment the receiver’s communication profile and for communication coaching. Coaching may be provided in real-time, as feedback, provided off-line, and the like. In an embodiment, this information may also be used as training data for communication models.

[0205] In an embodiment, the AIA may provide a facility for conversational agent modality, where the AIA may vary the type of interaction system utilized. For instance, a “highlight and suggest” interaction may be utilized, or the AIA may rewrite a sentence or passage given different constraints (for example, goals, proofing, clarity, effectiveness, and the like), or in a ‘more forward thinking’ mode, where the system chats with the user to work with them interactively to improve their communication, such as a friend or tutor would. For example, a user may generate a message, and receive an interactive dialog, such as “Hi! I’ve read your communication and have some tips ! Your grammar is fine, but I am concerned about organization. Can we start?” Once the user agrees, the AIA may provide guidance and dialog, such as “I’ve highlighted what I believe is the thesis statement of your message. I think this would make more sense moved up. Let me show you.”

[0206] The AIA may provide a host of messaging improvement facilities, where any number of them may be utilized and combined. For example, a user may utilize the AIA to write an email for applying for a job. The person may write a first pass email and then the AIA chooses the best modality to give them feedback. Maybe the user is a proficient writer and can look at comments made on the side, maybe the user is embarking on their first job search and the AIA may select a “conversational agent” mode and walk the user through why certain things should be changed and what is good, and why. With a modality selected, such as for “automatic rewrite”, AIA may review the draft, use language modules to check for grammatical errors, recognize that the goal of the email is to impress the receiver (for example, a hiring manager), and elect to transform the text into a more formal style. The AIA may access receiver communication profiles and determine that the receiver likes a certain pop culture reference and likes attachments in PDFs, so a subtle reference to a popular movie is slipped into the cover letter and the cover letter is converted from a word processing format into a PDF format. In addition, given the goal, the AIA may make sure to open the email with some ebullience to show that the user is really interested in the job and pulls content from their cover letter that would back up their enthusiasm for the job. For example, if they were applying for a job in a particular company, the system may pull out an internship at that company from last summer and use that as an example. The system may also check for missing information, suchas a signature at the bottom of the email and phone number where they can be reached. In this interchange, the AIA may utilize a number of communication improvement facilities in combination, such as modality selection, proofreading, effectiveness, style transformation (for example, includes vocabulary shifting), goal detection and optimization, receiver communication profile scanning (for example, inclusion of pop culture reference, affinity for a certain file format, and the like), effectiveness (for example, some parts are made more persuasive to show interest), argumentation (for example, supported by pulling facts from different sources automatically), check for missing information (for example, system has an idea of what a good letter looks like for this task), and the like.

[0207] In an embodiment, the AIA may be utilized in an educational environment, such as where a user is a student that is provided suggestions with respect to writing structure and style, with prompts to inappropriate tone, with additional context associated with new concepts, and the like. For instance, a student may be writing a paper on the civil war, and the AIA not only suggests modifications to the text with respect to improving the student’s quality of writing, but also suggests context to aid in the support for a student’s theme or bring related current events to the student’s attention.

[0208] In an embodiment, the AIA may be used for grammar and fluency correction. For instance, the AIA may utilize context to improve grammar and fluency, such as applying a different dictionary and grammar based on a user’s location, the specified audience (for example, use Australian English spelling and grammar norms by default if the users are (mostly) Australian), and the like. The AIA may utilize communication profiles to improve grammar and fluency, such as applying a different grammatical error correction model depending on the user’s native language (for example, applying a model trained on text with errors typical for Chinese learners of English to improve the accuracy of correction), adjusting spelling and grammar correction according to a receiver's sensitivity or opinion on certain writing issues. For example, based on an observed communication profile of the receiver, the system may know that he or she gets annoyed by lowercasing in short text messages or by sentences ending with prepositions, and enable the corresponding checks and applications of corrections if a user is writing to this person.

[0209] In an embodiment, the AIA may be used to improve clarity and effectiveness, such as through context and communication profiles. For instance, an alert of readability or vocabulary mismatch may be provided to a user based on the target audience (for example, too many idioms in a text for non-native speakers or inclusion of complex language in a text forchildren), a suggestion for an improvement to or automatic rewrite of the text may be provided to adjust readability and vocabulary, a rewrite of an email may be provided to maximize a positive outcome, and the like. For a given email, the AIA may suggest or apply the structure and revisions that increase the chance of a positive response or rewrite the text to make it more effective (for example, targeting specific sentences or generating a new email from scratch). A communication profile, prior communication, documents, calendar information, and the like, may be considered, such as were shared in the thread or elsewhere. For example, a sales email may be rewritten using a model trained on a corpus of similar sales emails with response and / or click data. In another example, improvements or a rewrite may be suggested to a support email so that a specific customer (receiver) is satisfied, such as using a model that is trained on a corpus of similar support emails tagged with satisfaction ratings and possibly also the receiver’ s communication profile.

[0210] The AIA may have the capability to modify a communication to appeal to a specific receiver, such as generating revisions or rewriting the communication using the vocabulary that the receiver would understand or using the structure and language that the receiver tends to use or reacts positively to (for example, based on his or her communication profile). Adding or removing content may be suggested, such as depending on the user’s goals and receiver’s background, such as using individual and shared context available to the system or adding references and content from documents (for example, previous interactions with the receiver), pop culture, current news, or other kind of world knowledge.

[0211] The AIA may have the capability to detect biased or inflammatory language and suggest more balanced rewrites when possible, such as warning the user if a specific receiver can be annoyed by certain forms of language (for example, structure, tone, certain terms, and the like), based on his or her communication profile. The AIA may detect a user’s emotional or physiological state to correlate with their current writing style, such as through biometric data from the user (for example, from a wearable device), visual indicators (for example, from facial indicators analyzed from a user-facing camera, such as mounted on a user’s laptop or smartphone), changes in writing style with respect past messaging and responses, and the like. For instance, a user may be wearing a smartwatch that monitors the temperature, heart rate, galvanic skin response, voice data (for example, vocal inflections, volume, speech patterns, and the like), and blood sugar, where changes in these levels are matched to emotional profile data, such as stored on the user’s communication profile, as determined from similar measurements in a population, and the like. The AIA may then be able to determine if theseindicators are adversely affecting the user’s communications, such as when a spike in body temperature, heart rate, and low blood sugar is detected, making a person more likely to communicate in a frustrated or angry manner. The AIA may then prompt the user to be aware of such potential influences, offer communication suggestions related to the same, or revise a communication. The AIA may also correlate detected emotions with the content and / or communication the user deals with and use this as training data for modules that predict reactions and for updating communication profiles.

