Generating suggested messages

The use of a machine learning model on a computing device to generate suggested messages addresses the challenge of composing coherent and relevant messages, enhancing communication efficiency and quality by reducing manual input and error, particularly benefiting users with disabilities.

WO2024233670A1PCT designated stage expired Publication Date: 2024-11-14GOOGLE LLC
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
PCT/US2024/028381
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-09
Filing Date
2024-05-08
Publication Date
2024-11-14

AI Technical Summary

Technical Problem

Users face challenges in efficiently composing coherent and contextually relevant messages, particularly in communication scenarios where manual typing is time-consuming and prone to errors, especially for individuals with disabilities or impairments.

Method used

A computing device employs a machine learning model to generate suggested messages based on a message thread, allowing users to select and send contextually appropriate responses, thereby reducing the effort and time required for message composition.

Benefits of technology

The solution streamlines communication by providing high-quality, contextually relevant messages, reducing the need for manual input and minimizing errors, while also considering user feedback to improve message quality and engagement.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure US2024028381_14112024_PF_FP_ABST
    Figure US2024028381_14112024_PF_FP_ABST
Patent Text Reader

Abstract

A method includes obtaining, by a computing device, a message thread. The computing device generates a set of suggested messages responsive to a recent message in the message thread by at least applying a machine learning model. Information in each of the set of suggested messages is related to information in the message thread. The computing device outputs, for display, at least a subset of suggested messages from the set of suggested messages. The computing device receives a user input that selects one suggested message from at least the subset of suggested messages as a selected message. The computing device populates a message field with the selected message.
Need to check novelty before this filing date? Find Prior Art

Description

GENERATING SUGGESTED MESSAGESCROSS-REFERENCE TO RELATED APPLICATION

[0001] This application claims the benefit of US Provisional Patent Application No. 63 / 501,050, filed May 9, 2023, which is incorporated by reference herein in its entirety.BACKGROUND

[0002] Users may communicate with each other via one or more computing devices. In some cases, the incoming communications (e.g., an email, a text message, a meeting request, etc.) may include textual information. A computing device may enable a user of a computing device to respond to the incoming communication by allowing the user to input textual information (e.g., using an input device), and send the textual information to other computing devices as a response.SUMMARY

[0003] Techniques are described by which a computing device may use a machine learning model to generate suggested messages, thereby facilitating communication between users. For example, a first computing device associated with a first user may apply a machine learning model to generate a set of messages based on a message thread between the first user and a second user. The first computing device may display at least a subset of the suggested messages to the first user. Responsive to the first user selecting one of the suggested messages, the first computing device may populate a message field with the selected message. In this way, the techniques of this disclosure may provide various advantages, such as saving time (e.g., by reducing the effort needed to compose a coherent and relevant message) and improving message quality (e.g., by leveraging a machine learning model to generate a high-quality message that is contextually appropriate, well-structured, and free of grammatical errors). In turn, the techniques of this disclosure may reduce in fewer interactions with and less resources (e.g., power, processor cycles, memory space, memory bus bandwidth, wireless bandwidth) expended by the first computing device.

[0004] In some examples, a method includes obtaining, by a computing device, a message thread. The message thread includes a set of messages between a first user associated with the computing device and a second user. The method further includesdetermining, by the computing device, a set of information types of a message from the set of messages by at least applying a machine learning model, and outputting, by the computing device and for display, at least a subset of information types from the set of information types. The method further includes receiving, by the computing device, a user input that selects one suggested information type from at least the subset of information types as a selected information type. The method further includes generating, by the computing device, a set of suggested messages responsive to the message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to the selected information type and information in the message thread. The method further includes outputting, by the computing device and for display, at least a subset of suggested messages from the set of suggested messages, and receiving, by the computing device, a user input that selects one suggested message from at least the subset of suggested messages as a selected message. The method further includes populating, by the computing device, a message field with the selected message.

[0005] In some examples, a computing device includes one or more processors. The computing device further includes a memory that stores instructions that, when executed by the one or more processors, cause the one or more processors to obtain a message thread. The message thread includes a set of messages between a first user associated with the computing device and a second user. The instructions further cause the one or more processors to determine a set of information types of a message from the set of messages by at least applying a machine learning model, and output, for display, at least a subset of information types from the set of information types. The instructions further cause the one or more processors to receive a user input that selects one suggested information type from at least the subset of information types as a selected information type. The instructions further cause the one or more processors to generate a set of suggested messages responsive to the message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to the selected information type and information in the message thread. The instructions further cause the one or more processors to output, for display, at least a subset of suggested messages from the set of suggested messages, and receive a user input that selects one suggested message from at least the subset of suggested messages as a selected message. The instructions further cause the one or more processors to populate a message field with the selected message.

[0006] In some examples, a computer-readable storage device is encoded with instructions that, when executed by one or more processors of a first computing device, cause the one or more processors to obtain a message thread. The message thread includes a set of messages between a first user associated with the computing device and a second user. The instructions further cause the one or more processors to determine a set of information types of a message from the set of messages by at least applying a machine learning model, and output, for display, at least a subset of information types from the set of information types. The instructions further cause the one or more processors to receive a user input that selects one suggested information type from at least the subset of information types as a selected information type. The instructions further cause the one or more processors to generate a set of suggested messages responsive to the message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to the selected information type and information in the message thread. The instructions further cause the one or more processors to output, for display, at least a subset of suggested messages from the set of suggested messages, and receive a user input that selects one suggested message from at least the subset of suggested messages as a selected message. The instructions further cause the one or more processors to populate a message field with the selected message.

[0007] In some examples, a computer program product for suggested message generation includes at least one non-transitory computer-readable medium including one or more instructions that, when executed by at least one processor, cause the at least one processor to obtain a message thread comprising a set of messages between a first user associated with a computing device and a second user. The one or more instructions further cause the at least one processor to determine a set of information types of a message from the set of messages by at least applying a machine learning model. The one or more instructions further cause the at least one processor to output, for display, at least a subset of information types from the set of information types, and receive a user input that selects one suggested information type from at least the subset of information types as a selected information type. The one or more instructions further cause the at least one processor to generate a set of suggested messages responsive to the message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to the selected information type and information in the message thread. The one or more instructionsfurther cause the at least one processor to output, for display, at least a subset of suggested messages from the set of suggested messages, and receive a user input that selects one suggested message from at least the subset of suggested messages as a selected message. The one or more instructions further cause the at least one processor to populate a message field with the selected message.

[0008] In some examples, an apparatus includes means for obtaining, by a computing device, a message thread including a set of messages between a first user associated with the computing device and a second user. The apparatus further includes means for generating, by the computing device, a set of suggested messages responsive to a message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to information in the message thread. The apparatus further includes means for outputting, for display, at least a subset of suggested messages from the set of suggested messages. The apparatus further includes means for receiving a user input that selects one suggested message from at least the subset of suggested messages as a selected message. The apparatus further includes means for populating a message field with the selected message.

[0009] The details of one or more examples are set forth in the accompanying drawings and the description below. Other features, objects, and advantages will be apparent from the description and drawings, and from the claims.BRIEF DESCRIPTION OF DRAWINGS

[0010] FIG. 1 is a conceptual diagram illustrating an example computing environment and graphical user interface for providing suggested messages in accordance with one or more techniques of this disclosure.

[0011] FIG. 2 is a block diagram illustrating one example of a computing device in accordance with one or more techniques of this disclosure.

[0012] FIG. 3 is a block diagram illustrating an example operation of a machine learning model in accordance with techniques of this disclosure.

