System

Abstract

A system is provided.SOLUTION: Means for collecting historical communication data, means for pre-processing and cleansing the historical communication data, means for training an artificial intelligence model that learns a writing style of each user by using the pre-processed data, means for providing a user terminal on which a response sentence creation button that can be operated by the user at the time of response is arranged, means for acquiring communication content received when the user presses the button and generating a response sentence based on the artificial intelligence model, means for distributing the generated response sentence to the user terminal, and means for transmitting communication when the user edits and confirms the response sentence. A system comprising: SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] Many users spend a lot of time drafting replies when using communication software to communicate electronically. Furthermore, because each user's writing style and vocabulary differ, creating consistent and efficient reply drafts is difficult, resulting in reduced work efficiency. The present invention aims to solve the above problems by reducing the time and efficiency of email drafting, while also enabling the automatic generation of reply drafts that reflect each user's writing style and perspective. [Means for solving the problem]

[0005] The present invention is a system including means for collecting past communication data, means for preprocessing and cleansing the past communication data, means for using the preprocessed data to train an artificial intelligence model that learns the writing style of each user, means for providing a user terminal having a response draft creation button that can be operated by the user when responding, means for acquiring the content of a communication received when the user presses the button and generating a response draft based on the artificial intelligence model, means for delivering the generated response draft to the user terminal, and means for sending a communication when the user edits and confirms the response draft. This allows for the rapid generation of reply drafts that reflect the writing style and position of each user based on the past communication data of the user, significantly reducing the user's effort.

[0006] "Communications data" refers to information sent or received electronically, such as email or messages.

[0007] A "collection means" is a method or device for collecting communication data based on specific criteria.

[0008] "Preprocessing" refers to cleansing and filtering collected data to remove unnecessary information and unify and format the data.

[0009] An "artificial intelligence model" is software that has the ability to learn from data and perform specific tasks automatically.

[0010] "Style" refers to the style and expression of writing used by a particular person or organization.

[0011] "Response Drafting Button" means the portion of the interface that a user operates to automatically generate a reply to a particular electronic communication.

[0012] A "user terminal" is a device that a user actually operates, such as a personal computer or smartphone.

[0013] A "draft response" is a candidate sentence for generating an appropriate reply to the received communication content.

[0014] The "distribution means" refers to a method or device for transmitting the generated response draft to the user terminal.

[0015] "Editing" refers to the act of a user modifying or revising a generated response draft. [Brief explanation of the drawings]

[0016] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12]FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0017] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0018] First, the terms used in the following description will be explained.

[0019] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0020] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0021] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0022] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0023] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0024] [First embodiment]

[0025] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0026] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0027] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0028] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0029] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0030] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0031] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0032] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0033] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0034] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0035] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0036] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0037] The present invention relates to a system for automatically generating reply drafts for electronic communications, which system functions around three central parties: a server, a terminal, and a user.

[0038] 1. Configuration

[0039] server

[0040] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the AI ​​model and generating sentences. Specifically, the server implements the following functions:

[0041] Data collection function: Collects past communication data from the company's internal email system.

[0042] Data preprocessing function: Cleanses collected data and removes unnecessary information.

[0043] Model training function: The cleansed data is used to train an artificial intelligence model that learns each user's writing style.

[0044] Text generation function: Analyzes the content of received communications and generates text suggestions based on artificial intelligence models.

[0045] Terminal

[0046] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[0047] Interface provision function: Displays a response draft creation button. When the user presses this button, the draft generation process begins.

[0048] Data transmission function: Sends the contents of received emails to the server.

[0049] Draft display function: Displays the draft sent from the server.

[0050] Draft editing function: Provides an interface for users to modify and edit the generated draft.

[0051] Sending function: The confirmed email draft is sent via the server.

[0052] User

[0053] A user is a person who operates a system and uses it according to a specific procedure, specifically by performing the following steps:

[0054] Check email: Check the email you received.

[0055] How to use the Create a Draft button: Press the Create a Draft button when replying.

[0056] Review and edit the transcript: Review the auto-generated transcript and make any necessary corrections.

[0057] Send Email: Send the revised / edited email draft.

[0058] 2. Program Processing

[0059] Server Processing

[0060] The server first collects past communication data from the company's email system. The collected data is pre-processed and cleansed to remove unnecessary information. An artificial intelligence model is then trained to learn each user's writing style and perspective. Once trained, the model is used to generate copy.

[0061] Terminal handling

[0062] When the user presses the "Create a draft" button, the device extracts the contents of the open email and sends it to the server. When the generated draft is returned from the server, the device displays it to the user. After the user confirms, edits, and confirms the draft, the device finally sends it.

[0063] User operations

[0064] The user opens the received email and, if a reply is necessary, presses the Create Message button. The user checks the displayed message and edits the content as necessary. The email is then sent with the final message.

[0065] 3. Specific Examples

[0066] Data collection and preprocessing

[0067] The server collects email data from the company's internal email server for the past year.

[0068] The server cleanses the collected data, removing unnecessary and personal information.

[0069] Training the model

[0070] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[0071] Copy generation and sending

[0072] The user opens the received email and presses the draft button.

[0073] The terminal extracts the mail contents and transmits them to the server.

[0074] The server generates a draft based on the received content and returns it to the terminal.

[0075] The terminal displays the generated draft to the user.

[0076] The user checks and edits the draft, and once it is confirmed, presses the send button.

[0077] The terminal transmits the confirmed email draft to the server, and finally transmits the email to the destination.

[0078] This system significantly reduces the user's workload and enables quick and accurate generation and sending of response emails.

[0079] The processing flow will be explained below.

[0080] Step 1:

[0081] The server collects past communication data from the company's email system, primarily email data from the past year, and collects it separately for each user.

[0082] Step 2:

[0083] The server stores the collected email data in a database, along with metadata such as the sender / receiver, date and time, and subject of the email.

[0084] Step 3:

[0085] The server performs spam filtering and personal information cleansing of the stored data, removing unnecessary and privacy-sensitive information.

[0086] Step 4:

[0087] The server then categorizes and tags the cleansed data, assigning tags to each email based on its content (e.g., formal, casual, urgent, etc.) to make the data easier to organize and analyze.

[0088] Step 5:

[0089] The server uses the cleansed and tagged data to train an artificial intelligence model, which uses natural language processing techniques to learn each user's writing style and vocabulary.

[0090] Step 6:

[0091] The terminal displays a message drafting button when the user opens the email client. When the user selects an email to reply to and presses the message drafting button, the terminal proceeds to the next step.

[0092] Step 7:

[0093] When the user presses the "Create a draft" button, the device extracts the contents of the open email, and this extracted data is used as key information for generating the draft.

[0094] Step 8:

[0095] The terminal sends the extracted email content to the server. The sent data includes the full text of the received email and metadata.

[0096] Step 9:

[0097] The server analyzes the content of the received email and generates drafts using an artificial intelligence model based on the corresponding user's writing style.

[0098] Step 10:

[0099] The server sends the generated message back to the terminal, which contains the necessary information for the reply message.

[0100] Step 11:

[0101] The terminal displays the received draft to the user, who can then check the content and make corrections or edits as necessary.

[0102] Step 12:

[0103] The user checks and edits the draft and finally confirms it. After confirming, they press the send button to proceed to the next step.

[0104] Step 13:

[0105] The terminal sends the finalized email draft to the server. The sent data includes the final email content.

[0106] Step 14:

[0107] The server sends the email to the specified recipient based on the final draft it received. After sending, it is saved in the database as a sending history.

[0108] By following these steps, users can easily and efficiently create and send correspondence drafts.

[0109] Example 1

[0110] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0111] In conventional electronic communication, creating replies is often done manually, resulting in a significant workload. Furthermore, some users have difficulty maintaining a consistent writing style and formality, resulting in inconsistencies in the quality of corporate communications. In these circumstances, there is a need for efficient and consistent replies. Furthermore, there is a need for a means to effectively utilize past communication data and automatically generate replies that reflect a user's unique writing style. To address this issue, the present invention provides a system for automatically generating efficient and consistent reply drafts.

[0112] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0113] In this invention, the server includes means for collecting historical communication data, means for pre-processing and cleansing the historical communication data, and means for using the pre-processed data to train a generative AI model that learns each user's writing style, thereby enabling efficient and consistent automated generation of reply drafts.

[0114] "Data collection" is the process of obtaining historical communication data from a particular system or server.

[0115] "Cleansing" is the process of removing unnecessary or incorrect information from collected data and arranging it into an appropriate format.

[0116] A "generative AI model" is an artificial intelligence model that learns the unique characteristics of users, such as their writing style and position, and automatically generates optimal reply drafts.