[0212] In an embodiment, communication profiles and available context may be utilized to compose and / or edit user content (for example, an electronic communication) in a way that maximizes the probability of the receiver(s) perceiving the received content as intended by the user and reacting positively. FIG. 4 depicts a workflow of asynchronous communication, including general advice to a user generating a communication and compositional and editing assistance to the user for the communication. The workflow begins with the user 404 (for example, a user sender) providing a communication input 402 to the AIA 400, where the AIA 400 receives the communication input 402 with a content and context module 414. The user 404 may provide an initial communication input 402 (for example, draft a message, upload a document, or just start typing or dictating). The user 404 may provide context by specifying the use case, goals, intent, desired reaction, and the like. The user 404 may provide prior communication history by opening a thread and / or granting the AIA 400 access to their inbox. The AIA 400 may infer more context from the communication environment, integrated systems, and recent user activity (for example, channel, schedule, user’s current state, and the like). The AIA 400 may classify the communication type and use case based on initial content and / or prior history. Techniques for classification may include rule-based, machine learning, deep learning, statistical models, and the like. The AIA 400 may present (at least in part) inferred context to the user 404 via communication output 408A to confirm or override. The AIA 400 may apply default context values (for example, default goals for a certain use case) if they are not provided by user 404 and cannot be reliably inferred.

[0213] The user 404 may specify a second user 412 (for example, a user receiver), such as by selecting from a contact list or picking applicable audience types, as described herein. The AIA 400 may identify the use case through interaction with the user 404, such as the goals of communication and desired / undesired reactions. The AIA 400 may begin the session as preconfigured or be launched with a certain use case parameter, infer the use case from the environment or prior communication, prompt the user 404 to specify the use case (for example,“feedback”, “personal request”, “confirmation”, “greeting”, “formal application”, “report”, and the like). Based on the use case, the AIA 400 may infer the goals and communication objectives and prompt the user 404 to confirm or modify them (for example, “convince,” “impress,” “entertain,” “inform,” “express gratitude,” and the like). Based on the use case and goals, the AIA 400 may identify or prompt user 404 to specify desired and undesired reactions of user 412 (for example, emotions, actions, and the like). For instance, a default desired reaction could be “a (positive) response”. A default undesired reaction could be “no response,” “negative response,” “anger,” “frustration,” and the like. The AIA 400 may provide for a default set of messaging goals, such as that a communication should be accurate (for example, error-free), not ambiguous (for example, clear), not harmful or offensive, doesn’t cause or escalate conflict, minimizes the risk of misunderstanding, and the like. In an embodiment, the goals of communication and desired / undesired reactions may be updated at any time throughout the interaction with the user, as described herein, based on updated inputs (for example, as the user 404 enters the subject for the communication or proceeds to type or dictate the communication).

[0214] The AIA 400 may identify other communication contexts, such as a communication channel, platform, or medium (for example, web, mobile, specific communication clients and collaborative editors, chat, text / voice, and the like), a communication type, form, genre (for example, email thread, short / instant message, shared document, social media post / comment, voice message, and the like), prior communications between the user 404 and user 412, location and time of day / week for the user(s), and the like. The AIA 400 may identify that the channel, time, or form of the communication are not optimal given the receiver and the use case / goals and suggest alternatives, such as based on the preferences explicitly provided by the user 412 or derived from receiver’s previous communication patterns observed and learned by the AIA, such as statistics of the receiver’s use of channels for different use cases, availability or schedule (for example, extracted from receiver’s calendar), typical time they’re checking and responding to incoming communication, communication forms they tend to respond to better, and the like. This recommendation may be updated as the user 404 works on the communication (for example, suggest an email or document form instead of an instant message if it exceeds certain size, suggest a short video recording if the communication is more informal). Extracted content and context obtained by the content and context module 414 is then passed to the preprocessing module 416, which encodes it into a common representation (for example, as a vector or matrix) to be used as an input into machine-learning models, deep-learning models, other statistical modes, and the like, within the communication transformation module 418.

[0215] The preprocessing module 416 prepares content and context, such as in association with accessed communication profiles 428 (for example, users and / or group profiles), for use by the communication transformation module 418. In an embodiment, the AIA 400 may interface with third-party communication platforms (or may be integrated with the third-party communication platform) to retrieve information about the user 412, such as a receiver ID (for example, email address) derived through extraction from the receiver field, from the thread or conversation (if it’s a response), and the like, from the user’s communication input 402. For example, accessing the receiver field through a Gmail API or browser extension. Alternatively, the AIA 400 may prompt the user 404 to specify the receiver / audience, such as by entering an email, selecting a contact from the list, selecting the audience or group from the list (for example, team X, family members, industry group Y), selecting the receiver or audience type from the list (for example, a close friend, a boss, expert audience, children audience, or native / non-native speakers), and the like. The AIA 400 may infer the receiver type or target group or audience based on the environment or context of the communication input 402, such as the communication client, user account, prior content of the thread or conversation, communication content, and the like, where for instance, the automatically identified receiver type or audience is presented to the user 404. The user 404 may be presented this information, and be provided an option to override, specify, narrow down, and the like, the selection. If the user 412 is known to the AIA, the preprocessing module 416 pulls the corresponding communication profile 428. If not, it may create a new one. The preprocessing module 416 may also associate the user 412 with relevant group communication profiles 428. In an embodiment, communication profile creation and visibility may be restricted to members of a certain group (for example, members of a corporate email domain, contacts specified by the user, and the like).

[0216] The preprocessing module 416 passes the encoded content and context, and the information accessed in communication profiles 428, to the communication transformation module 418 for optional processing, including general advice to the user through the advice module 420, communication composition assistance through the composition module 422, and editing of communication versions through the editing module 424, such as in conjunction with communication sources 430 as described herein. For instance, as part of providing general advice through the advice module, once the communication profile 428 is identified, AIA 400may display general characteristics of the receiver or audience and provide recommendations on how to communicate with them. Characteristics may include receiver’ s occupation or role, location, knowledge and interest areas, receiver’s language proficiency (for example, native or non-native English speaker), expected or average response time, availability, personality type or traits, and the like. Characteristics may be explicitly provided to AIA 400 by the receiver (for example, as part of filling in AIA 400 user profile), learned through surveys (for example, personality type or traits can be learned through a personality survey delivered to the receiver by AIA), which may be in-product surveys, obtained from external systems accessible to AIA (for example, receiver’s occupation may be extracted from a public personal information directory, professional contact network, web search and the like; receiver’s availability may be obtained from a calendar) or extracted by AIA 400 from receiver’s observed communication (for example, receiver’s knowledge and interests can be learned by analyzing the receiver’s communication through topic classification or keyword extraction algorithms and identifying the most frequent topics or keywords; receiver’s language proficiency can be identified by analyzing receiver’s vocabulary or running receiver’s communication through a statistical learning algorithm classifying input communication as native or non-native (produced by a native or a non-native speaker); expected response time may be based on historic data on receiver’s typical response time in the same or similar communication scenarios). Recommendations may include preferences and other information explicitly provided by the receiver such as preferred communication channels (such as email, instant messages, voice messages, specific communication platforms or applications), preferred communication format and style (such as formal or informal, brief or detailed, plain or rich text, logical or emotional, and the like), preferred communication time (for example, separately for different communication use cases, such as personal and business) and the like. Recommendations may also be derived by AIA 400 through analyzing a receiver’s communication patterns (for example, AIA 400 may recommend formats and styles that the receiver responds to most positively and consistently (based on the sentiment analysis of responses and the distribution of time to respond); alternatively, AIA 400 may recommend formats and styles that match / mimic the receiver’s prevalent communication styles and format).