[0013] FIG. 4 is a flow diagram illustrating an example operation for providing suggested messages in accordance with one or more techniques of this disclosure.DETAILED DESCRIPTION

[0014] FIG. 1 is a conceptual diagram illustrating an example computing environment 1 and a graphical user interface (“GUI 116”) for providing suggested messages relevant to a message thread, in accordance with one or more techniques of this disclosure. As shown in the example of FIG. 1, computing environment 1 includes one or more computing devices, such as computing devices 100A-100N (collectively, “computing devices 100”). Examples of computing devices 100 may include, but are not limited to, portable, mobile, or other devices, such as mobile phones (including smartphones), wearable devices (including smartwatches), laptop computers, desktop computers, tablet computers, vehicle head units (e.g., infotainment systems), smart television platforms, server computers, mainframes, etc. In some examples, computing devices 100 may represent a cloud computing system that provides one or more services via the network. That is, in some examples, computing devices 100 may be a distributed computing system. In any case, each of computing devices 100 may be substantially similar such that description of one may apply equally to any other.

[0015] As shown in the example of FIG. 1, computing device 100 A, includes one or more user interface (UI) devices (“UI device 102”). UI device 102 of computing device 100A may be configured to function as an input device and / or an output device for computing device 100A. UI device 102 may be implemented using various technologies. For instance, UI device 102 may be configured to receive input from a user through tactile, audio, and / or video feedback. Examples of input devices include a presence-sensitive display, a presence-sensitive or touch-sensitive input device, a mouse, a keyboard, a voice responsive system, video camera, microphone or any other type of device for detecting a command from a user. In some examples, a presence-sensitive display includes a touch- sensitive or presence-sensitive input screen, such as a resistive touchscreen, a surface acoustic wave touchscreen, a capacitive touchscreen, a projective capacitive touchscreen, a pressure sensitive screen, an acoustic pulse recognition touch screen, or another presence-sensitive technology. That is, UI device 102 of computing device 100A may include a presence-sensitive device that may receive tactile input from a user of computing device 100 A. UI device 102 may receive indications of the tactile input by detecting one or more gestures from the user (e.g., when the user touches or points to one or more locations of UI device 102 with a finger or a stylus pen).

[0016] UI device 102 may additionally or alternatively be configured to function as an output device by providing output to a user using tactile, audio, or video stimuli.Examples of output devices include a sound card, a video graphics adapter card, or any of one or more display devices, such as a liquid crystal display (LCD), dot matrix display, light emitting diode (LED) display, microLED, miniLED, organic light-emitting diode (OLED) display, e-ink, or similar monochrome or color display capable of outputting visible information to a user of computing device 100A. Additional examples of an output device include a speaker, a haptic device, or other device that can generate intelligible output to a user. For instance, UI device 102 may present output to a user of computing device 100 A as a graphical user interface that may be associated with functionality provided by computing device 100A. In this way, UI device 102 may present various user interfaces of applications executing at or accessible by computing device 100A (e.g., an electronic message application, an Internet browser application, etc.). A user of computing device 100 A may interact with a respective user interface of an application to cause computing device 100 A to perform operations relating to a function.

[0017] In some examples, UI device 102 of computing device 100 A may detect two- dimensional and / or three-dimensional gestures as input from a user of computing device 100A. For instance, a sensor of UI device 102 may detect the user's movement (e.g., moving a hand, an arm, a pen, a stylus, etc.) within a threshold distance of the sensor of UI device 102. UI device 102 may determine a two- or three-dimensional vector representation of the movement and correlate the vector representation to a gesture input (e.g., a hand-wave, a pinch, a clap, a pen stroke, etc.) that has multiple dimensions. In other words, UI device 102 may, in some examples, detect a multidimensional gesture without requiring the user to gesture at or near a screen or surface at which UI device 102 outputs information for display. Instead, UI device 102 may detect a multidimensional gesture performed at or near a sensor which may or may not be located near the screen or surface at which UI device 102 outputs information for display.

[0018] In the example of FIG. 1, computing device 100 A includes user interface (UI) module 104, application module 106A-106N (collectively “application module 106”). Modules 104 and 106 may perform operations described herein using hardware, software, firmware, or a mixture thereof residing in and / or executing at computing device 100A. Computing device 100 A may execute modules 104 and 106 with one processor or with multiple processors. In some examples, computing device 100 A may execute modules 104 and 106 as a virtual machine executing on underlying hardware. Modules 104 and106 may execute as one or more services of an operating system or computing platform or may execute as one or more executable programs at an application layer of a computing platform.

[0019] UI module 104, as shown in the example of FIG. 1, may be operable by computing device 100 A to perform one or more functions, such as receive input and send indications of such input to other components associated with computing device 100 A, such as application module 106. UI module 104 may also receive data from components associated with computing device 100A such as application module 106. Using the data received, UI module 104 may cause other components associated with computing device 100 A, such as UI device 102, to provide output based on the data. For instance, UI module 104 may receive data from one of application modules 106 to display a GUI.

[0020] Application modules 106, as shown in the example of FIG. 1, may include functionality to perform any variety of operations on computing device 100 A. For instance, application modules 106 may include a word processor, a text application, a web browser, a multimedia player, a calendar application, an operating system, a distributed computing application, a graphic design application, a video editing application, a web development application, or any other application. One of application modules 106 (e.g., application module 106 A) may be a text messaging or Short Message Service (SMS) application. For example, application module 106 A may include functionality to compose outgoing text message communications, receive incoming text message communications, respond to incoming text message communications, and other functions. Application module 106 A, in various examples, may provide data to UI module 104 causing UI device 102 to display a GUI.

[0021] In some examples, one or more of application module 106 may be operable to receive incoming communications from other devices (e.g., via the network). For instance, one or more of application module 106 may receive text messages for an account associated with a user of computing device 100 A, calendar alerts or meeting requests for a user of computing device 100 A, or other incoming communications.

[0022] In general, communications may include information (e.g., generated in response to input by users of the other devices). Examples of information include text (e.g., any combination of letters, words, numbers, punctuation, etc.), emoji, emoticons, images, animations, video, audio, or any other content that may be included in a communication. In the example of FIG. 1, application module 106 A may receive an incomingcommunication (e.g., a text message) from another computing device (e.g., computing device 100N).

[0023] Computing devices 100 may communicate with each other via a network. The network may include any public or private communication network, such as a cellular network, Wi-Fi network, a direct cell-to-satellite communication network, or other type of network for transmitting data between computing devices. In some examples, the network may represent one or more packet switched networks, such as the Internet. Computing devices 100 may send and receive data across the network using any suitable communication techniques. For example, computing devices 100 may each be operatively coupled to the network using respective network links. The network may include network hubs, network switches, network routers, etc., that are operatively inter-coupled thereby providing for the exchange of information between computing devices 100. In some examples, network links of the network may be Ethernet, ATM or other network connections. Such connections may include wireless and / or wired connections.

[0024] In general, a first user may use a communication platform provided by application module 106 to electronically communicate with a second user via computing devices 100. For example, computing device 100 A, which may be associated with the first user, may send messages to and receive messages from computing device 100N, which may be associated with the second user. When typing a message, the first user may manually provide inputs (e.g., to select letters, emojis, etc.) to computing device 100A. However, manually typing a message may be inconvenient or difficult. For example, manually typing a text message can be time-consuming, which may result in slower communication and frustration. Manually typing may also increase the likelihood of typos and autocorrect errors, leading to miscommunication or confusion. Additionally, some users, such as individuals with disabilities or impairments, may find manually drafting a text message to be particularly challenging, if not impossible.

[0025] In accordance with techniques of this disclosure, computing device 100A (and other computing devices 100) may include a suggestion module 108 that provides a set of suggested messages relevant to a message thread between various users. Suggestion module 108 may represent or otherwise apply an ML model 138 trained to generate contextually relevant and coherent messages based on one or more messages in the message thread. Suggestion module 108 may suggest the generated messages to the first user of computing device 100 A who may then select one of the suggested messages. Responsive to the selection, computing device 100 A may send the selected message tocomputing device 100N. In this way, the techniques may streamline communication between the first user and the second user (e.g., by reducing the number of interactions with computing device 100 A and the time required to type a message to send to computing device 100N).

[0026] Computing device 100 A may include suggestion module 108, which may include or otherwise apply ML model 138, such as a large language model. ML model 138 is described in greater detail below (e.g., with respect to FIG. 2). Briefly, however, ML model 138 may be trained on a large amount of data in order to generate coherent and contextually relevant messages. In some examples, ML model 138 may use complex algorithms and neural networks to analyze the structure, grammar, and context of language. ML model 138 may perform various natural language processing (NLP) tasks, such as machine translation, sentiment analysis, summarization, question answering, text generation, etc.