[0117] The "response drafting button" is a terminal interface element that the user operates to initiate the email reply drafting process.

[0118] "Reply generation" is the process of using a generative AI model to create appropriate reply suggestions based on collected and pre-processed data.

[0119] "Data transmission" is the process of sending and receiving extracted email content and generated drafts between the server and the terminal.

[0120] "Draft editing" is the process in which the user reviews the generated draft and makes corrections or edits as necessary.

[0121] "Communication sending" is the process of sending the finalized email draft to the intended recipient.

[0122] The present invention relates to a system for automatically generating reply drafts for electronic communications, which system functions around three main entities: a server, a terminal, and a user.

[0123] Server configuration and functions

[0124] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the generative AI model and generating sentences. Specifically, it implements the following functions:

[0125] Data collection function: The server collects historical email data from the company's mail system. For example, it periodically retrieves data for the past year using a Linux cron job or a Python script.

[0126] Data preprocessing function: The server uses a data cleansing tool (e.g., the Pandas library) to remove unnecessary information and personal information from the collected data.

[0127] Model training function: The server uses the cleansed data to train a generative AI model (e.g., GPT-3 or BERT model). This model learns each user's writing style and perspective and is used to automatically generate email reply suggestions.

[0128] Text generation function: The server generates an appropriate text based on the content of the received email and returns it to the terminal.

[0129] Device configuration and functions

[0130] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[0131] Interface provision function: The terminal displays a response draft creation button. When the user presses this button, the draft generation process begins.

[0132] Email content extraction function: When the user presses a button, the device extracts the contents of the open email and sends them to the server.

[0133] Draft reception function: Receives drafts generated from the server.

[0134] Text display function: The terminal displays the received text to the user, for example, using a GUI framework (JavaFX or Electron).

[0135] Copy editing function: Provides an interface (such as a rich text editor) for users to review and edit the copy.

[0136] Final sending function: The user resends the finalized draft to the server and finally sends the email.

[0137] User operation procedure

[0138] The user uses the system by following these steps:

[0139] 1. Check your email: Open your received email.

[0140] 2. How to use the Create a Draft button: Press the Create a Draft button when replying.

[0141] 3. Review and edit the transcript: Review the auto-generated transcript and make any necessary corrections.

[0142] 4. Send the email: Send the revised and edited email.

[0143] Specific examples

[0144] Data collection and preprocessing

[0145] The server collects email data from the company's internal email server for the past year.

[0146] The server cleanses the collected data, removing unnecessary and personal information.

[0147] Training the model

[0148] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[0149] Copy generation and sending

[0150] The user opens the received email and presses the draft button.

[0151] The terminal extracts the mail contents and transmits them to the server.

[0152] The server generates a draft based on the received content and returns it to the terminal.

[0153] The terminal displays the generated draft to the user.

[0154] The user checks and edits the draft, and once it is confirmed, presses the send button.

[0155] The terminal transmits the finalized email draft to the server, and finally transmits the email to the destination.

[0156] Prompt Sentence Examples

[0157] "I received an email inquiring about the status of a recent project. Please draft a response."

[0158] This system significantly reduces the user's workload and enables quick and accurate generation and sending of response emails.

[0159] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0160] Step 1: Collect historical communication data

[0161] The server collects email data from your company's mail server for the past year, using a Python script and the mail server's IMAP protocol to retrieve all emails for a specified period.

[0162] Input: Email server account information, period (e.g., past year)

[0163] Output: Captured email data (subject, body, sender / receiver information, etc.)

[0164] Step 2: Preprocess the data

[0165] The server cleanses the retrieved email data using the Pandas library, specifically by removing unnecessary header information and personal information and converting the data into an appropriate format.

[0166] Input: Captured email data

[0167] Output: Cleansed data (well-formatted data with unnecessary information removed)

[0168] Step 3: Training the generative AI model

[0169] The server uses the cleansed data to train a generative AI model (e.g., GPT-3 or BERT), applying machine learning algorithms to learn each user's writing style and vocabulary.

[0170] Input: Cleansed data

[0171] Output: A trained generative AI model

[0172] Step 4: Begin the response drafting process

[0173] The user presses the drafting button displayed on the terminal's email client, which starts the drafting process.

[0174] Input: User button presses

[0175] Output: Signal to start the copywriting process

[0176] Step 5: Extract the contents of the received email

[0177] The terminal extracts the content of the received email that the user currently has open, specifically, the subject, body, and sender information of the email.

[0178] Input: Incoming emails that the user has opened

[0179] Output: Extracted email content (subject, body, sender information)

[0180] Step 6: Send the email content to the server

[0181] The terminal sends the extracted email content to the server as an HTTP POST request.

[0182] Input: Extracted email content

[0183] Output: Request sent to server

[0184] Step 7: Server-generated copy

[0185] The server creates a prompt based on the content of the received email and generates a draft using a generative AI model. Specifically, it analyzes the content of the email, generates a prompt for the generative AI model, and then creates a reply draft based on that.

[0186] Input: Email content (subject, body, sender information)

[0187] Output: Generated reply draft

[0188] Step 8: Receiving and viewing the generated draft

[0189] The terminal displays the generated draft received from the server to the user, specifically, using a GUI framework to display the draft in an easy-to-read format.

[0190] Input: Generated reply

[0191] Output: The text displayed on the screen

[0192] Step 9: Review and edit your copy

[0193] The user checks the displayed draft and corrects or edits it as necessary. Specifically, the user edits the content using a rich text editor.

[0194] Input: Displayed text

[0195] Output: Edited draft

[0196] Step 10: Submit the final draft

[0197] When the user has finished editing, he / she presses the send button to confirm the final draft, which causes the device to resend the confirmed draft to the server and finally send the email.

[0198] Input: Edited text, click the send button

[0199] Output: Send the finalized draft and email it to the destination

[0200] (Application example 1)

[0201] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0202] Creating replies in conventional electronic communications requires a significant amount of time and effort from the user, making it difficult to respond quickly and accurately, especially during busy periods or when a large number of inquiries are received. Furthermore, it is difficult to maintain consistency in replies for each requester based on the user's writing style and past communication data. Furthermore, with the demand for rapid responses to inquiries, such as after-sales service, there is a lack of technology to efficiently generate response drafts.

[0203] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0204] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, and means for using the preprocessed data to train an artificial intelligence model that learns each user's writing style, thereby enabling means for automatically analyzing the content of the inquiry, means for generating a reply draft based on the content of the inquiry, means for displaying the generated reply draft to the user in an editable form, and means for automatically sending the finalized reply.

[0205] "Past communication data" refers to data on electronic communications that a user has conducted in the past.

[0206] "Cleansing" is a process of removing unnecessary information from collected data to make the data clean.

[0207] An "artificial intelligence model" is an algorithm that uses machine learning techniques to learn a user's writing style and specific response patterns.

[0208] The "draft creation button" is an interface that allows the user to automatically generate a reply draft.

[0209] A "user terminal" is an electronic device operated by a user that has the means to communicate with a server and send and receive data.

[0210] A "draft response" is a draft reply that is automatically generated based on the content of the received communication.

[0211] "Automatic analysis of inquiry content" is a technology that understands the content of received inquiries and identifies appropriate categories and subjects.

[0212] "Reply draft generation" is the process of automatically creating an appropriate reply based on the analyzed inquiry content.

[0213] "Editable Draft Display" is a function that presents the generated reply draft to the user, allowing the user to modify the text as needed.

[0214] "Auto-send" is a function that allows the system to automatically send a confirmed reply.

[0215] The present invention relates to a system for automatically generating reply drafts in electronic communications, which functions around three main entities: a server, a terminal, and a user.

[0216] 1. Configuration

[0217] server

[0218] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the generative AI model and generating text suggestions. Specifically, it implements the following functions:

[0219] Data collection function: The server collects past communication data from the company's internal communication systems.

[0220] Data preprocessing function: The server cleanses the collected data and removes unnecessary information.

[0221] Model training function: The server uses the cleansed data to train a generative AI model to learn each user's writing style.

[0222] Message generation function: The server analyzes the received communication content and generates a reply message based on the generative AI model.

[0223] Automatic analysis of inquiry content: The server automatically analyzes the inquiry content and identifies the appropriate category and subject.

[0224] Terminal

[0225] The terminal is a device operated by the user, which communicates with the server to send and receive data. The terminal has the following functions:

[0226] Interface provision function: The terminal displays a response draft creation button, which the user can press to start the draft generation process.

[0227] Data transmission function: The terminal transmits the received communication content to the server.

[0228] Draft display function: The terminal displays the draft sent from the server.

[0229] Draft editing function: The terminal provides an interface for users to modify and edit the generated draft.