[0217] As a part of general advice, the AIA 400 may suggest a different time to send a communication output 408B, such as if the receiver is determined to be occupied (for example, based on a retrieved schedule if the AIA 400 has access to it), if the receiver is more likely to provide sufficient attention at the current time where the receiver may better respond to anincoming communication sent at a different time (for example, based on their stated preferences or past behavior recorded in their communication profile, such as when the receiver most frequently checks and responds to incoming communication). Suggested time may be different for different communication use cases, such as personal and business (for example, if receiver’s explicit preferences or observed behavior differ in these use cases). The AIA 400 may suggest a different channel to send a communication (for example, email, instant messages, voice messages, specific communication platforms or applications), such as if the receiver is likely to see it and respond faster through a different channel (based on their stated preferences or past behavior recorded in their communication profile, such as the frequency of use of different channels for different communication use cases (among the channels AIA 400 integrates with or has access to)), if the topic, structure, or length of the communication are a better fit for a different communication channel (for example, sensitive topics may better be discussed over phone or longer communications may better be sent over email rather than IM), and the like.

[0218] As part of general advice, given a message produced by the user, the AIA 400 may inform the user of the predicted receiver’s reaction to such message. AIA 400 may alert the user in case the predicted reaction differs significantly from a desired / expected reaction. The desired / expected reaction may be derived from the user’ s intent or pre-defined in the AIA 400 for certain communication use cases (for example, if the user’ s intent is “persuade,” AIA 400 may consider “response with a confirmation” as desired / expected reaction; if the use case is a social media post, AIA 400 may have likes, shares, and comments pre-defined as desired / expected reactions).

[0219] In an embodiment, the AIA 400 may provide advice to a user who has sent a communication with feedback from the receiver of the communication, such as described herein. For instance, a user may be provided with reaction feedback, where the AIA 400 provides advice on how to follow up with a second communication to the receiver, such as changing the tone of the communication, adding more information, providing less information, making the communication less formal, removing language identified as undesirable, abusive, or offensive by the feedback, and the like. The feedback may also be included in an update to the user and / or the receiver’s profile so that any preferences or issues raised in the feedback can be incorporated into future communications.

[0220] The communication transformation module 418 may provide interactive feedback 406, such as prompting the user to specify a user / audience and the context that AIA 400 cannotget or infer from the environment, prior user input, user settings, available communication draft, and history, and the like. The AIA 400 may prompt the user via communication output 408A to confirm or override the receiver / audience and the context the AIA 400 has identified. The user may update the content, context, and receivers at any time during the session. The AIA 400 may detect changes, re-process inputs and adjust its feedback; present general recommendations that the user can react to (for example, acknowledge, dismiss, mark as helpful or incorrect, and the like); use user feedback on the general recommendations to improve them in the future; present template and / or textual completion options (for example, the user may preview and select one or none of them); offer completions and continuously update completion suggestions as the user is entering text (for example, typing or dictating); and the like. In more structured communication use cases, the AIA 400 may guide the user via the communication output 408A through a single template / outline and have the user answer questions / fill in certain blocks of content. The AIA 400 may generate the edits for the text the user has entered and automatically re-check it and update the edits when the user modifies the text, rewrites the content highlights changes, and allows the user to review and revert individual ones, and the like. The user may interact with and provide feedback on AIA's individual suggestions (for example, general recommendations, templates, completions, edits, and the like). The AIA 400 may use the data from user interactions (for example, accepted or dismissed suggestions) and feedback (for example, incorrect or irrelevant suggestion reports) to improve future suggestions (for example, by using this data to re-train / tune the corresponding machine, deep, or statistical learning models).

[0221] As further depicted in FIG. 5, the composition module 422 helps the user 404 compose communications with less effort and with greater speed, such as by generating (at least a portion of) communication content (for example, text / language) from a limited communication input 402, guiding the user by suggesting structure and / or placeholders to fill in content, and the like. It may also ensure the quality and effectiveness of communication by optimizing the generated language and structure for specific contexts and communication receivers. The composition module 422 may implement predictive language generation based on initial communication input 402 and available content and context 502, use case, and the like. It may suggest communication templates and generate textual completions, such as predicting the next word, phrase, sentence, paragraph, or an entire message / document / communication. It may present options for the user to confirm or select from. It may optimize generated language to a specific receiver(s) / audience and / or personalizeit to mimic the user’s communication style. The composition module 422 may use communication templates 504 for various communication scenarios, statistical algorithms for generating textual predictions (for example, machine learning, deep learning, language models, and the like), and other (prospective) forms of natural language generation. Language generation by the composition module 422 may optimize compositional suggestions to the user for utility and impact (effectiveness). Utility optimization can be achieved by maximizing the probability of the user 404 accepting the system-generated text (or one of the suggestions if multiple options are presented) and penalizing compositional selections for inapplicable, irrelevant, or too general suggestions (for example, penalizing templates with higher percent of placeholders vs. actual text, penalizing suggestions that apply to a broad range of use cases vs. the actual use case, and the like). Impact and effectiveness optimization with respect to the user’s communication goals can be achieved by maximizing the probability of a positive outcome or reaction by the receiver, given the context, use case, and the like, for the communication. The probability of a positive (desired) reaction may be estimated by a machine, deep, or other statistical learning model trained on a large set of reaction data (records of past communication content with meta-data such as context, senders, receivers / audience, and the actual reactions). Variations of the content generated by the composition module may be run through the reaction prediction model and filtered or ranked based on the predicted probability of the desired reaction before being presented to the user.

[0222] The system may accumulate a great amount of content (for example, messages, documents, and the like) with meta-data: context (for example, a resume for an engineering position at a startup or a marketing email in China) and senders / recipients (for example, identifiers, emails, or references to communication profiles). Receivers may be an audience (for example, IT recruiters and consumers in China). This data may be one general-purpose data set mixing everything or multiple specialized datasets, such as use case-specific (for example, only marketing emails) or receiver- specific (for example, email history of a specific user). Meta-data may also include the actual reactions (for example, the number of interview invitations for each resume, the number of clicks or product purchases triggered by each marketing email, the sentiment of a response to each user email, and the like). The system may train machine learning, deep learning, or other statistical models on this data and use these models in the reaction prediction / optimization module of the AIA. The system may have a model that predicts the reactions to a given user's message, context, and the receiver or audience (for example, returns the probability of different reactions. The system may also have a modelthat rewrites a given user's message to optimize for a desired reaction. It may take a draft of the message, context (including the desired reaction), and the receiver as input and produce a rewrite that maximizes the probability of the desired reaction. These models may be general- purpose (for example, trained on general-purpose data sets) or specialized (trained on specialized datasets). Afterward, the AIA 400 may capture the actual reaction to the user's communication, construct a reaction data record, and feed it into the model(s) to update / retrain them (for example, online learning).