[0027] Suggestion module 108 may obtain a message thread 110 between a first user associated with first computing device 100 A and a second user who may be associated with second computing device 100N or, in some examples, also associated with first computing device 100A. Message thread 110 may include a set (i.e., one or more) of messages 112A-112N (collectively, “messages 112”) exchanged between two or more users (e.g., the first user and the second user) within a communication platform (e.g., provided by application modules 106), such as a messaging application, a social media platform, etc. Only message 112A of messages 112 is shown and labeled in the example of FIG. 1 for ease of illustration. In some examples, message thread 110 may represent a chronological sequence of messages.

[0028] In general, suggestion module 108 may send information (e.g., incoming communications, location information, other contextual information, etc.) to ML model 138 only if computing device 100 receives permission from the user to send the information. For example, in situations discussed here in which computing device 100 may collect, transmit, or may make use of personal information about a user (e.g., incoming communications) the user may be provided with an opportunity to control whether programs or features of computing device 100 can collect user information (e.g., information about a user’s communications, a user’s social network, a user’s social actions or activities, a user’s profession, a user’s preferences, or a user’s current location), or to control whether and / or how computing device 100 may store and share user information.

[0029] In addition, certain data may be treated in one or more ways before it is stored, transmitted, or used so that personally identifiable information is removed. For example, a user’s identity may be treated so that no personally identifiable information can be determined about the user. Thus, the user may have control over how information is collected about the user and stored, transmitted, and / or used in accordance with techniques of this disclosure.

[0030] ML model 138 may use a recent message as input to generate suggested information types 114A-114N (collectively, “information types 114”). An information type of a message may be or otherwise include a topic, theme, point, subject, purpose, intent, keyword, question, answer, related words, related ideas, etc. Only information types 114A-114C of information types are shown and labeled in the example of FIG. 1 for ease of illustration.

[0031] ML model 138 may use a recent message in message thread 110 to prompt the generation of content, such as information types 114. For example, as shown in FIG. 1, a recent message may be message 112A from the second user stating, “Dinner later with the group around 8?” Suggestion module 108 may provide message 112A to ML model 138, which ML model 138 may then use as input to generate suggested information types 114 that are consistent with the input’s context.

[0032] UI module 104 may output, for display by UI device 102, a graphical user interface 116 (“GUI 116”). Responsive to ML model 138 generating information types 114, UI module 104 may cause GUI 116 to include at least a subset of information types 114. The first user may review information types 114 included in GUI 116 and select one of information types 114 (e.g., information type 114A). That is, application module 106 may receive a user input that selects one of information types 114 as a selected information type.

[0033] ML model 138 may use the selected information type to guide the generation of suggested messages 117A-117N (collectively, “suggested messages 117”). Only suggested messages 117A-117C of messages 117 are shown and labeled in the example of FIG. 1 for ease of illustration. For example, the user of computing device 100 may select information type 414A, which is the answer “yes” to the question “Dinner later with the group around 8?” ML model 138 may then generate suggested messages 117 that are responsive to recent message 112 A. In general, messages 117 may include text, emoticons, images, reactions, animations, videos, etc.

[0034] Information in each suggested message of the set of suggested messages 117 may be related to the selected information type, which in the example of FIG. 1 is information type 114 A. Information in each suggested message of the set of suggested messages 117 may also be related to information in message thread 110. For example, ML model 138 may generate suggested message 117A that states, “Yep, 8 sounds great!” ML model 138 may also generate suggested message 117B that states, “Dinner with the group sounds good and works for me.” ML model 138 may also generate suggested message 117C that states “Dinner sounds great!”

[0035] In general, suggested messages 117 may be responsive to the recent message. For example, suggested messages 117 may accurately address the subject matter, specific details, and the purpose of the recent message (and, optionally, to other portions of message thread 110). Additionally, suggested messages 117 may maintain coherence and continuity in a conversation by using information contained in message thread 110. That is, information in each of suggested messages 117 may be related to information in message thread 110.

[0036] Responsive to ML model 138 generating suggested messages 117, UI module 104 may cause GUI 116 to include at least a subset of suggested messages 117. In some examples, GUI 116 may include a message selection area 118 that includes the subset of suggested messages 117. For example, message selection area 118 may be displayed in response to application modules 106 receiving the subset of suggested messages 117 from suggestion module 108. Message selection area 118 may include the subset of suggested messages 117, such as suggested messages 117A-117C.

[0037] The first user may review the subset of suggested messages 117 included in GUI 116 and select one of suggested messages 117. In turn, application module 106 may receive a user input that selects one of suggested messages 117 as a selected message. Application module 106 may then populate a message field 120 with the selected message, enabling the first user to modify and / or send the selected message.

[0038] In some examples, GUI 116 may include feedback graphical elements 156. A user may interact with (e.g., select) one or more of feedback graphical elements 156 to provide feedback relating to suggested messages 117. ML model 138 may be trained using the user feedback. For example, ML model 138 may be trained to maximize positive user feedback and minimize negative user feedback.

[0039] Generating and sending one of suggested messages 117 based on a selection of suggested information types 114 may include one or more advantages. For example,while a user could reasonably send one of suggested information types 114 as a message to another user, suggested information types 114 may be too concise and direct, potentially leading to the impression that the user of computing device 100 doesn't care about the conversation. In other words, a very short message can be interpreted as dismissive, suggesting that the user sending the message is not interested in engaging in a conversation, even if the sender did not intend it that way. Thus, ML model 138 may generate suggested messages 117 such that a length of each suggested message of the set of suggested messages 117 is greater than a length of a selected information type. In this way, the techniques may help a user avoid being perceived as impolite or thoughtless, thereby helping maintain positive relationships and fostering effective communication.

[0040] Additionally, by enabling computing device 100 A to obtain, display, and send suggested messages 117, suggestion module 108 may reduce the amount of effort required for the first user of computing device 100 A to input and send a message to other users. For example, a user may simply tap or otherwise select one of suggested messages 117 instead of having to manually type or use voice recognition (which may be prone to errors and typos) to input a response. As a result, the techniques of this disclosure may not only decrease the time to compose messages but also improve the overall quality of messages.

[0041] Although primarily described herein as being sent by another user, such as a second user, a recent message may be by the first user of computing device 100. For example, suggestion module 108 may suggest a message as a follow-up to a message sent by the first user instead of as a response to a message from a second user. Thus, the example conversation illustrated in FIG. 1 is provided for purposes of explanation and is not intended to limit the scope of this disclosure.

[0042] Also, although primarily described herein as being associated with a second computing device (e.g., computing device 100N), a second user may be associated with the same computing device as the first user. For example, the first user and the second user may be communicating with each other asynchronously using computing device 100 A. For example, the first user may send, using computing device 100 A (e.g., a smart home device) a message, such as “What do you want for dinner?” ML model 138 may use this message as input to suggest information types 114, such as “American,” “Italian,” “Chinese,” etc. Responsive to the first user selecting one of information types 114, ML model 138 may generate and output suggested messages 117. For example, if the first user selects “American,” ML model 138 may generate and output “How abouthamburgers at 7?”, “There’s a highly-rated American bistro near us ”, “I can make a reservation at a steakhouse.”, etc. In some examples, computing device 100A may be a remote I / O device.

[0043] FIG. 2 is a block diagram illustrating a computing device 200. Computing device 200 may be one example of computing devices 100 in accordance with one or more techniques of this disclosure. FIG. 2 illustrates only one particular example of computing device 200, and many other examples of computing device 200 may be used in other instances and may include a subset of the components included in example computing device 200 or may include additional components not shown in FIG. 2.