[0230] Send function: Sends the final draft of the text that the user has confirmed.

[0231] User

[0232] A user is a person who operates the system and uses it according to a specific procedure, specifically by performing the following steps:

[0233] Confirming the inquiry: The user confirms the received inquiry email.

[0234] Operation of the Create Copy Button: When a user wants to reply, he or she presses the Create Copy button.

[0235] Review and edit the draft: The user reviews the automatically generated draft and corrects it if necessary.

[0236] Sending email: The user sends the corrected and edited reply.

[0237] 2. Hardware and Software Used

[0238] Hardware:

[0239] Server machine: A machine with a high-performance CPU and GPU

[0240] Device: Devices such as smartphones and PCs

[0241] software:

[0242] TensorFlow: Used for training and inferencing generative AI models

[0243] NLTK: Used to preprocess data for natural language processing

[0244] sklearn:Used for splitting data function

[0245] 3. Data processing and data calculation

[0246] The server preprocesses past communication data collected by the server and removes unnecessary information. Using this cleansed data, the server trains an AI model that learns each user's writing style. The generated AI model analyzes the content of received communications and automatically generates reply drafts. It also automatically analyzes the content of inquiries and provides appropriate reply drafts.

[0247] 4. Specific Examples

[0248] As an example of actual system operation, the following prompt statements can be used:

[0249] Inquiries regarding this purchase

[0250] "The product never arrived"

[0251] Please tell me the return procedure.

[0252] In this way, the present invention significantly reduces the user's workload and enables quick and accurate generation and transmission of response emails.

[0253] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0254] Step 1:

[0255] The server collects past communication data. Specifically, it collects email data from the company's internal communication system for the past year and stores it in a database. At this stage, the input is internal email data, and the output is the collected, unprocessed data.

[0256] Step 2:

[0257] The server preprocesses and cleanses the collected data. Specifically, it uses the natural language processing library NLTK to remove unnecessary words (stop words) and personal information from the collected data. At this stage, the collected raw data is the input, and the cleansed data is the output.

[0258] Step 3:

[0259] The server uses the preprocessed data to train a generative AI model. Specifically, it uses TensorFlow to segment the data and an LSTM model to learn each user's writing style. At this stage, the input is the cleansed data and the output is a trained AI model.

[0260] Step 4:

[0261] When a user receives an inquiry email, the terminal displays the received email and the user presses the button to create a response draft. The received email content is the input, and the content is sent to the server as the output.

[0262] Step 5:

[0263] The server analyzes the content of the received email and automatically generates a reply. Specifically, it analyzes the inquiry content and generates an appropriate reply using a trained AI model. At this stage, the received email content is the input, and the generated reply is the output.

[0264] Step 6:

[0265] The terminal displays the generated reply draft to the user. The user can then review this draft and make corrections or edits as necessary. For example, in response to an inquiry such as "The arrival of the product is delayed," the following reply is generated: "We apologize for any inconvenience caused by your order. We are currently checking the delivery status. Could you please wait a moment?"

[0266] Step 7:

[0267] When the user finally confirms the draft, the terminal sends this confirmed draft to the server, which then automatically sends the confirmed draft as a reply to the original query. At this stage, the confirmed draft is the input, and a reply email is sent to the user as the output.

[0268] The above steps significantly reduce the user's workload and enable quick and accurate generation and sending of response emails.

[0269] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0270] The present invention relates to a system for automatically generating reply proposals for electronic communications, in particular incorporating an emotion engine. The system functions around three parties: a server, a terminal, and a user.

[0271] 1. Configuration

[0272] server

[0273] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the AI ​​model and generating sentences. Specifically, the server implements the following functions:

[0274] Data collection function: Collects past communication data from the company's internal email system.

[0275] Data preprocessing function: Cleanses collected data and removes unnecessary information.

[0276] Model training function: The cleansed data is used to train an artificial intelligence model that learns each user's writing style.

[0277] Text generation function: Analyzes the content of received communications and generates text suggestions based on artificial intelligence models.

[0278] Emotion recognition: Recognizes user emotions using an emotion engine and reflects them in the generated text.

[0279] Terminal

[0280] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[0281] Interface provision function: Displays a response draft creation button and emotion recognition results. When the user presses this button, the draft generation process begins.

[0282] Data transmission function: Sends the contents of received emails to the server.

[0283] Text display function: Displays the text sent from the server and emotion recognition results.

[0284] Draft editing function: Provides an interface for users to modify and edit the generated draft.

[0285] Sending function: The confirmed email draft is sent via the server.

[0286] User

[0287] A user is a person who operates a system and uses it according to a specific procedure, specifically by performing the following steps:

[0288] Check email: Check the email you received.

[0289] How to use the Create a Draft button: Press the Create a Draft button when replying.

[0290] Review and edit the text and emotion recognition results: Review the automatically generated text and emotion recognition results and make corrections as necessary.

[0291] Send Email: Send the revised / edited email draft.

[0292] 2. Program Processing

[0293] Server Processing

[0294] The server first collects past communication data from the company's email system. The collected data is pre-processed and cleansed, and unnecessary information is removed. An artificial intelligence model is then trained to learn each user's writing style and perspective. Once trained, the model is used to generate text suggestions. It also uses an emotion engine to recognize user emotions and reflect them in the generated text.

[0295] Terminal handling

[0296] When the user presses the "Create a message" button, the device extracts the contents of the open email and sends it to the server. When the generated message and emotion recognition results are returned from the server, the device displays them to the user. After the user confirms, edits, and confirms the message and emotion recognition results, the device finally sends the message.

[0297] User operations

[0298] The user opens the received email and, if a reply is necessary, presses the "Create a message" button. The user checks the displayed message and emotion recognition results, and edits the content as necessary. The email is then sent with the finalized message.

[0299] 3. Specific Examples

[0300] Data collection and preprocessing

[0301] The server collects email data from the company's internal email server for the past year.

[0302] The server cleanses the collected data, removing unnecessary and personal information.

[0303] Training the model

[0304] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[0305] emotion recognition

[0306] The server uses an emotion engine to recognize the user's emotions, and this emotion data is reflected in the copy generation.

[0307] Copy generation and sending

[0308] The user opens the received email and presses the draft button.

[0309] The terminal extracts the mail contents and transmits them to the server.

[0310] The server generates a draft based on the received content and the emotion recognition results and sends it back to the terminal.

[0311] The terminal displays the generated draft and emotion recognition results to the user.

[0312] The user checks and edits the draft, and once it is confirmed, presses the send button.

[0313] The terminal transmits the confirmed email draft to the server, and finally transmits the email to the destination.

[0314] This system significantly reduces the user's workload and enables the system to generate and send response emails quickly and accurately. Furthermore, the introduction of an emotion engine allows the system to provide appropriate message suggestions based on the user's emotions, enabling more human-like communication.

[0315] The processing flow will be explained below.

[0316] Step 1:

[0317] The server collects past communication data from the company's internal email system, primarily email data from the past year, and collects it separately for each user.

[0318] Step 2:

[0319] The server stores the collected email data in a database, along with metadata such as the sender / receiver, date and time, and subject of the email.

[0320] Step 3:

[0321] The server cleanses the stored data, filtering out spam and removing unnecessary and personal information, and uses natural language processing technology to format the data.

[0322] Step 4:

[0323] The server categorizes and tags the cleansed data, assigning tags to each email based on its content (e.g., formal, casual, urgent, etc.) to make the data easier to organize and analyze.

[0324] Step 5:

[0325] The server uses the cleansed and tagged data to train an artificial intelligence model, which uses natural language processing techniques to learn each user's writing style and vocabulary, enabling personalized responses.

[0326] Step 6:

[0327] The server trains an emotion engine to recognize user emotions by analyzing keystrokes, mouse movements, and voice inputs, and incorporates this emotion information into the copy generation model.

[0328] Step 7:

[0329] When the user opens the email client, the device displays a message creation button. When the user selects the email to reply to and presses the message creation button, the device proceeds to the next step.

[0330] Step 8:

[0331] When the user presses the "Create a copy" button, the device extracts the contents of the open email, and this extracted data is used as key information for generating the copy.

[0332] Step 9:

[0333] The terminal transmits the extracted email content and emotion data generated by the emotion engine to the server. The transmitted data includes the full text of the received email, metadata, and emotion information.

[0334] Step 10:

[0335] The server analyzes the content and emotional information of the received email and generates a draft message using an AI model based on the corresponding user's writing style, taking into account the user's emotional state.

[0336] Step 11:

[0337] The server then sends the generated message back to the device, which includes the necessary information for the email reply and the recommended text from the emotion engine.