[0223] Communication templates 504 may be especially applicable in frequent or repeated communication scenarios, such as short requests or replies, common communication types (for example, introductions, greetings, invitations, resumes, applications, or reports), and the like. Communication templates 504 may not be as applicable for creative writing, describing personal experience, communications containing complex ideas or argumentation, and the like. The composition module 422 may select communication templates 504 from clusters of semantically close or equivalent templates for different communication use cases. Communication templates 504 may consist of text and placeholders or a combination thereof. Some may be ready for immediate use (for example, short requests and replies), and some may be a sequence of placeholders or an outline defining structural and logical blocks that the user would have to flesh out. Communication templates may include short communications (for example, requests, replies, greetings, and the like), use-case specific templates (for example, separate for different channels, communication types, and other context categories / conditions; clusters of variations catering to different communication profiles, and the like). User-specific or group-specific communication templates may be frequently used by the user or by a group that the user belongs to or is targeting (for example, team- or company-specific templates). Communication templates 504 may contain a mix of manually created and automatically generated templates. Templates may be manually created either by the users (for example, administrators providing organization-specific templates for all users in the organization) or by AIA developers / maintainers (for example, a database of professionally edited templates for various common communication scenarios / use cases). Some communication templates 504 may be generated automatically by processing a large body of documents and messages, normalizing and canonicalizing them using different kinds of parsing, chunking or text similarity techniques, and then clustering equivalent ones (for example, with semi-supervised machine learning). In an embodiment, a hybrid approach to template curation may also beutilized, where people manually verify / edit automatically generated templates before making them available.

[0224] In an embodiment, selection of communication templates 504 may be implemented as a machine, deep, or other statistical learning model / technique that finds the most likely templates given the initial content, such as previous communications in the communication thread, and known context, such as a user’s intent. The template selector module 506 identifies a subset of communication templates to consider, such as based on available context, (for example, communication type, receivers, for example, team or organization) and applies a machine learning model to score these templates and find the highest scoring one(s). As a postprocessing step, the template selector may discard or penalize communication templates 504 that are too general and ensure diversity in a resulting set of communication template candidates, such as for further personalization and targeting and also to provide the user 404 a choice.

[0225] In an embodiment, automatically generating textual completions 508 may be implemented using machine, deep, or other statistical learning models conditioned on the prefix word sequence (for example, current communication draft), n previous communication(s) in the thread / conversation, additional context, and the like. They may generate the next word, phrase, sentence, paragraph, or an entire message or document. In an embodiment, representations of the previous content and context are fed into a natural language generation model that is trained on large-scale communication data that may be mixed with users’ prior communication and / or effective communications of other people like the user (for example, with respect to industry group, and the like), and / or effective examples of communication in the same or similar scenario (for example, with respect to communication type, use case, and the like) such as giving a higher weight to the mixed-in data. The outputs of the model are ranked based on the receiver’s communication profile and other contextual information. Output candidates can be ranked by the probability of them triggering desired reaction or outcome or based on their similarity to the user’s communication style. The highest-ranking outputs are suggested for the user to confirm / select. The user may confirm / select a completion to use as- is or modify.

[0226] The composition module 422 may utilize an automatic content fill function 512 in association with the appropriate placeholders utilizing the language generation facility, such as with filling in personal user data (for example, full name, contact details, addresses, and other data that the user may frequently share in communication), public facts or world knowledge(for example, filling in the full date when only a day of week is mentioned, filling in the public addresses of meeting places mentioned), and the like. It may also prompt the user to fill in certain placeholders in communication templates suggested by AIA.

[0227] The personalization and targeting facility 514, such as in conjunction with targeted and personalized content suggestions 518, may be able to provide personalization and targeting of new content, and filter generated language options (for example, select variations from a cluster of templates or textual completion candidates) based on the user’s communication profile (for example, to match their communication style) or based on the receiver’s communication profile (pick the ones that may appeal more to the receiver to increase the likelihood of a positive reaction). The personalization and targeting facility 514 may provide for reaction prediction 516, such as in conjunction with reaction data stored in communication profiles 428. The personalization and targeting facility 514 may also dynamically modify selected templates, such as by adjusting vocabulary and / or style to mimic the user’s communication and / or optimize for a specific receiver or audience. The personalization and targeting facility may use a style transfer module implemented with a machine, deep, or other statistical learning model or models, such as a sequence-to- sequence model or models that translate generated language variations from generic style into the user’s authentic communication style. The personalization and targeting facility may use a reaction prediction module implemented with a machine, deep, or other statistical learning model trained on reaction data (communication records enriched with meta-data including reactions of specific receivers) to estimate the probability of a positive reaction and filter or rank generated language variations based on that. Language generation models may also be trained on reaction-enriched data to generate variations that maximize the probability of positive reaction in the first place.

[0228] As further depicted if FIG. 6, the editing module 424 analyzes and modifies an existing fragment or the entire communication to optimize it for a specific receiver(s) or audience and use case. The editing module 424 helps the user 404 ensure the quality of their communication by analyzing existing language and generating changes that improve its accuracy, clarity, and effectiveness with respect to communication context and communication profiles. The editing module 424 may include features (checks) that analyze and improve different aspects of communication quality. At least some of these checks may be context- and communication party-dependent. Checks may utilize a rule-based and statistical module 602, a machine learning model(s) 602, deep learning-based models, and the like, or hybrid of different processing module functionality. The editing module 424 may generate modificationsand present them to the user 404 to review and confinn or reject individually. It may also modify existing language and present (mark or notify of) the changes it has made - and the user may be able to revert individual changes.

[0229] The editing module 424 may provide for context- and profile- dependent checks, such as with respect to accuracy, clarity, effectiveness, and the like. Accuracy checks may include spelling and grammar correction by applying the norms of a specific language dialect, such based on the receiver’s dialect, spelling, and grammar correction applied to different parameters or models. Accuracy checks may depend on the user’s communication profile (for example, depending on the user’s first language and / or language proficiency), and adjust a model’s parameters, such as using a model trained on a corpus of texts by the people with same first language / proficiency, resulting in an optimized check for a specific error distribution and frequency. Clarity and effectiveness may include preparing the communication in a way the receiver can listen to (for example, without shutting down, and building trust), such as by optimizing trigger words (for example, removing or replacing negative trigger words, adding more positive trigger words, and the like, such as described herein), prompting the user to add positive feedback that the receiver would appreciate (depending on the use case), replacing non-inclusive language that may make the receiver feel excluded, flagging aggressive language (for example, suggesting the user to wait for a later time to send the communication), applying the style preferred by the receiver, for example, formal vs. informal, brief vs. detailed, and the like. Clarity and effectiveness may include preparing the communication in a way the receiver can understand (for example, building clarity), such as by improving the communication for readability / clarity (for example, adjusting readability (sentence and word complexity) and vocabulary to the receiver’s language proficiency level (including native / non- native status), adjusting for formatting based on receiver’s preferences, predicting reaction of the receiver (for example, in the case of direct communication) or audience (in case of a broadcast communication), warning the user when a negative or undesired reaction is likely (for example, highlighting content language that can cause negative emotions for a specific receiver or group, highlighting content that may not engage the receiver (for example, too long, irrelevant, and the like), warning when it’s not a good time to contact the receiver (for example, they are unavailable, busy, sick, stressed, such as based on the recent communication history, calendar integration, explicit receiver settings, observed receiver state, and the like), modifying communication to increase the likelihood of a positive reaction (for example, through recommendations, suggesting changes, or a complete rewrite)), and the like. The Al A 400 mayuse reaction prediction like the one provided in the composition module 422 to rank and filter to generate edits. For instance, if predicted reactions do not match desired reactions and / or match undesired reactions, the AIA 400 may warn the user and suggest modifications to the draft, for example to highlight suboptimal fragments (for example, identified by the reaction prediction model), offer vocabulary suggestions (for example, word replacements) that maximize the likelihood of a desired reaction, offer structural changes (for example, size, outline, transitions, paragraph splitting, and the like) that maximize the probability of desired reaction, offer a complete rewrite of the draft, and the like. In an embodiment, the AIA 400 may make corrections to the communication automatically, without user intervention.