[0044] As shown in FIG. 2, computing device 200 may include one or more user interface devices 202 (“UI device 202”), one or more processors 222 (“processor 222”), one or more storage devices 234 (“storage device 224”), one or more communication units 226 (“communication unit 226”). As also shown in FIG. 2, UI device 202 may include one or more input devices 228 (“input device 228”) and one or more output devices 230 (“output device 230”). As also shown in FIG. 2, storage device 224 may include a user interface module 204 (“UI module 204”), one or more application modules 206 (“application module 206”), a suggestion module 208, an operating system 234 (“OS 234”), a database 236, a ML model 238, and a model trainer 240.

[0045] In some examples, UI device 202 may be a presence-sensitive display configured to detect input (e.g., touch and non-touch input) from a user of respective computing device 200. UI device 202 may output information to a user in the form of a UI, which may be associated with functionality provided by computing device 200. Such UIs may be associated with computing platforms, operating systems, applications, and / or services executing at or accessible from computing device 200 (e.g., electronic message applications, chat applications, Internet browser applications, mobile or desktop operating systems, social media applications, electronic games, menus, and other types of applications).

[0046] Processor 222 may implement functionality and / or execute instructions within computing device 200. For example, processor 222 may receive and execute instructions that provide the functionality of modules 204-208 and OS 234. These instructions executed by processor 222 may cause computing device 200 to store and / or modify information within storage device 224 of processor 222 during program execution. Processor 222 may execute instructions of modules 204-208 and OS 234 to perform oneor more operations. That is Modules 204-208 and OS 234 may be operable by processor 222 to perform various functions described herein.

[0047] Storage device 224 within computing device 200 may store information for processing during operation of computing device 200 (e.g., computing device 200 may store data accessed by modules 204-208 and OS 234 during execution at computing device 200). In some examples, storage device 224 may be a temporary memory, meaning that a primary purpose of storage device 224 is not long-term storage. Storage device 224 on computing device 200 may be configured for short-term storage of information as volatile memory and therefore not retain stored contents if powered off. Examples of volatile memories include random access memories (RAM), dynamic random access memories (DRAM), static random access memories (SRAM), and other forms of volatile memories known in the art.

[0048] Storage device 224 may include one or more computer-readable storage media. Storage device 224 may be configured to store larger amounts of information than volatile memory. Storage device 224 may further be configured for long-term storage of information as non-volatile memory space and retain information after power on / off cycles. Examples of non-volatile memories include magnetic hard disks, optical discs, floppy discs, flash memories, or forms of electrically programmable memories (EPROM) or electrically erasable and programmable (EEPROM) memories. Storage device 224 may store program instructions and / or information (e.g., within database 236) associated with modules 204-208 and OS 234.

[0049] Communication unit 226 of computing device 200 may communicate with one or more external devices via one or more wired and / or wireless networks by transmitting and / or receiving network signals on the one or more networks. Examples of communication units 226 include a network interface card (e.g. such as an Ethernet card), an optical transceiver, a radio frequency transceiver, a GPS receiver, or any other type of device that can send and / or receive information. Other examples of communication unit 226 may include short wave radios, cellular data radios, wireless network radios, as well as universal serial bus (USB) controllers.

[0050] Input device 228 of computing device 200 may receive input. Examples of input are tactile, audio, and video input. Input device 228 of computing device 200, in one example, includes a presence-sensitive display, a fingerprint sensor, touch-sensitive screen, mouse, keyboard, voice responsive system, video camera, microphone or any other type of device for detecting input from a human or machine.

[0051] Input devices 228 may include one or more sensors. Numerous examples of sensors exist and include any input component configured to obtain environmental information about the circumstances surrounding computing device 200 and / or physiological information that defines the activity state and / or physical well-being of a user of computing device 200. In some examples, a sensor may be an input component that obtains physical position, movement, and / or location information of computing device 200. For instance, sensors may include one or more location sensors (e.g., GNSS components, Wi-Fi components, cellular components), one or more temperature sensors, one or more motion sensors (e.g., multi-axial accelerometers, gyros), one or more pressure sensors (e.g., barometer), one or more ambient light sensors, and one or more other sensors (e.g., microphone, camera, infrared proximity sensor, hygrometer, and the like). Other sensors may include a heart rate sensor, magnetometer, glucose sensor, hygrometer sensor, olfactory sensor, compass sensor, step counter sensor, to name a few other non-limiting examples.

[0052] Output device 230 of computing device 200 may generate one or more outputs. Examples of outputs are tactile, audio, and video output. Output device 230 of computing device 200, in one example, includes a presence-sensitive display, sound card, video graphics adapter card, speaker, liquid crystal display (LCD), or any other type of device for generating output to a human or machine.

[0053] Communication channels 232 (“COMM channel 232”) may interconnect each of the components 202, 222, 224, and 226 for inter-component communications (physically, communicatively, and / or operatively). In some examples, communication channel 232 may include a system bus, a network connection, an inter-process communication data structure, or any other method for communicating data.

[0054] Computing device 200 may include OS 234. OS 234 may control the operation of components of computing device 200. For example, OS 234 may facilitate the communication of modules 204-208 with processor 222, storage device 224, and communication units 226. In some examples, OS 234 may manage interactions between software applications (e.g., application module 206) and a user of computing device 200. OS 234 may have a kernel that facilitates interactions with underlying hardware of computing device 200 and provides a fully formed application space capable of executing a wide variety of software applications having secure partitions in which each of the software applications executes to perform various operations. In some examples, UI module 204 may be considered a component of OS 234.

[0055] In the example of FIG. 2, suggestion module 208 may include a hardware device having various hardware, firmware, and software components. However, FIG. 2 illustrates only one particular example of suggestion module 208, and many other examples of suggestion module 208 may be used in accordance with techniques of this disclosure. In some examples, components of suggestion module 208 may be located in a singular location. In other examples, one or more components of suggestion module 208 may be in different locations (e.g., connected via the network of FIG. 1). That is, in some examples suggestion module 208 may be part of a conventional computing device, while in other examples, suggestion module 208 may be part of a distributed or “cloud” computing system.

[0056] Suggestion module 208 may apply a machine learning model, such as ML model 238, to generate information types and suggested messages relevant to a message thread. In some examples, ML model 238 can be or include one or more artificial neural networks (also referred to simply as neural networks). A neural network can include a group of connected nodes, which also can be referred to as neurons or perceptrons. A neural network can be organized into one or more layers. Neural networks that include multiple layers can be referred to as “deep” networks. A deep network can include an input layer, an output layer, and one or more hidden layers positioned between the input layer and the output layer. The nodes of the neural network can be connected or non-fully connected.

[0057] ML model 238 can be or include one or more transformer-based neural networks, such as a large language model (LLM). Transformer-based neural networks may refer to a type of deep learning architecture specifically designed for handling sequential data, such as text or time series. In other words, transformer-based neural networks like LLMs may be configured to perform natural language processing (NLP) tasks, such as questionanswering, machine translation, text summarization, and sentiment analysis. Transformerbased neural networks may utilize a self-attention mechanism, which allows the model to weigh the importance of different elements in a given input sequence relative to each other. The self-attention mechanism may help ML model 238 effectively capture long- range dependencies and complex relationships between elements, such as words in a sentence.

[0058] ML model 238 may include an encoder and a decoder that operate to process and generate sequential data, such as text. Both the encoder and decoder may include one or more of self-attention mechanisms, position-wise feedforward networks, layernormalization, or residual connections. In some examples, the encoder may process an input sequence and create a representation that captures the relationships and context among the elements in the sequence. The decoder may then obtain the representation generated by the encoder and produce an output sequence. In some examples, the decoder may generate the output one element at a time (e.g., one word at a time), using a process called autoregressive decoding, where the previously generated elements are used as input to predict the next element in the sequence.

[0059] ML model 238 may generate a set of suggested messages by determining a set of information types of a recent message. An information type of a message may be or otherwise include a topic, theme, point, subject, purpose, intent, keyword, question, answer, related words, related ideas, etc. In some examples, ML model 238 may determine the information type by leveraging a self-attention mechanism to capture the relationships and dependencies between words in the input sequence. For example, ML model 238 may tokenize (e.g., split) a sequence of words or subwords, which ML model 238 may convert into vectors (e.g., numerical representations) that ML model 238 can process. ML model 238 may use the self-attention mechanism to weigh the importance of each token in relation to the others. In this way, ML model 238 may identify patterns and relationships between the tokens, and in turn the words corresponding to the tokens, that indicate one or more information types of the recent message.