[0338] Step 12:

[0339] The device displays the received draft to the user, who can then check the content and make corrections or edits as necessary.

[0340] Step 13:

[0341] The user checks and edits the draft and finally confirms it. After confirming it, the user presses the submit button to proceed to the next step.

[0342] Step 14:

[0343] The terminal sends the finalized email draft to the server. The sent data includes the final email content.

[0344] Step 15:

[0345] The server sends the email to the specified recipient based on the final draft received. After sending, it is recorded in the database as a sending history.

[0346] Example 2

[0347] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0348] Modern business communication requires fast and appropriate responses to large volumes of emails. However, for users who receive emails frequently, coming up with an appropriate response for each email can be a significant burden. It is also important for email replies to reflect the user's emotions, which requires the user to incorporate their own feelings into the text. Furthermore, it is even more difficult to reply in a consistent style using past communication data. To address these challenges, a system is needed that automatically generates emotionally relevant response drafts and allows users to easily edit and send them.

[0349] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0350] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, means for training an artificial intelligence model that learns each user's writing style using the preprocessed data, means for reflecting emotion data obtained from an emotion recognition engine in message draft generation, and means for delivering the generated response drafts and emotion recognition results to the user terminal. This allows for the automatic generation of quick and appropriate email responses, and the user only needs to edit them, significantly reducing the burden on the user. Furthermore, by reflecting the emotion recognition results, more human-like communication can be achieved.

[0351] "Past communication data" refers to records of communication such as emails sent and received in the past, and includes the content of exchanges between users.

[0352] "Preprocessing" refers to the process of removing unnecessary information from collected data and preparing the data in a format that is easy to analyze.

[0353] "Cleansing" refers to the process of removing noise and unnecessary parts from data and extracting only the necessary information.

[0354] "Artificial intelligence model" refers to a program or algorithm that learns a user's writing style or specific patterns through machine learning or deep learning.

[0355] "Response drafting button" refers to a button that a user operates on the interface to cause the system to initiate automatic response drafting.

[0356] A "user terminal" is a device operated by a user, and refers to computer equipment such as a personal computer or smartphone.

[0357] An "emotion recognition engine" refers to an algorithm or software that analyzes and classifies user emotions from text data.

[0358] "Emotion data" refers to user emotion information extracted by an emotion recognition engine, and includes emotion categories such as "joy," "sadness," and "anger," for example.

[0359] "Message generation" refers to the process of using an artificial intelligence model to automatically create a message response based on the received content and sentiment data.

[0360] "Delivery" refers to sending the generated text and emotion recognition results to a user terminal via a network.

[0361] "Communication" refers to the act of sending and receiving e-mails, etc. over a network.

[0362] An embodiment of this invention provides a system that collects and cleans past communication data, trains an artificial intelligence model based on that data, generates drafts using an emotion recognition engine, and allows users to review, edit, and send them. The components of this system and the specific processing flow are described below.

[0363] Server Roles

[0364] The server plays a central role in the system, collecting data, preprocessing, training models, generating text, and recognizing emotions using the following hardware and software:

[0365] Hardware: A powerful server computer (e.g., a server with dual Xeon processors)

[0366] Software: Databases (e.g., MongoDB), generative AI models (e.g., GPT-3), emotion recognition engines (e.g., Microsoft Azure Cognitive Services)

[0367] Data collection and preprocessing

[0368] The server collects email data from the company's internal email system for the past year. This is achieved by using the email server's API to retrieve emails and store them in a local database. The collected data then goes through a cleansing process to remove unnecessary parts (such as email headers and signatures) and extract only the body of the email.

[0369] Training the model

[0370] The server uses the cleansed email data to train a generative AI model that learns the writing style and tone of each user's emails using large datasets and advanced machine learning algorithms.

[0371] emotion recognition

[0372] The server uses an emotion recognition engine to analyze the user's emotions. Specifically, it uses emotional data extracted from the content of the email (e.g., "joy," "sadness," "anger," etc.) to generate more appropriate text.

[0373] Copy generation and sending

[0374] When a user checks an incoming email and presses the message creation button if a reply is necessary, the device extracts the email content and sends it to the server. The server generates a message based on the received email content and the emotion recognition results. The generated message and emotion recognition results are sent back to the device and displayed to the user.

[0375] The user can check and edit the displayed message, correct it as appropriate, and then press the send button to send the final message. At this time, the device sends the message via the server.

[0376] Specific examples

[0377] For example, you can use prompts like the following to instruct a generative AI model to generate specific sentences:

[0378] "Analyze the content of the received email and generate a polite reply that includes an apology. Based on the results of the emotion engine, create a reply that takes into consideration the user's emotions."

[0379] Based on this specific prompt, the server can automatically generate appropriate reply ideas using a generative AI model, significantly reducing the user's effort. Furthermore, by incorporating emotion recognition, more human-like communication can be achieved.

[0380] In this way, the present invention helps users to respond to emails quickly and accurately, and improves the efficiency of email reply work.

[0381] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0382] Step 1: Collect data

[0383] The server collects email data from the company's internal email system for the past year. Specifically, it uses the email server's API to retrieve emails from the past year. The retrieved email data is stored in a local database (e.g., MongoDB) in JSON format.

[0384] Input: Email data from the internal email system (period: past year)

[0385] Output: Email data stored in a local database (JSON format)

[0386] Step 2: Preprocessing the data

[0387] The server cleanses the collected email data. Specifically, it uses regular expressions to remove unnecessary information such as email headers and signatures, and extracts only the body of the email. The preprocessed data is then saved in JSON format.

[0388] Input: Collected email data (JSON format)

[0389] Output: Cleansed email data (JSON format)

[0390] Step 3: Train the model

[0391] The server trains a generative AI model (e.g., GPT-3) using the cleansed email data. Specifically, it splits the data into batches and trains the model over multiple epochs. The trained model learns the writing style and tone of each user.

[0392] Input: Cleansed email data (JSON format)

[0393] Output: A trained generative AI model

[0394] Step 4: Process incoming email

[0395] The user checks the received email and presses the message creation button if a reply is necessary. The terminal extracts the contents of the opened email and sends them to the server.

[0396] Input: Received email content (text format)

[0397] Output: Email content sent to the server (text format)

[0398] Step 5: Recognize emotions

[0399] The server sends the received email content to an emotion recognition engine to analyze the user's emotions. The recognized emotion data is classified into categories such as "joy," "sadness," and "anger," and is used in the copy generation process.

[0400] Input: Received email content (text format)

[0401] Output: Recognized emotion data (categorical format)

[0402] Step 6: Generate copy

[0403] The server provides a prompt to the generative AI model based on the content of the received email and the emotion data, and generates a draft. Example prompt: "Analyze the content of the received email and generate a polite reply that includes an apology. Based on the results of the emotion engine, please create a draft that takes the user's emotions into consideration."

[0404] Input: received email content (text format), recognized emotion data (category format), prompt text

[0405] Output: Generated draft (text format)

[0406] Step 7: Distributing the copy

[0407] The server delivers the generated text and emotion recognition results to the terminal, which then displays them to the user.

[0408] Input: Generated text (text format), emotion recognition results (category format)

[0409] Output: Text and emotion recognition results delivered to the device (text format)

[0410] Step 8: Edit and submit your copy

[0411] The user checks the draft displayed on the device and edits it as necessary. After that, the user presses the send button, and the edited and confirmed draft is sent from the device via the server. It is finally sent as an email.

[0412] Input: Generated draft (text format)

[0413] Output: Final email draft sent (email format)

[0414] (Application example 2)

[0415] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0416] While automated systems for generating reply proposals in electronic communications can significantly reduce time and effort, conventional systems often do not take emotion recognition into account, which can result in inappropriate replies. Furthermore, while there is a growing need for fast and appropriate replies to user comments in content distribution services, current methods suffer from slow response times. The present invention aims to solve these problems by generating fast and appropriate replies while taking user emotions into account.

[0417] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0418] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, and means for training an artificial intelligence model that learns each user's writing style using the preprocessed data. This makes it possible to generate an optimal writing style for each user based on past data. Also, in this invention, the server includes means for generating a response draft based on received content and emotion recognition results, and means for automatically analyzing comments and generating reply drafts in a content distribution service. This makes it possible to take user emotions into consideration and generate appropriate and prompt reply drafts in a content distribution service.

[0419] "Historical communications data" is a collection of previous emails, messages, and other communications records.

[0420] "Means of collection" refers to the functions and processes for importing past communication data into the server.

[0421] "Preprocessing and cleansing means" refers to functions and processes for removing unnecessary information from collected data and preparing the data.

[0422] An "artificial intelligence model" is a computer program that uses machine learning techniques to analyze data and is trained to perform specific tasks.