[0230] The AIA 400 may provide for a reaction capture module 426 to capture the receiver’s reaction 410 to improve the effectiveness of generated or modified communications using reaction data, such as through updating communication profiles, collecting reaction data for training (in batch or online mode) the models that predict reactions and optimize communication for certain reactions, and the like. The AIA’s reaction capture module 426 may capture reactions 410 through a communication platform (for example, responses, time to respond, sentiment of responses, and the like) or integrations with other systems (for example, link clicks, or other receiver actions triggered by the communication). Reaction capture module 426 may capture reactions 410 directly from the receiver (for example, if the receiver is also an AIA 400 user), ask the user 404 to report the receiver reaction 410 as the outcome of a certain communication, and the like. Receiver reaction data may also be obtained and processed in a batch outside of the AIA 400 workflow (for example, extracted through an integration or an API from an external system, from an archive of prior communication, and the like).

[0231] The AIA 400 may include a reaction prediction and optimization facility that may utilize reaction data, such as consisting of records of communication acts (ent, p, ext, ro), where ent is content (for example, message, document, text, audio, video, in raw form or included meta data such as for formatting, markup, annotations, tags, and the like), p = (s, r) are references / identifiers of communication parties, where s is a user sender and r is a user receiver / audience (for example, r may be an individual or a group or a category of receivers, s may be a human or a computational system capable of generating natural language, s, r may include properties and characteristics of the parties or link to communication profiles of the parties, r may include the type of relationship between the recipient(s) and the sender, ext is context (for example, may include the use case, user goals, references to communicationhistory (such as previous communications in a thread) or related documents, current psychoemotional state of communication parties, their availability, location, and the like), ro is a reaction or outcome that the content ent triggered with the receiver(s) r. Reaction / outcome ro may be a response or no response, time to response, number of subsequent communications in the thread (for example, before the goal of communication was achieved), sentiment or type of the response (positive / negative, confirmation / rej ection), receiver emotions, use case-specific response (for example, containing certain elements / words), receiver actions (for example, clicks on the link in the communication, signups), additions or deletions of participants in a thread or conversation, and the like. In case r is plural (a group / audience), ro may be an aggregate, such as the total number of receiver actions, positive and negative responses, clicks, and the like. Response data may be one or multiple data sets (for example, use case or receiverspecific). Reaction data may be used to train a machine, deep, or other statistic learning model that would predict the most likely reaction-outcome ro for a given (ent, p, ext).

[0232] Referring to FIG. 7, this reaction prediction 516 may be used to inform the user sender 404 of the likely reaction-outcome, such as from the user sender 404 send content 702 to a user receiver 412. The user receiver 412 then produces a reaction 704 which is then captured 706. The captured reaction 706 may produce reaction data 708 to be stored, as well as being provided for reaction prediction 516. Referring to FIG. 8, the reaction prediction may be used to rank different variations of content 810 using a ranking module 806 given a desirable / acceptable set of reactions / outcomes, and the like. In an embodiment, the reaction prediction 516 may interact with a reaction module consumer 802, such as for language generation and / or sender output processing.

[0233] FIG. 8 shows the reaction module consumer 802, which is responsible for language generation and sender input processing. Reaction module consumer 802 is configured to create content variations and send these variations, along with the context and recipient / audience information, to the reaction prediction 516 within a reaction module. The reaction prediction 516 is configured to evaluate the generated content variations, predicting the likely reactions based on data about the recipient / audience and context. Following this, the content variations are ranked by the ranking module 806 based on the predicted and expected reactions. These ranked content variations, accompanied by their scores, are then returned to the reaction module consumer 802 for further processing.

[0234] Simultaneously, the actual reactions 812 from the user receiver 412 are captured by the capturing the reaction module 808. Reactions 812 are stored as reaction data 804, whichserves multiple purposes. This data is utilized for batch learning, where large datasets are used to train models offline, and for online learning, where new reaction data continuously updates and improves the predictive models such as reaction prediction 516.

[0235] The reaction prediction 516 can then inform the user sender 404 of the likely reaction outcomes, optimizing the content for maximum effectiveness. The reaction data 804 can be used in training models that generate or modify content to achieve desired reactions. These models can be general-purpose or tailored to specific contexts or receivers, ensuring the communication is contextually appropriate and receiver specific. Content 810, refined and optimized based on the reaction data, is then delivered to the user receiver 412. This optimized content aims to achieve the desired reactions, and the cycle continues with the capture of new reactions, creating a feedback loop that perpetually enhances communication effectiveness.

[0236] Thus, FIG. 8 shows an interconnected system where the reaction module consumer 802 generates and processes content, the reaction prediction 516 predicts and ranks reactions, and the capturing module 808 gathers real-time reaction data. This data is stored and utilized by the reaction data 804 for continuous model training and content optimization.

[0237] Referring to FIG. 9, reaction data may also be used to train a machine learning model such as language generation or modification model 904 that generates output content 902 or modifies content in a way that maximizes the probability of a certain (set of) ro. For instance, given the communication parties p, context ext, target reaction(s) ro, and the space of possible content variations CNT, the model could find: ent* = argmax P (enti I p, ext, ro), enti 6 CNT

[0238] The models may be either general-purpose or context-, receiver- or reactionspecific. For example, context- specific models may be trained on reaction data sets consisting of records of communication acts in a certain use case. Receiver-specific models may be trained on communication addressed to a specific receiver or audience and representing their reactions. The models may be used in Al-assisted composition, editing, or communication analysis applications. The models may be trained offline (for example, batch learning using large reaction data sets) or online (AIA’ s reaction capture module may feed new reaction data records into existing models to improve their accuracy). The models may run on the server side (for example, in the cloud) or may be embedded on communication devices.