[0060] ML model 238 may generate suggested messages based on the determined one or more information types of the recent message. For example, information in each of the suggested messages generated by ML model 238 may be related (e.g., logically, semantically, etc.) to at least a subset of the set of information types of the recent message. For example, if ML model 238 determines that the set of information types of a recent message includes “vacation,” “swimming,” and “weekend,” ML model 238 may generate a suggested message like “Do you want to go to the beach this weekend?” and other similar messages.

[0061] Although primarily described herein as being a transformer-based neural network, ML model 238 may be or otherwise include one or more other types of neural networks. For example, ML model 238 may be or include an autoencoder. In some examples, the aim of an autoencoder is to learn a representation (e.g., a lower- dimensional encoding) for a set of data, typically for the purpose of dimensionality reduction. For example, in some examples, an autoencoder can seek to encode the input data and then provide output data that reconstructs the input data from the encoding. Recently, the autoencoder concepthas become more widely used for learning generative models of data. In some examples, the autoencoder can include additional losses beyond reconstructing the input data. ML model 238 may be or include one or more other forms of artificial neural networks such as, for example, deep Boltzmann machines, deep belief networks, stacked autoencoders, etc. Any of the neural networks described herein can be combined (e.g., stacked) to form more complex networks.

[0062] In some examples, ML model 238 can be or include one or more feed forward neural networks. In feed forward networks, the connections between nodes do not form a cycle. For example, each connection can connect a node from an earlier layer to a node from a later layer. In some examples, ML model 238 can be or include one or more recurrent neural networks. In some examples, at least some of the nodes of a recurrent neural network can form a cycle.

[0063] Recurrent neural networks can be especially useful for processing input data that is sequential in nature. For example, a recurrent neural network can pass or retain information from a previous portion of the input data sequence to a subsequent portion of the input data sequence through the use of recurrent or directed cyclical node connections. Sequential input data may include words in a sentence (e.g., for natural language processing, speech detection or processing, etc.). In some examples, sequential input data can include time-series data (e.g., sensor data versus time or imagery captured at different times). Example recurrent neural networks may include long short-term memory (LSTM) recurrent neural networks, gated recurrent units, bi-direction recurrent neural networks, continuous time recurrent neural networks, neural history compressors, echo state networks, Elman networks, Jordan networks, recursive neural networks, Hopfield networks, fully recurrent networks, sequence-to- sequence configurations, etc.

[0064] In some examples, ML model 238 can be or include one or more convolutional neural networks. In some examples, a convolutional neural network can include one or more convolutional layers that perform convolutions over input data using learned filters. Filters can also be referred to as kernels. Convolutional neural networks can be especially useful for vision problems such as when the input data includes imagery such as still images or video. However, convolutional neural networks can also be applied for natural language processing.

[0065] Computing device 200 may include model trainer 240 that trains (e.g., pre-train, fine-tune, etc.) ML model 238. Model trainer 240 may pre-train ML model 238 on a large and diverse corpus of text from various sources, such as websites, books, articles, andother text repositories. This dataset may cover a wide range of topics and domains to ensure ML model 238 learns diverse linguistic patterns and contextual relationships. Model trainer 240 may train ML model 238 to optimize an objective function. The objective function may be or include a loss function, such as cross-entropy loss, that compares (e.g., determines a difference between) output data generated by the model from the training data and labels (e.g., ground-truth labels) associated with the training data. For example, the objective function of ML model 238 may be to correctly predict the next word in a sequence of words or correctly fill in missing words as much as possible.

[0066] In some examples, ML model 238 may determine a message value for each suggested message of a set of suggested messages. The message value may indicate a predicted utility of a suggested message to a user for each of the suggested messages. In these examples, ML model 238 may compare the message value for each of the suggested messages to determine whether to display the suggested messages to a user. For example, for each of the suggested messages, ML model 238 may determine a message value from 0 to 1. In this example, a message value of 0 may be associated with a lowest confidence in the message being useful or of interest to a user, and a message value of 1 may be associated with a highest confidence in the message being useful or of interest to a user. It should be understood that other ranges of message values (e.g., 0 to 100, 0 to -10, etc.), relationships between message values and amount of confidence (e.g., a message value of 0 being associated with a highest confidence and a message value of 1 being associated with a lowest confidence), and / or the like are contemplated.

[0067] If the message value of a suggested message satisfies a message value condition, ML model 238 (or another module of computing device 200, such as application module 206), may classify the suggested message as belonging to the subset of suggested messages to be outputted to the user. In some examples, the message value may satisfy the message value condition when the message value satisfies a message value threshold. For example, if ML model 238 determines, for a particular suggested message, a message value of 0.8, and if the message value threshold is 0.7, then the particular suggested message may be in the subset of suggested messages that ML model 238 transmits to UI module 204 for display to a user.

[0068] On the other hand, if ML model 238 determines a message value that does not satisfy the message value threshold, such as a message value of 0.6, then the particular suggested message may not be in the subset of suggested messages, and computingdevice 200 may execute a different action, such as discard the particular suggested message. In some examples, UI module 204 may output, for display by UI device 202, at least the subset of suggested messages in an order based on the message value for each suggested message of at least the subset of suggested messages.

[0069] In some examples, rather than comparing the message values to a message value threshold, computing device 200 may compare the message values to each other. For example, ML model 238 (or another module of computing device 200) may classify only the suggested messages having the highest message values as belonging to the subset of suggested messages. In some examples, the subset of suggested messages may only include a relatively small number of suggested messages (e.g., 2, 3, 4, etc.) to avoid overwhelming the user of computing device 200 with options.

[0070] Similarly, ML model 238 may determine a type value for each information type of a set of information types. The type value may indicate a predicted utility of a suggested information type to a user for each of the information types. In these examples, ML model 238 may compare the type value for each of the information types to determine whether to display the information type to a user. ML model 238 may compare the type values of the information types to a value type threshold to determine satisfaction of a value type condition for outputting the information types to a user. Additionally or alternatively, ML model 238 may compare the type values of the information types to each other and only output the subset of information types having the highest type values.

[0071] In general, ML model 238 may excel at performing NLP tasks, such as generating text and other content. However, with respect to specific types of content (e.g., specific information types), ML model 238 may have an increased likelihood of generating false or inaccurate information. To address the issue of generating false information, ML model 238 may be configured to exclude the generation of content relating to a set of excluded information types. The set of excluded information types may include one or more of phone numbers, addresses, web addresses, etc.

[0072] In some examples, model trainer 240 may fine-tune ML model 238 by using the feedback in the training process. For example, UI device 202 may receive a user input that selects feedback (e.g., thumbs up, thumbs down, etc.) relating to at least one suggested message of the set of suggested messages. In some examples, the feedback may indicate whether the suggested messages are accurate or inaccurate, correct or incorrect, high quality or low quality, etc. Model trainer 240 may obtain this feedback from the user of computing device 200 (as well as feedback from other users) and use the feedback fortraining. For example, model trainer 240 may convert the feedback into labeled data for supervised training. Additionally or alternatively, model trainer 240 may fine-tune ML model 238 by monitoring the relationship between the performance of ML model 238 and user feedback, and iterate the fine-tuning process as necessary (e.g., to receive more positive user feedback and less negative user feedback). In this way, the techniques of this disclosure may establish a feedback loop that continuously improves the quality of the output of ML model 238.

[0073] By leveraging ML algorithms, suggestion module 208 may decrease the time and / or effort required to compose a message to send to another user while increasing the likelihood that suggested messages will actually be of value to the user. That is, by presenting the user of computing device 200 with a set of suggested messages, the techniques of this disclosure may reduce the likelihood that the user will need to manually type a message.