[0423] A "training method" is a function or process that uses historical communication data to teach an artificial intelligence model a particular writing style or pattern.

[0424] The "response draft generation button" is an interface element that the user can operate to start the process of automatically generating a reply draft.

[0425] A "user terminal" is a device operated by a user for sending e-mails and other communications.

[0426] "Received content" refers to the content of messages or emails received on the user's terminal.

[0427] "Emotion recognition results" are information obtained by inferring and analyzing a user's emotions from the content of comments and messages.

[0428] "Comment analysis" is the process of analyzing received comments to understand their content and generate appropriate responses.

[0429] A "means for automatically generating reply drafts" is a function or process for analyzing a comment or message and automatically generating a reply to that comment or message.

[0430] "Delivery means" refers to the function or process for transmitting the generated reply draft to the user terminal.

[0431] "Edit and confirm" is the process by which a user corrects an automatically generated draft and confirms that it should be sent.

[0432] The "means of transmitting communication" is the function or process for transmitting the revised or finalized text.

[0433] MODE FOR CARRYING OUT THE INVENTION

[0434] The present invention is a system for automatically generating reply proposals to comments in a content distribution service. Here, the specific method and operation of the system will be described in detail.

[0435] Server Processing

[0436] The server collects and preprocesses past communication data. During the data collection process, past comment data is retrieved through an API endpoint and saved in JSON format. The collected data is cleansed to remove unnecessary information. The preprocessed data is then used to train a generative AI model that learns each user's writing style. The artificial intelligence technology used here uses OpenAI's API.

[0437] Emotion Recognition and Sentence Generation

[0438] When the server receives a new comment from a user, it analyzes the content and performs emotion recognition. This emotion recognition is performed based on the text extracted from the user's comment, using OpenAI's emotion analysis model. The analysis results are reflected in the generation of reply proposals. An appropriate reply proposal is generated based on the received content and the emotion recognition results. In this process, a prompt sentence is used to instruct the generative AI model to generate a proposal.

[0439] Terminal handling

[0440] The device communicates with the server in response to user operations. When a user receives a comment and replies to it, the device sends the comment content to the server. When the server returns a reply draft and emotion recognition results, the device displays them to the user. The user can check the displayed draft and edit it as necessary. The edited and confirmed draft is sent from the device to the server, and a reply to the comment is finally sent.

[0441] User operations

[0442] When a user receives a comment on a content distribution service and needs to reply to that comment, they press the response draft button. This operation sends the comment content from the device to the server. The server generates a reply draft based on the comment content and emotion recognition results and returns it to the user's device. The user can then check the draft and edit it as necessary. Finally, by pressing the send button, the edited draft is finalized and sent back to the comment poster.

[0443] Specific examples

[0444] For example, if a user receives a comment such as "This news was really helpful!", the device sends the comment to the server, which then uses the following prompt to instruct the generative AI model to generate a sentence:

[0445] Example prompt sentence:

[0446] A user says 'This news was really helpful!'. You could reply with (emotion: joy):

[0447] The generated message, "Thank you for your feedback! We look forward to working with you in the future," is displayed on the user's device. The user can review this message, revise it as necessary, and then send it. This allows for an appropriate and prompt reply that takes the user's feelings into consideration.

[0448] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0449] Step 1:

[0450] The server collects past communication data.

[0451] Input: Historical comment data retrieved from an API endpoint.

[0452] How it works: The server collects data through the API and stores it in JSON format.

[0453] Output: Archived historical comment data.

[0454] Step 2:

[0455] The server pre-processes and cleanses the collected communication data.

[0456] Input: Collected historical comment data.

[0457] How it works: The server strips out unnecessary information from the data, leaving only the necessary fields. Specifically, it extracts the comment content and the identity of the commenter.

[0458] Output: Cleansed communication data.

[0459] Step 3:

[0460] The server uses the preprocessed data to train a generative AI model.

[0461] Input: Cleansed communication data.

[0462] How it works: The server uses the preprocessed data to train a generative AI model that learns each user's writing style. It uses OpenAI's API to generate and train the AI ​​model.

[0463] Output: A fully trained generative AI model.

[0464] Step 4:

[0465] The user receives the comment and presses the response draft button.

[0466] Input: The comment content received by the user.

[0467] How it works: The user presses the button on their device to create a response draft and send it to the server.

[0468] Output: The comment sent.

[0469] Step 5:

[0470] The server analyzes the received content and recognizes the emotion.

[0471] Input: The comment submitted.

[0472] How it works: The server uses OpenAI's sentiment analysis model to analyze the content of comments and recognize their sentiment.

[0473] Output: Recognized emotion data.

[0474] Step 6:

[0475] The server generates a draft response based on the received content and the emotion recognition results.

[0476] Input: Received comment content and recognized emotion data.

[0477] How it works: The server uses the generative AI model to generate a prompt, sends it to the AI ​​model, and generates an appropriate reply.

[0478] Output: The generated response draft.

[0479] Step 7:

[0480] The server delivers the generated response draft to the user terminal.

[0481] Input: The generated response draft.

[0482] Operation: The server sends the generated draft to the user terminal.

[0483] Output: A draft response displayed on the user's terminal.

[0484] Step 8:

[0485] The user reviews the draft response and edits it if necessary.

[0486] Input: Proposed response displayed on the terminal.

[0487] Action: The user reviews the draft response and makes any necessary corrections.

[0488] Output: Edited draft response.

[0489] Step 9:

[0490] The user presses the send button to send the reply draft.

[0491] Input: Edited draft response.

[0492] Action: The user presses the send button and the confirmed response draft is sent via the server.

[0493] Output: Final draft response sent.

[0494] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0495] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search<url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0496] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0497] [Second embodiment]

[0498] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0499] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0500] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0501] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0502] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0503] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0504] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0505] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0506] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0507] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0508] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0509] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0510] The present invention relates to a system for automatically generating reply drafts for electronic communications, which system functions around three central parties: a server, a terminal, and a user.

[0511] 1. Configuration

[0512] server

[0513] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the AI ​​model and generating sentences. Specifically, the server implements the following functions:

[0514] Data collection function: Collects past communication data from the company's internal email system.

[0515] Data preprocessing function: Cleanses collected data and removes unnecessary information.

[0516] Model training function: The cleansed data is used to train an artificial intelligence model that learns each user's writing style.

[0517] Text generation function: Analyzes the content of received communications and generates text suggestions based on artificial intelligence models.

[0518] Terminal

[0519] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[0520] Interface provision function: Displays a response draft creation button. When the user presses this button, the draft generation process begins.

[0521] Data transmission function: Sends the contents of received emails to the server.

[0522] Draft display function: Displays the draft sent from the server.

[0523] Draft editing function: Provides an interface for users to modify and edit the generated draft.

[0524] Sending function: The confirmed email draft is sent via the server.

[0525] User

[0526] A user is a person who operates a system and uses it according to a specific procedure, specifically by performing the following steps:

[0527] Check email: Check the email you received.

[0528] How to use the Create a Draft button: Press the Create a Draft button when replying.

[0529] Review and edit the transcript: Review the auto-generated transcript and make any necessary corrections.

[0530] Send Email: Send the revised / edited email draft.

[0531] 2. Program Processing

[0532] Server Processing

[0533] The server first collects past communication data from the company's email system. The collected data is pre-processed and cleansed to remove unnecessary information. An artificial intelligence model is then trained to learn each user's writing style and perspective. Once trained, the model is used to generate copy.

[0534] Terminal handling

[0535] When the user presses the "Create a draft" button, the device extracts the contents of the open email and sends it to the server. When the generated draft is returned from the server, the device displays it to the user. After the user confirms, edits, and confirms the draft, the device finally sends it.

[0536] User operations

[0537] The user opens the received email and, if a reply is necessary, presses the Create Message button. The user checks the displayed message and edits the content as necessary. The email is then sent with the final message.

[0538] 3. Specific Examples

[0539] Data collection and preprocessing

[0540] The server collects email data from the company's internal email server for the past year.

[0541] The server cleanses the collected data, removing unnecessary and personal information.

[0542] Training the model

[0543] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[0544] Copy generation and sending

[0545] The user opens the received email and presses the draft button.

[0546] The terminal extracts the mail contents and transmits them to the server.

[0547] The server generates a draft based on the received content and returns it to the terminal.

[0548] The terminal displays the generated draft to the user.

[0549] The user checks and edits the draft, and once it is confirmed, presses the send button.

[0550] The terminal transmits the confirmed email draft to the server, and finally transmits the email to the destination.

[0551] This system significantly reduces the user's workload and enables quick and accurate generation and sending of response emails.

[0552] The processing flow will be explained below.