[0239] In an embodiment, the AIA 400 may provide a receiver with a modified version of a received communication, such as based on the communication profde of the user and the preferences of the receiver, such as described herein. The receiver may benefit from suchmodification if the user has characteristics and behaviors that causes the communication to be non-optimal in some way, such as because the user tends to communicate in an emotional manner, the user has a language characteristic that makes understanding difficult (for example, the user is a non-native speaker, uses idioms that are unfamiliar to the receiver, uses slang that is offensive to the receiver), and the like. The AIA 400 may monitor communications being received, and flag those that may need modification, where, such as based on a user’s preferences, the communication may be modified before it is presented to the receiver. In an embodiment, the AIA 400 may indicate that the communication has been modified, indicate what portions of the communication has been modified, show the modifications in an annotated version of the original communication, present both a clean modified version and an annotated version of the original communication, and the like. Based on past communications from a user, and past reactions and preferences by the receiver, the AIA 400 may implement modifications to a communication without first notifying the receiver. Alternately, the receiver may be notified that a modified version is available, where the receiver may be offered options for reading or listening to the communication. In an embodiment, the AIA 400 may implement modifications based primarily on the receiver’ s communication profile, such as regardless of the user / source of the communication (for example, modify / augment anything the receiver reads / hears, including incoming communication, online or printed content, and the like). Based on the receiver’s language proficiency, the AIA 400 may adjust readability (for example, simplify vocabulary, rephrase idioms, and the like). Based on the receiver’s environment, preferences, available time, and the like, the AIA 400 may condense / summarize content. If the AIA 400 knows the receiver is sensitive to specific content (for example, based on his / her communication profile), the AIA 400 may alert the receiver or mask this content. In the process, the AIA 400 may observe and collect receiver reactions to various types of content as training data, learn from their feedback and interactions with its functionality that augments the content receivers consume, and the like.

[0240] In an embodiment, the AIA 400 may augment an incoming communication to a receiver, such as through a user interface, for transforming incoming communications (for example, emails, instant message, and the like), shared documents, or other content that receivers consume, into a more digestible form. Transformations of incoming communications may include removing wordy or redundant language that does not carry meaning so that reading takes less time and effort; such transformations can take the form of classification machine or deep learning algorithms to strike out words or phrases or sequence to sequencealgorithms which automatically convert one text to another (for example, neural machine translation or paraphrase). Data employed ranges may be transformed from manually or automatically curated lists of commonly overused phrases (either generally, by domain or by communication party) to parallel data sets which consist of original wordy sentences or documents and their less wordy rewrites.

[0241] Transformations may provide summarizing longer messages or documents for quick preview. Possible applications include, and are not limited to, providing a summary or set of key points for long emails, messages, or documents, or automatically generating more concise versions of long passages to make it easier for the reader to understand, and the like. Methods may include machine and deep learning techniques which can leverage parallel corpora of long sentences or passages with their summarized rewrites, as well as leveraging vast amounts of unlabeled data.

[0242] In an example embodiment, and referring to FIG. 10, a communication platform or application 1002 may receive an original incoming communication with meta-data 1004, which may be classified with a classification module 1006 and evaluated for triggers with an optimization trigger module 1008, which my take input data from stored user preferences and communication profiles 1010. A transform module 1012 may provide a plurality of transformations, as described herein, to produce an optimized incoming communication.

[0243] Transformations may include formatting texts and adding structure, such as highlighting key phrases, splitting long paragraphs, converting lists into bullet points. Techniques may include using machine learning or deep learning or other statistical techniques and treating the problem as a classification or sequence labeling task.

[0244] Transformations may include rewriting messages or documents using vocabulary and readability (for example, splitting long sentences; replacing long or rare words with shorter, more common synonyms; replacing idioms with literal / universally understandable equivalent phrases, and the like) to a level that the user would understand, depending on their language proficiency level and background. Methods may include using manually and / or automatically curated dictionaries and reference sources as part of rule or statistical approaches. Methods may leverage a user’s profile, user and conversation context and knowledge, and the like, of both conversation parties to automatically identify (for example, using techniques such as machine or deep learning) types of language, and then employ rule or statistical techniques to generate the appropriate the context.

[0245] Transformations may include rewriting messages or documents using style, tone, argumentation, coherence, and structure, and the like, that the user would understand and react positively to, such as depending on the context, goals, and recipient background. Methods may include machine and deep learning techniques which can leverage parallel corpora of long sentences or passages with their summarized rewrites, as well as leveraging vast amounts of unlabeled data.

[0246] Transformations may include masking or rephrasing the language that the user would not like to see (for example, offensive language, negative triggers, cliches, and the like) of certain words based on user's stated preferences or previously observed reactions. Such transformations may take the form of classification machine or deep learning algorithms to strike out words or phrases or sequence to sequence algorithms which automatically convert one text to another, such as neural machine translation or paraphrase. Data employed ranges from manually or automatically curated lists of commonly overused phrases may be employed (either generally, by domain, by communication party, and the like) to parallel data sets which consists of phrases, sentences, or documents with the target language and their corrections.

[0247] Transformations may include adding missing context, such as explaining terms, abbreviations, slang, idioms, and the like, that can be unfamiliar to the user. Methods may include using manually and automatically curated dictionaries and reference sources as part of rule or statistical approaches. Methods may leverage the user’ s profile, user and conversation context and knowledge of both conversation parties to automatically identify (for example, using techniques such as machine or deep learning) parts of communication which need more context, and then employ rule or statistical techniques to generate the appropriate the context.

[0248] Transforming incoming communications may methods include, but are not limited to, rule-based, machine learning, deep learning, or other statistical supervised, semi-supervised and unsupervised techniques. Readers may choose message or document types that they would like to be automatically modified. Alternatively, users may request the system to modify specific messages or documents on demand, such as by applying specific filters or transformations. Users may be able to view the original version even if the message / document was automatically modified.

[0249] In an embodiment, and with reference to FIG. 11 , the AIA provides feedback for communication interactions, including audio and / or video communications. The process begins with receiving raw or preprocessed audio / video data at step 1102 and capturing and processing it at step 1104. At step 1106, participants, verbal and non-verbal content, andcontext are identified. Following this, at step 1108, the AIA maintains a digital representation of the conversation, which is continuously updated at step 1110 to reflect the conversation thread. The AIA then analyzes incoming and outgoing messages at step 11 12, providing realtime assistance and generating feedback at step 1114 to improve communication effectiveness.

[0250] In an embodiment, the user may be able to specify whether they want to transform incoming communications automatically or manually, showing the original before or after transformation, and the like. In a non-limiting example, and referring to FIG. 12, based on receiver preferences 1202, at first process step 1204, the user interface may intercept an incoming communication to a receiver and classify the communication, such as by a type of communication, who the user is, and the like. In a second process step 1206, the user interface may decide whether to transform the communication, such as automatically based on user preferences or decided by the receiver based on an interaction between the receiver and the user interface. In a third process step 1208, the user interface determines that the incoming communication should not be immediately transformed, showing the receiver the original communication, and offering the ability to transform. In a fourth process step 1210, the user interface determines that the incoming communication should be immediately transformed before displaying the communication to the receiver and offering to also show the receiver the original communication. This is meant to be merely a non-limiting example of how the AIA 400 may be configured to transform an incoming communication, illustrating some basic features that may be provided to the user. Transformation types, such as described herein, may also be selectable by the user through a user interface, such as improvements in readability, creating greater conciseness, providing summarization, formatting the communication, providing language cleanup, injecting context into the communication, and the like.