[0074] FIG. 3 is a block diagram illustrating an example operation of a machine learning model in accordance with techniques of this disclosure. As shown in FIG. 3, an example training process 350 is used to train a ML model 338, which may be similar if not substantially similar to ML model 238. ML model 338 is trained on training data 351 that includes input data 352 that has labels 353. Training processes 350 is one example training process; other training processes may be used as well.

[0075] In some examples, ML model 338 may be trained by optimizing an objective function, such as objective function 355. For example, in some examples, objective function 355 may be or include a loss function, such as cross-entropy loss. One or more of various optimization techniques may be performed to optimize objective function 355. For example, the optimization technique(s) may minimize or maximize objective function 355. Example optimization techniques include Hessian-based techniques and gradientbased techniques, such as, for example, coordinate descent; gradient descent (e.g., stochastic gradient descent); subgradient methods; etc. Other optimization techniques include black box optimization techniques and heuristics.

[0076] In some examples, backward propagation of errors may be used in conjunction with an optimization technique (e.g., gradient based techniques) to train ML model 338 (e.g., when ML model 338 is a multi-layer model such as an artificial neural network). For example, an iterative cycle of propagation and model parameter (e.g., weights) update may be performed to train ML model 338. Example backpropagation techniques include truncated backpropagation through time, Levenberg- Marquardt backpropagation, etc.

[0077] In some examples, ML model 338 described herein may be trained using unsupervised learning techniques to increase the effectiveness of ML model 338. Unsupervised learning may include inferring a function to describe hidden structure from unlabeled data. For example, a classification or categorization may not be included in the data. Unsupervised learning techniques may be used to produce ML models capable of performing clustering, anomaly detection, learning latent variable models, or other tasks.

[0078] ML model 338 may be trained to increase the effectiveness of ML model 338 using semi-supervised techniques which combine aspects of supervised learning and unsupervised learning. ML model 338 may be trained or otherwise generated through evolutionary techniques or genetic algorithms. In some examples, ML model 338 described herein may be trained using reinforcement learning. In reinforcement learning, an agent (e.g., model) may take actions in an environment and learn to maximize rewards and / or minimize penalties that result from such actions. Reinforcement learning may differ from the supervised learning problem in that correct input / output pairs are not presented, nor sub-optimal actions explicitly corrected.

[0079] In some examples, ML model 338 described herein may include or otherwise be impacted by a number of hyperparameters, such as, for example, learning rate, number of layers, number of nodes in each layer, number of leaves in a tree, number of clusters; etc. Hyperparameters may affect model performance. Hyperparameters may be hand selected or may be automatically selected through application of techniques such as, for example, grid search; black box optimization techniques (e.g., Bayesian optimization, random search, etc.); gradient-based optimization; etc. Example techniques and / or tools for performing automatic hyperparameter optimization include Hyperopt; Auto-WEKA; Spearmint; Metric Optimization Engine (MOE); etc.

[0080] In some examples, various techniques may be used to optimize and / or adapt the learning rate when the model is trained. Example techniques and / or tools for performing learning rate optimization or adaptation include Adagrad; Adaptive Moment Estimation (ADAM); Adadelta; RMSprop; etc. In some examples, transfer learning techniques may be used to provide an initial model from which to begin training of ML model 338 described herein.

[0081] In some examples, ML model 338 described herein may be included in different portions of computer-readable code on a computing device. In one example, ML model 338 may be included in a particular application or program and used (e.g., exclusively) by such particular application or program. Thus, in one example, a computing device mayinclude a number of applications and one or more of such applications may contain its own respective machine learning library and ML model(s).

[0082] In another example, ML model 338 described herein may be included in an operating system of a computing device (e.g., in a central intelligence layer of an operating system) and may be called or otherwise used by one or more applications that interact with the operating system. In some examples, each application may communicate with the central intelligence layer (and model(s) stored therein) using an application programming interface (API) (e.g., a common, public API across all applications).

[0083] In some examples, the central intelligence layer may communicate with a central device data layer. The central device data layer may be a centralized repository of data for the computing device. The central device data layer may communicate with a number of other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, and / or additional components. In some examples, the central device data layer may communicate with each device component using an API (e.g., a private API).

[0084] FIG. 4 is a flowchart illustrating example operations of an example computing device in accordance with one or more aspects of the present disclosure. Although the example operation of FIG. 4 is described as being performed by computing device 200 of FIG. 2 and in the context of FIG. 1, in other examples some or all of the example operation may be performed by another computing device.

[0085] In accordance with techniques of this disclosure, suggestion module 208 of computing device 200 may obtain (with the user’s explicit consent) message thread 110 (or another message thread) (400). Message thread 110 may include a set of messages exchanged between a user associated with computing device 200 and a user associated with another computing device (e.g., computing device 100N). The messages may be exchanged via a communication platform provided by application module 206.

[0086] Suggestion module 208 may provide at least a subset of the set of messages (e.g., at least a portion of message thread 110) to ML model 238, which may be an LLM (402). ML model 138 may use a recent message from the messages as input to generate suggested information types 114. Responsive to ML model 238 generating information types 114, UI module 204 may output, for display by UI device 202, at least a subset of information types (404). For example, UI module 204 may cause GUI 116 to include at least a subset of information types 114, such as information types 114A-114C. The first user may review information types 114 included in GUI 116 and select one ofinformation types 114 (e.g., information type 114A). That is, application module 206 may receive a user input that selects one of information types 114 as a selected information type (406).

[0087] ML model 138 may use the selected information type as input to generate suggested messages 117 that are responsive to recent message 112A (or another recent message that may be by the user of computing device 200 or another user) in message thread 110 (408). ML model 238 may generate suggested messages 117 such that information in each of suggested messages is related to the selected information type and information in message thread 110. Suggested messages 117 may include text, emojis, emoticons, images, reactions, animations, videos, etc. a length of each suggested message of the set of suggested messages 117 may be greater than a length of the selected information type. In some examples, ML model 238 may exclude the generation of content relating to a set of excluded information types, such as phone numbers, addresses, web addresses, etc.

[0088] In some examples, ML model 238 may determine a message value that indicates a predicted utility of a suggested message to a user for each of suggested messages 117. Computing device 200 may display a subset of suggested messages 117 based on the message values of suggested messages 117. For example, ML model 238, or another module of computing device 200 (e.g., application module 206) may classify each suggested message that has a message value satisfying a message value condition as belonging to the subset of suggested messages. In other words, the suggested messages that have message values satisfying a message value condition may belong to a subset of suggested messages 117 for display to a user of computing device 200. In some examples, satisfaction of the message value condition may be based on whether the message value for a suggested message of suggested messages 117 satisfies a threshold message value.

[0089] UI module 204 may output, for display by UI device 202, at least a subset of suggested messages 117 (410). For example, GUI 116 may include suggested messages 117A-114C. In some examples, the order of at least the subset of suggested messagesl 14 may be based on the message value for each suggested message of at least the subset of suggested messages 117. Application module 206 may receive a user input that selects one suggested message from at least the subset of suggested messages 117 (412).Application module 206 may then populate message field 120 (or another message field) with the selected message (414).

[0090] In some examples, model trainer 240 may receive user input that selects feedback relating to at least one suggested message of the set of suggested messages 117. Model trainer 240 may train ML model 238 using the feedback.

[0091] This disclosure includes various examples, such as the following examples.

[0092] Example 1 : A method includes obtaining, by a computing device, a message thread comprising a set of messages between a first user associated with the computing device and a second user; determining, by the computing device, a set of information types of a message from the set of messages by at least applying a machine learning model; outputting, by the computing device and for display, at least a subset of information types from the set of information types; receiving, by the computing device, a user input that selects one suggested information type from at least the subset of information types as a selected information type; generating, by the computing device, a set of suggested messages responsive to the message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to the selected information type and information in the message thread; outputting, by the computing device and for display, at least a subset of suggested messages from the set of suggested messages; receiving, by the computing device, a user input that selects one suggested message from at least the subset of suggested messages as a selected message; and populating, by the computing device, a message field with the selected message.

[0093] Example 2: The method of example 1, wherein a length of each suggested message from the set of suggested messages is greater than a length of the selected information type.