[0553] Step 1:

[0554] The server collects past communication data from the company's email system, primarily email data from the past year, and collects it separately for each user.

[0555] Step 2:

[0556] The server stores the collected email data in a database, along with metadata such as the sender / receiver, date and time, and subject of the email.

[0557] Step 3:

[0558] The server performs spam filtering and personal information cleansing of the stored data, removing unnecessary and privacy-sensitive information.

[0559] Step 4:

[0560] The server then categorizes and tags the cleansed data, assigning tags to each email based on its content (e.g., formal, casual, urgent, etc.) to make the data easier to organize and analyze.

[0561] Step 5:

[0562] The server uses the cleansed and tagged data to train an artificial intelligence model, which uses natural language processing techniques to learn each user's writing style and vocabulary.

[0563] Step 6:

[0564] The terminal displays a message drafting button when the user opens the email client. When the user selects an email to reply to and presses the message drafting button, the terminal proceeds to the next step.

[0565] Step 7:

[0566] When the user presses the "Create a draft" button, the device extracts the contents of the open email, and this extracted data is used as key information for generating the draft.

[0567] Step 8:

[0568] The terminal sends the extracted email content to the server. The sent data includes the full text of the received email and metadata.

[0569] Step 9:

[0570] The server analyzes the content of the received email and generates drafts using an artificial intelligence model based on the corresponding user's writing style.

[0571] Step 10:

[0572] The server sends the generated message back to the terminal, which contains the necessary information for the reply message.

[0573] Step 11:

[0574] The terminal displays the received draft to the user, who can then check the content and make corrections or edits as necessary.

[0575] Step 12:

[0576] The user checks and edits the draft and finally confirms it. After confirming, they press the send button to proceed to the next step.

[0577] Step 13:

[0578] The terminal sends the finalized email draft to the server. The sent data includes the final email content.

[0579] Step 14:

[0580] The server sends the email to the specified recipient based on the final draft it received. After sending, it is saved in the database as a sending history.

[0581] By following these steps, users can easily and efficiently create and send correspondence drafts.

[0582] Example 1

[0583] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0584] In conventional electronic communication, creating replies is often done manually, resulting in a significant workload. Furthermore, some users have difficulty maintaining a consistent writing style and formality, resulting in inconsistencies in the quality of corporate communications. In these circumstances, there is a need for efficient and consistent replies. Furthermore, there is a need for a means to effectively utilize past communication data and automatically generate replies that reflect a user's unique writing style. To address this issue, the present invention provides a system for automatically generating efficient and consistent reply drafts.

[0585] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0586] In this invention, the server includes means for collecting historical communication data, means for pre-processing and cleansing the historical communication data, and means for using the pre-processed data to train a generative AI model that learns each user's writing style, thereby enabling efficient and consistent automated generation of reply drafts.

[0587] "Data collection" is the process of obtaining historical communication data from a particular system or server.

[0588] "Cleansing" is the process of removing unnecessary or incorrect information from collected data and arranging it into an appropriate format.

[0589] A "generative AI model" is an artificial intelligence model that learns the unique characteristics of users, such as their writing style and position, and automatically generates optimal reply drafts.

[0590] The "response drafting button" is a terminal interface element that the user operates to initiate the email reply drafting process.

[0591] "Reply generation" is the process of using a generative AI model to create appropriate reply suggestions based on collected and pre-processed data.

[0592] "Data transmission" is the process of sending and receiving extracted email content and generated drafts between the server and the terminal.

[0593] "Draft editing" is the process in which the user reviews the generated draft and makes corrections or edits as necessary.

[0594] "Communication sending" is the process of sending the finalized email draft to the intended recipient.

[0595] The present invention relates to a system for automatically generating reply drafts for electronic communications, which system functions around three main entities: a server, a terminal, and a user.

[0596] Server configuration and functions

[0597] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the generative AI model and generating sentences. Specifically, it implements the following functions:

[0598] Data collection function: The server collects historical email data from the company's mail system. For example, it periodically retrieves data for the past year using a Linux cron job or a Python script.

[0599] Data preprocessing function: The server uses a data cleansing tool (e.g., the Pandas library) to remove unnecessary information and personal information from the collected data.

[0600] Model training function: The server uses the cleansed data to train a generative AI model (e.g., GPT-3 or BERT model). This model learns each user's writing style and perspective and is used to automatically generate email reply suggestions.

[0601] Text generation function: The server generates an appropriate text based on the content of the received email and returns it to the terminal.

[0602] Device configuration and functions

[0603] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[0604] Interface provision function: The terminal displays a response draft creation button. When the user presses this button, the draft generation process begins.

[0605] Email content extraction function: When the user presses a button, the device extracts the contents of the open email and sends them to the server.

[0606] Draft reception function: Receives drafts generated from the server.

[0607] Text display function: The terminal displays the received text to the user, for example, using a GUI framework (JavaFX or Electron).

[0608] Copy editing function: Provides an interface (such as a rich text editor) for users to review and edit the copy.

[0609] Final sending function: The user resends the finalized draft to the server and finally sends the email.

[0610] User operation procedure

[0611] The user uses the system by following these steps:

[0612] 1. Check your email: Open your received email.

[0613] 2. How to use the Create a Draft button: Press the Create a Draft button when replying.

[0614] 3. Review and edit the transcript: Review the auto-generated transcript and make any necessary corrections.

[0615] 4. Send the email: Send the revised and edited email.

[0616] Specific examples

[0617] Data collection and preprocessing

[0618] The server collects email data from the company's internal email server for the past year.

[0619] The server cleanses the collected data, removing unnecessary and personal information.

[0620] Training the model

[0621] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[0622] Copy generation and sending

[0623] The user opens the received email and presses the draft button.

[0624] The terminal extracts the mail contents and transmits them to the server.

[0625] The server generates a draft based on the received content and returns it to the terminal.

[0626] The terminal displays the generated draft to the user.

[0627] The user checks and edits the draft, and once it is confirmed, presses the send button.

[0628] The terminal transmits the finalized email draft to the server, and finally transmits the email to the destination.

[0629] Prompt Sentence Examples

[0630] "I received an email inquiring about the status of a recent project. Please draft a response."

[0631] This system significantly reduces the user's workload and enables quick and accurate generation and sending of response emails.

[0632] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0633] Step 1: Collect historical communication data

[0634] The server collects email data from your company's mail server for the past year, using a Python script and the mail server's IMAP protocol to retrieve all emails for a specified period.

[0635] Input: Email server account information, period (e.g., past year)

[0636] Output: Captured email data (subject, body, sender / receiver information, etc.)

[0637] Step 2: Preprocess the data

[0638] The server cleanses the retrieved email data using the Pandas library, specifically by removing unnecessary header information and personal information and converting the data into an appropriate format.

[0639] Input: Captured email data

[0640] Output: Cleansed data (well-formatted data with unnecessary information removed)

[0641] Step 3: Training the generative AI model

[0642] The server uses the cleansed data to train a generative AI model (e.g., GPT-3 or BERT), applying machine learning algorithms to learn each user's writing style and vocabulary.

[0643] Input: Cleansed data

[0644] Output: A trained generative AI model

[0645] Step 4: Begin the response drafting process

[0646] The user presses the drafting button displayed on the terminal's email client, which starts the drafting process.

[0647] Input: User button presses

[0648] Output: Signal to start the copywriting process

[0649] Step 5: Extract the contents of the received email

[0650] The terminal extracts the content of the received email that the user currently has open, specifically, the subject, body, and sender information of the email.

[0651] Input: Incoming emails that the user has opened

[0652] Output: Extracted email content (subject, body, sender information)

[0653] Step 6: Send the email content to the server

[0654] The terminal sends the extracted email content to the server as an HTTP POST request.

[0655] Input: Extracted email content

[0656] Output: Request sent to server

[0657] Step 7: Server-generated copy

[0658] The server creates a prompt based on the content of the received email and generates a draft using a generative AI model. Specifically, it analyzes the content of the email, generates a prompt for the generative AI model, and then creates a reply draft based on that.

[0659] Input: Email content (subject, body, sender information)

[0660] Output: Generated reply draft

[0661] Step 8: Receiving and viewing the generated draft

[0662] The terminal displays the generated draft received from the server to the user, specifically, using a GUI framework to display the draft in an easy-to-read format.

[0663] Input: Generated reply

[0664] Output: The text displayed on the screen

[0665] Step 9: Review and edit your copy

[0666] The user checks the displayed draft and corrects or edits it as necessary. Specifically, the user edits the content using a rich text editor.

[0667] Input: Displayed text

[0668] Output: Edited draft

[0669] Step 10: Submit the final draft

[0670] When the user has finished editing, he / she presses the send button to confirm the final draft, which causes the device to resend the confirmed draft to the server and finally send the email.