[0251] Augmenting incoming communications to a user may be provided through a user interface, such as providing the ability for the user to modify the communication after it’ s sent, ability for the receiver to interact with and provide feedback on or react to any part of the communication, ability for the receiver to transform incoming communications, collecting user preferences in the context of incoming communication, such as through iterative questions and tracking reactions, and the like. The user interface may send the content of the communications as a link reference to an object stored on a server (for example, in the cloud) that renders on the client side in an interactive form allowing: 1) a user receiving a communication to interact with and provide feedback on the communication or any part of the communication and 2) a user sending a communication to modify the communication at any point after it was sent (forexample, if the user thinks of something else, finds a mistake, calms down and wishes they wouldn't have set the communication, and the like). This would enable users to control their content over time, even when sent through third-party applications (for example, email applications), where users could change and / or delete their content at any point in the future and provide a receiver with an option to modify communications they receive, such as based on their preferences.

[0252] In an embodiment, a sender portion of the user interface may provide specific feedback to help users compose communications and to increase effectiveness by targeting communications to the receiver specific profiles and preferences, such as with respect to vocabulary, style, context, accuracy, the best time to send the communication, and the like.

[0253] In an embodiment, a receiver portion of the user interface may display communications in an interactive format (for example, embedded or a pop-up web page object) allowing receivers to interact with and provide feedback on the communication even if they are not using the AIA 400 client application, which may also provide users a chance to modify the communication after it’s sent. For instance, receivers may react to incoming communications by highlighting any fragment and selecting an emoji or writing a comment to it that would be immediately available to the user (for example, the user would get notified and be able to see the emoji and / or comment by opening the communication). If nothing is highlighted, the reaction may be applied to the entire communication. If the communication was sent to multiple people or shared with a group, the receiver may choose to share the reaction only with the user, with all receivers, (for example, everyone on the thread), with a subset of receivers, or keep it private but visible to AIA. AIA 400 may process the reactions and use them to improve the rules and models optimizing incoming communication for the receiver and assisting the user(s) in communication with the receiver.

[0254] The user interface may enable the receiver to transform incoming messages, providing reader (receiver) with the ability to modify communications they receive, such as with the transformations described herein. Each transformation type, such as described herein, may be provided with a separate filter that can be turned on or off. Additionally, the receiver may specify the types of incoming communication they would like the filters to be applied to automatically, while other communications are modified on demand by applying specific filters. Alternatively, users may request the system to modify specific incoming communications (for example, messages, documents, and the like, as described herein) on demand, such as by applying specific filters and / or transformations. Users may be able to viewthe original version of an incoming message even if the communication was automatically modified. In the process, user preferences may be collected in the context of incoming communications through iterative questions and tracking of reactions.

[0255] In an embodiment, the system may send a message (for example, content) over a network, sending a link to the message that is stored on AIA 400 servers or cloud. If the receivers have the AIA 400 installed (for example, as a browser or communication application extension), they will see the actual message instead of the link (for example, AIA 400 will display the actual message instead of the link). The receivers that don't have the AIA 400 installed, may only see the link but will be able to follow this link to view the message and interact with it (for example, in the browser). The sender may modify the message at any time, and all the receivers may be enabled to immediately see the updated version. If some receiver has an optimized version of the message open (for example, a feature of the AIA 400 augmenting incoming communication), this optimized version may automatically re-generate when the sender modifies the original message. Receivers may interact with the message and provide instant feedback (for example, by selecting a portion of the message and reacting with an emoticon or a comment). Receivers may control who sees the feedback, such as only the sender, everybody on the thread, or just the receiver and AIA (for example, so that AIA 400 could better optimize future messages for the receiver). Additionally, the users may have all the AIA 400 features for optimizing outgoing communication and augmenting incoming communication.

[0256] In an embodiment, the user interface may deliver iterative questions to a user with respect to the receiver’s communication preferences and / or feedback on specific communications to hone the AIA’s understanding of each receiver’s preferences. These preferences may be used by the system to modify incoming communications to the receiver and / or to help the sending user(s) optimize their communication with the receiver. Receivers may be motivated to provide feedback so that the AIA 400 could better transform communications to their liking. Receivers may choose to share their feedback on incoming communications with users to provide receivers with improved versions of the modified communications of some communications they receive based on the shared feedback and preferences provided.

[0257] In transforming an incoming communication, the AIA 400 may record preferences, it may intercept incoming communication, classify it, and decide whether it should be transformed before being displayed to the reader and what transformations should be applied,as based on the receiver’s preferences. Then filters can then be run based on preferences (for example, may be a combination of rule -based and machine-leaming-based filter implementations). The AIA 400 can then present the transformed version to the reader based on the preferences. In an embodiment, user preferences may enable the selection of other functions, such as the receiver being able to switch between the transformed and the original versions (for example, by clicking a link or tab), be able to highlight fragments and react with an emoji or provide a comment, request to modify a specific communication that AIA 400 did not automatically modify, and the like. In an embodiment, AIA 400 may learn to modify communications of a certain type based on what the receiver requests to modify most frequently. In that case, AIA 400 may ask whether the receiver wants to always (or never) modify similar communications, or communication from certain users.

[0258] In an embodiment, the user interface and user experience may be implemented through integration with third-party platforms and communication channels (for example, as an extension or a plug-in), as a feature of a communication platform or client (for example, built into a messenger, email app, and the like), through a mobile platform (for example, communication may be composed in a dedicated mobile app), stored in the cloud, and the like. Implementation may utilize a link sent through a communication channel (for example, text message), where when the receiver clicks the link, it may launch a dedicated app or open the communication in a web browser. If the receiver is not on the platform or doesn’t have the extension / app, they could get a link that would open the communication in a web browser, such as with interactive features.

[0259] If the sender modifies a communication that has already been sent, it may be updated on all clients that have it open. Synchronization between the server-side copy and the ones displayed to the sender and the receivers may be implemented using technology such as operational transformation, differential synchronization or conflict-free replicated data types. Transformed receiver-side versions may also be re-generated.

[0260] In an embodiment, the user interface may capture receiver preferences, such as for contributing to their profiles, through asking a set of questions / presenting settings during user onboarding (for example, how they would like the texts to be modified, what they would not want to see, which kind of communication they would like to always or never modify, and the like), through prompting the user to provide more answers and settings later, through tracking their communication and reactions and suggesting settings that match their communication style / behavior and reactions, and the like.

[0261] The user interface may also provide for tracking communication and content consumption habits and preferences for a receiver, providing insights and recommendations for better productivity and communication impact.

[0262] In an embodiment, the AIA 400 may augment outgoing voice and video messages. For instance, the AIA 400 may intercept a voice or a video message, check it for correctness, clarity, and effectiveness in the background, and alert the user in case any significant issues are found, and suggest improvements (for example, via text or voice output). In the process, the system may ask the user for additional details to get more context and identify a receiver if necessary for targeted or more effective suggestions. The AIA 400 may augment conversations, such as to help the user understand the reactions of the receiver(s), interpret hidden and non-verbal signals, suggest what to say next to mitigate or prevent possible tension or misunderstanding, suggest adjusting vocabulary or talking speed based on the receiver’s language proficiency, and the like. The AIA 400 may augment incoming communication, such as transforming incoming communications or other content the receiver consumes into a more digestible form. For instance, the AIA 400 may summarize longer texts and messages so that processing them takes less time, rewrite texts and messages using vocabulary and readability level that the receiver would understand (for example, depending on his or her language proficiency level and background), rewrite texts and messages with respect to communication structure (for example, prose vs. bullet listings), and the like.