[0094] Example 3: The method of example 1 or 2, wherein the machine learning model is a large language model.

[0095] Example 4: The method of any of examples 1 to 3, further includes determining, by the computing device, a message value indicating a predicted utility of a suggested message to the first user for each suggested message from the set of suggested messages by at least applying the machine learning model; and classifying, by the computing device, each suggested message that has a message value satisfying a message value condition as belonging to the subset of suggested messages.

[0096] Example 5: The method of example 4, wherein satisfying the message value condition is based on whether the message value for the suggested message satisfies a threshold message value.

[0097] Example 6: The method of example 4 or 5, wherein outputting at least the subset of suggested messages comprises outputting, by the computing device and for display, at least the subset of suggested messages in an order based on the message value for each suggested message from at least the subset of suggested messages.

[0098] Example 7: The method of any of examples 1 to 6, wherein at least one suggested message of the set of suggested messages includes at least one of text, emojis, emoticons, images, reactions, animations, and videos.

[0099] Example 8: The method of any of examples 1 to 7, wherein the message is by the first user.

[0100] Example 9: The method of any of examples 1 to 8, wherein the message is by the second user.

[0101] Example 10: The method of any of examples 1 to 9, further includes receiving, by the first computing device, a user input that selects feedback relating to at least one suggested message from the set of suggested messages; and training, by the first computing device, the machine learning model using the feedback.

[0102] Example 11 : The method of any of examples 1 to 10, wherein applying the machine learning model includes excluding generation of content relating to a set of excluded information types including one or more of phone numbers, addresses, or web addresses.

[0103] Example 12: A computing device includes: one or more processors; a memory that stores instructions that, when executed by the one or more processors, cause the one or more processors to: obtain a message thread comprising a set of messages between a first user associated with the computing device and a second user; determine a set of information types of a message from the set of messages by at least applying a machine learning model; output, for display, at least a subset of information types from the set of information types; receive a user input that selects one suggested information type from at least the subset of information types as a selected information type; generate a set of suggested messages responsive to the message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to the selected information type and information in the message thread; output, for display, at least a subset of suggested messages from the set of suggested messages; receive a user input that selects one suggested message from at least the subset of suggested messages as a selected message; and populate a message field with the selected message.

[0104] Example 13: The computing device of example 12, wherein a length of each suggested message from the set of suggested messages is greater than a length of the selected information type.

[0105] Example 14: The computing device of example 12 or 13, wherein the machine learning model is a large language model.

[0106] Example 15: The computing device of any of examples 12 to 14, wherein the instructions further cause the one or more processors to: determine a message value indicating a predicted utility of a suggested message to the first user for each suggested message from the set of suggested messages by at least applying the machine learning model; and classify each suggested message that has a message value satisfying a message value condition as belonging to the subset of suggested messages.

[0107] Example 16: The computing device of example 19, wherein satisfying the message value condition is based on whether the message value for the suggested message satisfies a threshold message value.

[0108] Example 17: The computing device of example 15 or 16, wherein the instructions cause the one or more processors to output at least the subset of suggested messages by outputting, for display, at least the subset of suggested messages in an order based on the message value for each suggested message from at least the subset of suggested messages.

[0109] Example 18: The computing device of any of examples 12 to 17, wherein at least one suggested message of the set of suggested messages includes at least one of text, emojis, emoticons, images, reactions, animations, and videos.

[0110] Example 19: The computing device of any of examples 12 to 18, wherein the message is by the first user.

[0111] Example 20: The computing device of any of examples 12 to 19, wherein the message is by the second user.

[0112] Example 21 : The computing device of any of examples 12 to 20, wherein the instructions further cause the one or more processors to: receive a user input that selects feedback relating to at least one suggested message from the set of suggested messages; and train the machine learning model using the feedback.

[0113] Example 22: The computing device of any of examples 12 to 21, wherein the instructions cause the one or more processors to apply the machine learning model by excluding generation of content relating to a set of excluded information types including one or more of phone numbers, addresses, or web addresses.

[0114] Example 23: A non-transitory computer-readable storage medium encoded with instructions that, when executed by one or more processors of a first computing device, cause the one or more processors to: obtain a message thread comprising a set of messages between a first user associated with the computing device and a second user; determine a set of information types of a message from the set of messages by at least applying a machine learning model; output, for display, at least a subset of information types from the set of information types; receive a user input that selects one suggested information type from at least the subset of information types as a selected information type; generate a set of suggested messages responsive to the message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to the selected information type and information in the message thread; output, for display, at least a subset of suggested messages from the set of suggested messages; receive a user input that selects one suggested message from at least the subset of suggested messages as a selected message; and populate a message field with the selected message.

[0115] Example 24: The non-transitory computer-readable storage medium of example 23, wherein a length of each suggested message from the set of suggested messages is greater than a length of the selected information type.

[0116] Example 25: The non-transitory computer-readable storage medium of example 23 or 24, wherein the machine learning model is a large language model.

[0117] Example 26: The non-transitory computer-readable storage medium of any of examples 23 to 25, wherein the instructions further cause the one or more processors to: determine a message value indicating a predicted utility of a suggested message to the first user for each suggested message from the set of suggested messages by at least applying the machine learning model; and classify each suggested message that has a message value satisfying a message value condition as belonging to the subset of suggested messages.

[0118] Example 27: The non-transitory computer-readable storage medium of example 26, wherein satisfying the message value condition is based on whether the message value for the suggested message satisfies a threshold message value.

[0119] Example 28: The non-transitory computer-readable storage medium of example 26 or 27, wherein the instructions cause the one or more processors to output at least the subset of suggested messages by outputting, for display, at least the subset of suggestedmessages in an order based on the message value for each suggested message from at least the subset of suggested messages.

[0120] Example 29: The non-transitory computer-readable storage medium of any of examples 23 to 28, wherein at least one suggested message of the set of suggested messages includes at least one of text, emojis, emoticons, images, reactions, animations, and videos.

[0121] Example 30: The non-transitory computer-readable storage medium of any of examples 23 to 29, wherein the message is by the first user.

[0122] Example 31 : The non-transitory computer-readable storage medium of any of examples 23 to 30, wherein the message is by the second user.

[0123] Example 32: The non-transitory computer-readable storage medium of any of examples 23 to 31, wherein the instructions further cause the one or more processors to: receive a user input that selects feedback relating to at least one suggested message from the set of suggested messages; and train the machine learning model using the feedback.

[0124] Example 33: The non-transitory computer-readable storage medium of any of examples 23 to 32, wherein the instructions cause the one or more processors to apply the machine learning model by excluding generation of content relating to a set of excluded information types including one or more of phone numbers, addresses, or web addresses.

[0125] Example 34: A computer program product for suggested message generation, the computer program product comprising at least one non-transitory computer-readable medium including one or more instructions that, when executed by at least one processor, cause the at least one processor to: obtain a message thread comprising a set of messages between a first user associated with a computing device and a second user; determine a set of information types of a message from the set of messages by at least applying a machine learning model; output, for display, at least a subset of information types from the set of information types; receive a user input that selects one suggested information type from at least the subset of information types as a selected information type; generate a set of suggested messages responsive to the message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to the selected information type and information in the message thread; output, for display, at least a subset of suggested messages from the set of suggested messages; receive a user input that selects one suggested message from at least the subset of suggested messages as a selected message; and populate a message field with the selected message.

[0126] Example 35: The computer program product of example 34, wherein a length of each suggested message from the set of suggested messages is greater than a length of the selected information type.

[0127] Example 36: The computer program product of example 34 or 35, wherein the machine learning model is a large language model.

[0128] Example 37: The computer program product of any of examples 34 to 36, wherein the one or more instructions further cause the at least one processor to: determine a message value indicating a predicted utility of a suggested message to the first user for each suggested message from the set of suggested messages by at least applying the machine learning model; and classify each suggested message that has a message value satisfying a message value condition as belonging to the subset of suggested messages.

[0129] Example 38: The computer program product of example 37, wherein satisfying the message value condition is based on whether the message value for the suggested message satisfies a threshold message value.