[0671] Input: Edited text, click the send button

[0672] Output: Send the finalized draft and email it to the destination

[0673] (Application example 1)

[0674] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0675] Creating replies in conventional electronic communications requires a significant amount of time and effort from the user, making it difficult to respond quickly and accurately, especially during busy periods or when a large number of inquiries are received. Furthermore, it is difficult to maintain consistency in replies for each requester based on the user's writing style and past communication data. Furthermore, with the demand for rapid responses to inquiries, such as after-sales service, there is a lack of technology to efficiently generate response drafts.

[0676] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0677] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, and means for using the preprocessed data to train an artificial intelligence model that learns each user's writing style, thereby enabling means for automatically analyzing the content of the inquiry, means for generating a reply draft based on the content of the inquiry, means for displaying the generated reply draft to the user in an editable form, and means for automatically sending the finalized reply.

[0678] "Past communication data" refers to data on electronic communications that a user has conducted in the past.

[0679] "Cleansing" is a process of removing unnecessary information from collected data to make the data clean.

[0680] An "artificial intelligence model" is an algorithm that uses machine learning techniques to learn a user's writing style and specific response patterns.

[0681] The "draft creation button" is an interface that allows the user to automatically generate a reply draft.

[0682] A "user terminal" is an electronic device operated by a user that has the means to communicate with a server and send and receive data.

[0683] A "draft response" is a draft reply that is automatically generated based on the content of the received communication.

[0684] "Automatic analysis of inquiry content" is a technology that understands the content of received inquiries and identifies appropriate categories and subjects.

[0685] "Reply draft generation" is the process of automatically creating an appropriate reply based on the analyzed inquiry content.

[0686] "Editable Draft Display" is a function that presents the generated reply draft to the user, allowing the user to modify the text as needed.

[0687] "Auto-send" is a function that allows the system to automatically send a confirmed reply.

[0688] The present invention relates to a system for automatically generating reply drafts in electronic communications, which functions around three main entities: a server, a terminal, and a user.

[0689] 1. Configuration

[0690] server

[0691] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the generative AI model and generating text suggestions. Specifically, it implements the following functions:

[0692] Data collection function: The server collects past communication data from the company's internal communication systems.

[0693] Data preprocessing function: The server cleanses the collected data and removes unnecessary information.

[0694] Model training function: The server uses the cleansed data to train a generative AI model to learn each user's writing style.

[0695] Message generation function: The server analyzes the received communication content and generates a reply message based on the generative AI model.

[0696] Automatic analysis of inquiry content: The server automatically analyzes the inquiry content and identifies the appropriate category and subject.

[0697] Terminal

[0698] The terminal is a device operated by the user, which communicates with the server to send and receive data. The terminal has the following functions:

[0699] Interface provision function: The terminal displays a response draft creation button, which the user can press to start the draft generation process.

[0700] Data transmission function: The terminal transmits the received communication content to the server.

[0701] Draft display function: The terminal displays the draft sent from the server.

[0702] Draft editing function: The terminal provides an interface for users to modify and edit the generated draft.

[0703] Send function: Sends the final draft of the text that the user has confirmed.

[0704] User

[0705] A user is a person who operates the system and uses it according to a specific procedure, specifically by performing the following steps:

[0706] Confirming the inquiry: The user confirms the received inquiry email.

[0707] Operation of the Create Copy Button: When a user wants to reply, he or she presses the Create Copy button.

[0708] Review and edit the draft: The user reviews the automatically generated draft and corrects it if necessary.

[0709] Sending email: The user sends the corrected and edited reply.

[0710] 2. Hardware and Software Used

[0711] Hardware:

[0712] Server machine: A machine with a high-performance CPU and GPU

[0713] Device: Devices such as smartphones and PCs

[0714] software:

[0715] TensorFlow: Used for training and inferencing generative AI models

[0716] NLTK: Used to preprocess data for natural language processing

[0717] sklearn:Used for splitting data function

[0718] 3. Data processing and data calculation

[0719] The server preprocesses past communication data collected by the server and removes unnecessary information. Using this cleansed data, the server trains an AI model that learns each user's writing style. The generated AI model analyzes the content of received communications and automatically generates reply drafts. It also automatically analyzes the content of inquiries and provides appropriate reply drafts.

[0720] 4. Specific Examples

[0721] As an example of actual system operation, the following prompt statements can be used:

[0722] Inquiries regarding this purchase

[0723] "The product never arrived"

[0724] Please tell me the return procedure.

[0725] In this way, the present invention significantly reduces the user's workload and enables quick and accurate generation and transmission of response emails.

[0726] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0727] Step 1:

[0728] The server collects past communication data. Specifically, it collects email data from the company's internal communication system for the past year and stores it in a database. At this stage, the input is internal email data, and the output is the collected, unprocessed data.

[0729] Step 2:

[0730] The server preprocesses and cleanses the collected data. Specifically, it uses the natural language processing library NLTK to remove unnecessary words (stop words) and personal information from the collected data. At this stage, the collected raw data is the input, and the cleansed data is the output.

[0731] Step 3:

[0732] The server uses the preprocessed data to train a generative AI model. Specifically, it uses TensorFlow to segment the data and an LSTM model to learn each user's writing style. At this stage, the input is the cleansed data and the output is a trained AI model.

[0733] Step 4:

[0734] When a user receives an inquiry email, the terminal displays the received email and the user presses the button to create a response draft. The received email content is the input, and the content is sent to the server as the output.

[0735] Step 5:

[0736] The server analyzes the content of the received email and automatically generates a reply. Specifically, it analyzes the inquiry content and generates an appropriate reply using a trained AI model. At this stage, the received email content is the input, and the generated reply is the output.

[0737] Step 6:

[0738] The terminal displays the generated reply draft to the user. The user can then review this draft and make corrections or edits as necessary. For example, in response to an inquiry such as "The arrival of the product is delayed," the following reply is generated: "We apologize for any inconvenience caused by your order. We are currently checking the delivery status. Could you please wait a moment?"

[0739] Step 7:

[0740] When the user finally confirms the draft, the terminal sends this confirmed draft to the server, which then automatically sends the confirmed draft as a reply to the original query. At this stage, the confirmed draft is the input, and a reply email is sent to the user as the output.

[0741] The above steps significantly reduce the user's workload and enable quick and accurate generation and sending of response emails.

[0742] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0743] The present invention relates to a system for automatically generating reply proposals for electronic communications, in particular incorporating an emotion engine. The system functions around three parties: a server, a terminal, and a user.

[0744] 1. Configuration

[0745] server

[0746] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the AI ​​model and generating sentences. Specifically, the server implements the following functions:

[0747] Data collection function: Collects past communication data from the company's internal email system.

[0748] Data preprocessing function: Cleanses collected data and removes unnecessary information.

[0749] Model training function: The cleansed data is used to train an artificial intelligence model that learns each user's writing style.

[0750] Text generation function: Analyzes the content of received communications and generates text suggestions based on artificial intelligence models.

[0751] Emotion recognition: Recognizes user emotions using an emotion engine and reflects them in the generated text.

[0752] Terminal

[0753] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[0754] Interface provision function: Displays a response draft creation button and emotion recognition results. When the user presses this button, the draft generation process begins.

[0755] Data transmission function: Sends the contents of received emails to the server.

[0756] Text display function: Displays the text sent from the server and emotion recognition results.

[0757] Draft editing function: Provides an interface for users to modify and edit the generated draft.

[0758] Sending function: The confirmed email draft is sent via the server.

[0759] User

[0760] A user is a person who operates a system and uses it according to a specific procedure, specifically by performing the following steps:

[0761] Check email: Check the email you received.

[0762] How to use the Create a Draft button: Press the Create a Draft button when replying.

[0763] Review and edit the text and emotion recognition results: Review the automatically generated text and emotion recognition results and make corrections as necessary.

[0764] Send Email: Send the revised / edited email draft.

[0765] 2. Program Processing

[0766] Server Processing

[0767] The server first collects past communication data from the company's email system. The collected data is pre-processed and cleansed, and unnecessary information is removed. An artificial intelligence model is then trained to learn each user's writing style and perspective. Once trained, the model is used to generate text suggestions. It also uses an emotion engine to recognize user emotions and reflect them in the generated text.

[0768] Terminal handling

[0769] When the user presses the "Create a message" button, the device extracts the contents of the open email and sends it to the server. When the generated message and emotion recognition results are returned from the server, the device displays them to the user. After the user confirms, edits, and confirms the message and emotion recognition results, the device finally sends the message.