[0263] In an embodiment, the AIA 400 may provide the ability to augment real-time conversations being carried out through augmented reality (AR) or virtual reality (VR) platforms through an AIA AR / VR communication facility, which provides communication assistance (augmented communication) to users (senders as well as receivers). The AR / VR communication facility may be implemented through an application or application extension for a third-party AR / VR platform or a feature of an AR / VR platform, and support both asynchronous communication (improving communications that are outgoing, recorded, dictated, and the like) and real-time conversations. Although the AR / VR communication facility is described primarily with respect to AR and VR technology platforms, and the sender’s point of view, the AR / VR communication facility may work with any wearable device including a microphone to capture dictated or recorded communication(s), a privacy output display capability to provide assistance and suggestions to the user (for example, an AR or VR display, smart eyeglasses, heads-up display, smart contact lenses, head-worn virtual retina display, and the like), and an in-ear speaker. Optionally, other functional resources may beutilized by the AR / VR communication facility, such as a camera to read body signals in a conversation, identify or recognize specific people in a conversation, recognize speech (from mouth movements), recognize eye movements, recognize written text, intercept video communications, and the like. Body sensors (implantable or not) may be used to capture emotions, level of stress, and the like, that may be associated with communication. A braincomputer interface may be used to capture reactions, intentions, thoughts, and the like. In an embodiment, the AR / VR communication facility may be implemented with a smart speaker to intercept voice communications, analyze them, and either translate them into better-formed text and then read back or display them in a target application for confirmation or iterations. The AR / VR communication facility may provide feedback / suggestions via voice or a companion application display (for example, smartphone, smart-glasses, and the like) before finalizing and sending a recorded communication. The AR / VR communication facility may be implemented with a VR platform to augment conversations between people as avatars in virtual reality through capturing and improving recorded communication and assisting in realtime conversations. In an embodiment, observed communication can be tracked and analyzed to generate insights and provide feedback to the user over time, such as processed on the device, processed offline, processed in a hybrid client-server system (for example, partially processed on the device), and the like, such as with offline or embedded models.

[0264] In an embodiment, the AR / VR communication facility may augment outgoing recorded or asynchronous voice or video communication. For instance, the AR / VR communication facility may intercept a voice or a video communication the user has recorded through a smart speaker or a smart device camera (for example, the ability to connect to a network via different wireless protocols can operate to some extent...

Claims

CLAIMSWhat is claimed is1. A method of electronic communication assistance comprising: receiving a partial electronic communication at an artificial intelligence assistant computing facility from a first electronic identifier associated with a first user, the partial electronic communication comprising a communication content associated with the first electronic identifier associated with the first user and a second electronic identifier associated with a second user; extracting a communication context from the partial electronic communication; encoding the partial electronic communication for processing, creating an encoded partial electronic communication; retrieving from a communication profile database a first communication profile for the first user using the first electronic identifier associated with the first user, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user using the second electronic identifier associated with the second user, wherein the second communication profile comprises a second user communication attribute; processing the encoded partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using at least one of the communication context, the first user communication attribute, or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

2. The method of claim 1 further comprising transmitting the changed electronic communication to the first electronic identifier associated with the first user, and / or transmitting the changed electronic communication to the second electronic identifier associated with the second user.

3. The method of claim 1, wherein the compositional change is derived from representations of previous content and context from a plurality of user profiles stored in thecommunication profile database, which are like at least one of the first communication profile or the second communication profile.

4. The method of claim 1, wherein the processor is trained on large-scale data mixed with prior communication and effective communications from a plurality of user profiles.

5. The method of claim 1 , wherein the processor performs operations based on instructions corresponding to a machine learning model for creating the compositional change.

6. The method of claim 1 , wherein the compositional change is an auto-generated textual completion; the auto-generated textual completion being a phrasal completion, and wherein the processor is configured to generate the compositional change by optimizing generated language as determined by the processor from the second user communication attribute.

7. The method of claim 1, wherein the processor is configured to generate the compositional change by replicating a communication style of the first user as determined by the processor from the first user communication attribute.

8. The method of claim 1, wherein the partial electronic communication includes a communication goal, and the processor is configured to generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal.

9. The method of claim 8, wherein the processor is configured to generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion.

10. The method of claim 9, wherein the processor is configured to select the communication template using a machine learning model to find a most effective communication template based at least in part on the communication content.

11. The method of claim 10, wherein the plurality of communication templates includes at least one automatically generated template generated by the processor.

12. The method of claim 10, wherein the processor is configured to select the communication template by using the machine learning model to score the plurality of communication templates based at least in part on the communication content, first user communication attribute, second user communication attribute, or communication context.

13. A method of electronic communication assistance comprises: intercepting an electronic communication from further transmission at an artificial intelligence assistant computing facility, wherein the electronic communication was transmitted from a first electronic identifier associated with a first user to second electronic identifier associated with a second user, the electronic communication comprising a communication content and comprising or associated with the first electronic identifier associated with the first user; extracting a communication context from the electronic communication; encoding the electronic communication for processing creating an encoded electronic communication; retrieving from a communication profile database a first communication profile for the first user using the first electronic identifier, wherein the first communication profile comprises a first user communication attribute; retrieving from the communication profile database a second communication profile for the second user using the second electronic identifier, wherein the second communication profile comprises a second user communication attribute; processing the encoded electronic communication with a processor to generate a compositional change for the communication content of the electronic communication using at least one of the communication context, the first user communication attribute, or the second user communication attribute to generate the compositional change; and generating a changed electronic communication from the electronic communication and the compositional change, wherein the changed electronic communication comprises annotations to indicate the compositional change to the electronic communication.

14. The method of claim 13, wherein the compositional change is derived from representations of previous content and context from a plurality of user profiles stored in the communication profile database, which are like at least one of the first communication profile or the second communication profile.

15. The method of claim 13, wherein the electronic communication includes a communication goal, and the processor is configured to generate the compositional change by optimizing for impact and effectiveness of generated language with respect to the communication goal.

16. The method of claim 15, wherein the processor is configured to generate the compositional change further using a communication template selected from a plurality of communication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion.

17. A server computer comprising a processor and a computer-readable storage device that stores instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving a partial electronic communication at an artificial intelligence assistant computing facility from an electronic identifier associated with a user, the partial electronic communication comprising a communication content and comprising or associated with the electronic identifier; retrieving a communication profile, wherein the communication profile comprises a user communication attribute; processing the partial electronic communication with a processor to generate a compositional change for the communication content of the partial electronic communication using the user communication attribute to generate the compositional change; and generating a changed electronic communication from the partial electronic communication and the compositional change.

18. The server computer of claim 17, wherein the processor is configured to generate the compositional change further using a communication template selected from a plurality ofcommunication templates comprising at least one of prepared text or placeholder locations for defining structural elements for user completion.

19. The server computer of claim 18, wherein the plurality of communication templates may include at least one automatically generated template generated by the processor.

20. The server computer of claim 19, wherein the processor is configured to select the communication template by using a machine learning model to score the plurality of communication templates based at least in part on the communication content, communication context, first user communication attribute, or second user communication attribute.

Images