[0130] Example 39: The computer program product of example 37 or 38, wherein the one or more instructions cause the at least one processor to output at least the subset of suggested messages by outputting, for display, at least the subset of suggested messages in an order based on the message value for each suggested message from at least the subset of suggested messages.

[0131] Example 40: The computer program product of any of examples 34 to 39, wherein at least one suggested message of the set of suggested messages includes at least one of text, emojis, emoticons, images, reactions, animations, and videos.

[0132] Example 41 : The computer program product of any of examples 34 to 40, wherein the message is by the first user.

[0133] Example 42: The computer program product of any of examples 34 to 41, wherein the message is by the second user.

[0134] Example 43: The computer program product of any of examples 34 to 42, wherein the one or more instructions further cause the at least one processor to: receive a user input that selects feedback relating to at least one suggested message from the set of suggested messages; and train the machine learning model using the feedback.

[0135] Example 44: The computer program product of any of examples 34 to 43, wherein the one or more instructions cause the at least one processor to apply the machine learning model by excluding generation of content relating to a set of excluded information types including one or more of phone numbers, addresses, or web addresses.

[0136] In one or more examples, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored on or transmitted over, as one or more instructions or code, a computer-readable medium and executed by a hardware-based processing unit. Computer-readable media may include computer-readable storage media, which corresponds to a tangible medium such as data storage media, or communication media including any medium that facilitates transfer of a computer program from one place to another, e.g., according to a communication protocol. In this manner, computer-readable media generally may correspond to (1) tangible computer-readable storage media, which is non-transitory or (2) a communication medium such as a signal or carrier wave. Data storage media may be any available media that can be accessed by one or more computers or one or more processors to retrieve instructions, code and / or data structures for implementation of the techniques described in this disclosure. A computer program product may include a computer-readable medium.

[0137] By way of example, and not limitation, such computer-readable storage media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage, or other magnetic storage devices, flash memory, or any other medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer. Also, any connection is properly termed a computer- readable medium. For example, if instructions are transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. It should be understood, however, that computer-readable storage media and data storage media do not include connections, carrier waves, signals, or other transient media, but are instead directed to non-transient, tangible storage media. Disk and disc, as used herein, includes compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray disc, where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0138] Instructions may be executed by one or more processors, such as one or more digital signal processors (DSPs), general purpose microprocessors, application specific integrated circuits (ASICs), field programmable logic arrays (FPGAs), or other equivalentintegrated or discrete logic circuitry. Accordingly, the term “processor,” as used herein may refer to any of the foregoing structures or any other structure suitable for implementation of the techniques described herein. In addition, in some aspects, the functionality described herein may be provided within dedicated hardware and / or software modules. Also, the techniques could be fully implemented in one or more circuits or logic elements.

[0139] The techniques of this disclosure may be implemented in a wide variety of devices or apparatuses, including a wireless handset, an integrated circuit (IC) or a set of ICs (e.g., a chip set). Various components, modules, or units are described in this disclosure to emphasize functional aspects of devices configured to perform the disclosed techniques, but do not necessarily require realization by different hardware units. Rather, as described above, various units may be combined in a hardware unit or provided by a collection of intraoperative hardware units, including one or more processors as described above, in conjunction with suitable software and / or firmware.

[0140] Various examples have been described. These and other examples are within the scope of the following claims.

Claims

WHAT IS CLAIMED IS:

1. A method comprising: obtaining, by a computing device, a message thread comprising a set of messages between a first user associated with the computing device and a second user; determining, by the computing device, a set of information types of a message from the set of messages by at least applying a machine learning model; outputting, by the computing device and for display, at least a subset of information types from the set of information types; receiving, by the computing device, a user input that selects one suggested information type from at least the subset of information types as a selected information type; generating, by the computing device, a set of suggested messages responsive to the message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to the selected information type and information in the message thread; outputting, by the computing device and for display, at least a subset of suggested messages from the set of suggested messages; receiving, by the computing device, a user input that selects one suggested message from at least the subset of suggested messages as a selected message; and populating, by the computing device, a message field with the selected message.

2. The method of claim 1, wherein a length of each suggested message from the set of suggested messages is greater than a length of the selected information type.

3. The method of claim 1 or 2, wherein the machine learning model is a large language model.

4. The method of any of claims 1 to 3, further comprising: determining, by the computing device, a message value indicating a predicted utility of a suggested message to the first user for each suggested message from the set of suggested messages by at least applying the machine learning model; and classifying, by the computing device, each suggested message that has a message value satisfying a message value condition as belonging to the subset of suggested messages, wherein satisfying the message value condition is based on whether the message value for the suggested message satisfies a threshold message value.

5. The method of claim 4, wherein outputting at least the subset of suggested messages comprises outputting, by the computing device and for display, at least the subset of suggested messages in an order based on the message value for each suggested message from at least the subset of suggested messages.

6. The method of any of claims 1 to 5, wherein at least one suggested message from the set of suggested messages comprises at least one of text, emojis, emoticons, images, reactions, animations, and videos.

7. The method of any of claims 1 to 6, further comprising: receiving, by the computing device, a user input that selects feedback relating to at least one suggested message from the set of suggested messages; and training, by the computing device, the machine learning model using the feedback.

8. A computing device comprising: one or more processors; a memory that stores instructions that, when executed by the one or more processors, cause the one or more processors to: obtain a message thread comprising a set of messages between a first user associated with the computing device and a second user; determine a set of information types of a message from the set of messages by at least applying a machine learning model; output, for display, at least a subset of information types from the set of information types; receive a user input that selects one suggested information type from at least the subset of information types as a selected information type; generate a set of suggested messages responsive to the message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to the selected information type and information in the message thread; output, for display, at least a subset of suggested messages from the set of suggested messages; receive a user input that selects one suggested message from at least the subset of suggested messages as a selected message; and populate a message field with the selected message.

9. The computing device of claim 8, wherein a length of each suggested message from the set of suggested messages is greater than a length of the selected information type.

10. The computing device of claim 8 or 9, wherein the instructions further cause the one or more processors to: determine a message value indicating a predicted utility of a suggested message to the first user for each suggested message from the set of suggested messages by at least applying the machine learning model; and classify each suggested message that has a message value satisfying a message value condition as belonging to the subset of suggested messages, wherein satisfying the message value condition is based on whether the message value for the suggested message satisfies a threshold message value.

11. The computing device of claim 10, wherein the instructions further cause the one or more processors to output at least the subset of suggested messages by outputting, for display, at least the subset of suggested messages in an order based on the message value for each suggested message from at least the subset of suggested messages.

12. The computing device of any of claims 8 to 11, wherein the instructions further cause the one or more processors to: receive a user input that selects feedback relating to at least one suggested message from the set of suggested messages; and train the machine learning model using the feedback.

13. A non-transitory computer-readable storage medium encoded with instructions that, when executed by one or more processors of a computing device, cause the one or more processors to perform any of the methods of claim 1-7.

14. A computer program product for suggested message generation, the computer program product comprising at least one non-transitory computer-readable medium including one or more instructions that, when executed by at least one processor, cause the at least one processor to: obtain a message thread comprising a set of messages between a first user associated with a computing device and a second user; determine a set of information types of a message from the set of messages by at least applying a machine learning model; output, for display, at least a subset of information types from the set of information types; receive a user input that selects one suggested information type from at least the subset of information types as a selected information type; generate a set of suggested messages responsive to the message from the set of messages by at least applying the machine learning model, wherein information in each suggested message from the set of suggested messages is related to the selected information type and information in the message thread; output, for display, at least a subset of suggested messages from the set of suggested messages; receive a user input that selects one suggested message from at least the subset of suggested messages as a selected message; and populate a message field with the selected message.

Citation Information

Patent Citations

  • Reply recommendation apparatus and system and method for text construction

    US20160306800A1

  • Method and apparatus for recommending reply message

    US20160308794A1

  • Smart replies using an on-device model

    US20180089588A1

  • Providing message response suggestions

    US20200374243A1