[0770] User operations

[0771] The user opens the received email and, if a reply is necessary, presses the "Create a message" button. The user checks the displayed message and emotion recognition results, and edits the content as necessary. The email is then sent with the finalized message.

[0772] 3. Specific Examples

[0773] Data collection and preprocessing

[0774] The server collects email data from the company's internal email server for the past year.

[0775] The server cleanses the collected data, removing unnecessary and personal information.

[0776] Training the model

[0777] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[0778] emotion recognition

[0779] The server uses an emotion engine to recognize the user's emotions, and this emotion data is reflected in the copy generation.

[0780] Copy generation and sending

[0781] The user opens the received email and presses the draft button.

[0782] The terminal extracts the mail contents and transmits them to the server.

[0783] The server generates a draft based on the received content and the emotion recognition results and sends it back to the terminal.

[0784] The terminal displays the generated draft and emotion recognition results to the user.

[0785] The user checks and edits the draft, and once it is confirmed, presses the send button.

[0786] The terminal transmits the confirmed email draft to the server, and finally transmits the email to the destination.

[0787] This system significantly reduces the user's workload and enables the system to generate and send response emails quickly and accurately. Furthermore, the introduction of an emotion engine allows the system to provide appropriate message suggestions based on the user's emotions, enabling more human-like communication.

[0788] The processing flow will be explained below.

[0789] Step 1:

[0790] The server collects past communication data from the company's internal email system, primarily email data from the past year, and collects it separately for each user.

[0791] Step 2:

[0792] The server stores the collected email data in a database, along with metadata such as the sender / receiver, date and time, and subject of the email.

[0793] Step 3:

[0794] The server cleanses the stored data, filtering out spam and removing unnecessary and personal information, and uses natural language processing technology to format the data.

[0795] Step 4:

[0796] The server categorizes and tags the cleansed data, assigning tags to each email based on its content (e.g., formal, casual, urgent, etc.) to make the data easier to organize and analyze.

[0797] Step 5:

[0798] The server uses the cleansed and tagged data to train an artificial intelligence model, which uses natural language processing techniques to learn each user's writing style and vocabulary, enabling personalized responses.

[0799] Step 6:

[0800] The server trains an emotion engine to recognize user emotions by analyzing keystrokes, mouse movements, and voice inputs, and incorporates this emotion information into the copy generation model.

[0801] Step 7:

[0802] When the user opens the email client, the device displays a message creation button. When the user selects the email to reply to and presses the message creation button, the device proceeds to the next step.

[0803] Step 8:

[0804] When the user presses the "Create a copy" button, the device extracts the contents of the open email, and this extracted data is used as key information for generating the copy.

[0805] Step 9:

[0806] The terminal transmits the extracted email content and emotion data generated by the emotion engine to the server. The transmitted data includes the full text of the received email, metadata, and emotion information.

[0807] Step 10:

[0808] The server analyzes the content and emotional information of the received email and generates a draft message using an AI model based on the corresponding user's writing style, taking into account the user's emotional state.

[0809] Step 11:

[0810] The server then sends the generated message back to the device, which includes the necessary information for the email reply and the recommended text from the emotion engine.

[0811] Step 12:

[0812] The device displays the received draft to the user, who can then check the content and make corrections or edits as necessary.

[0813] Step 13:

[0814] The user checks and edits the draft and finally confirms it. After confirming it, the user presses the submit button to proceed to the next step.

[0815] Step 14:

[0816] The terminal sends the finalized email draft to the server. The sent data includes the final email content.

[0817] Step 15:

[0818] The server sends the email to the specified recipient based on the final draft received. After sending, it is recorded in the database as a sending history.

[0819] Example 2

[0820] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0821] Modern business communication requires fast and appropriate responses to large volumes of emails. However, for users who receive emails frequently, coming up with an appropriate response for each email can be a significant burden. It is also important for email replies to reflect the user's emotions, which requires the user to incorporate their own feelings into the text. Furthermore, it is even more difficult to reply in a consistent style using past communication data. To address these challenges, a system is needed that automatically generates emotionally relevant response drafts and allows users to easily edit and send them.

[0822] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0823] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, means for training an artificial intelligence model that learns each user's writing style using the preprocessed data, means for reflecting emotion data obtained from an emotion recognition engine in message draft generation, and means for delivering the generated response drafts and emotion recognition results to the user terminal. This allows for the automatic generation of quick and appropriate email responses, and the user only needs to edit them, significantly reducing the burden on the user. Furthermore, by reflecting the emotion recognition results, more human-like communication can be achieved.

[0824] "Past communication data" refers to records of communication such as emails sent and received in the past, and includes the content of exchanges between users.

[0825] "Preprocessing" refers to the process of removing unnecessary information from collected data and preparing the data in a format that is easy to analyze.

[0826] "Cleansing" refers to the process of removing noise and unnecessary parts from data and extracting only the necessary information.

[0827] "Artificial intelligence model" refers to a program or algorithm that learns a user's writing style or specific patterns through machine learning or deep learning.

[0828] "Response drafting button" refers to a button that a user operates on the interface to cause the system to initiate automatic response drafting.

[0829] A "user terminal" is a device operated by a user, and refers to computer equipment such as a personal computer or smartphone.

[0830] An "emotion recognition engine" refers to an algorithm or software that analyzes and classifies user emotions from text data.

[0831] "Emotion data" refers to user emotion information extracted by an emotion recognition engine, and includes emotion categories such as "joy," "sadness," and "anger," for example.

[0832] "Message generation" refers to the process of using an artificial intelligence model to automatically create a message response based on the received content and sentiment data.

[0833] "Delivery" refers to sending the generated text and emotion recognition results to a user terminal via a network.

[0834] "Communication" refers to the act of sending and receiving e-mails, etc. over a network.

[0835] An embodiment of this invention provides a system that collects and cleans past communication data, trains an artificial intelligence model based on that data, generates drafts using an emotion recognition engine, and allows users to review, edit, and send them. The components of this system and the specific processing flow are described below.

[0836] Server Roles

[0837] The server plays a central role in the system, collecting data, preprocessing, training models, generating text, and recognizing emotions using the following hardware and software:

[0838] Hardware: A powerful server computer (e.g., a server with dual Xeon processors)

[0839] Software: Databases (e.g., MongoDB), generative AI models (e.g., GPT-3), emotion recognition engines (e.g., Microsoft Azure Cognitive Services)

[0840] Data collection and preprocessing

[0841] The server collects email data from the company's internal email system for the past year. This is achieved by using the email server's API to retrieve emails and store them in a local database. The collected data then goes through a cleansing process to remove unnecessary parts (such as email headers and signatures) and extract only the body of the email.

[0842] Training the model

[0843] The server uses the cleansed email data to train a generative AI model that learns the writing style and tone of each user's emails using large datasets and advanced machine learning algorithms.

[0844] emotion recognition

[0845] The server uses an emotion recognition engine to analyze the user's emotions. Specifically, it uses emotional data extracted from the content of the email (e.g., "joy," "sadness," "anger," etc.) to generate more appropriate text.

[0846] Copy generation and sending

[0847] When a user checks an incoming email and presses the message creation button if a reply is necessary, the device extracts the email content and sends it to the server. The server generates a message based on the received email content and the emotion recognition results. The generated message and emotion recognition results are sent back to the device and displayed to the user.

[0848] The user can check and edit the displayed message, correct it as appropriate, and then press the send button to send the final message. At this time, the device sends the message via the server.

[0849] Specific examples

[0850] For example, you can use prompts like the following to instruct a generative AI model to generate specific sentences:

[0851] "Analyze the content of the received email and generate a polite reply that includes an apology. Based on the results of the emotion engine, create a reply that takes into consideration the user's emotions."

[0852] Based on this specific prompt, the server can automatically generate appropriate reply ideas using a generative AI model, significantly reducing the user's effort. Furthermore, by incorporating emotion recognition, more human-like communication can be achieved.

[0853] In this way, the present invention helps users to respond to emails quickly and accurately, and improves the efficiency of email reply work.

[0854] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0855] Step 1: Collect data

[0856] The server collects email data from the company's internal email system for the past year. Specifically, it uses the email server's API to retrieve emails from the past year. The retrieved email data is stored in a local database (e.g., MongoDB) in JSON format.

[0857] Input: Email data from the internal email system (period: past year)

[0858] Output: Email data stored in a local database (JSON format)

[0859] Step 2: Preprocessing the data

[0860] The server cleanses the collected email data. Specifically, it uses regular expressions to remove unnecessary information such as email headers and signatures, and extracts only the body of the email. The preprocessed data is then saved in JSON format.

[0861] Input: Collected email data (JSON format)

[0862] Output: Cleansed email data (JSON format)

[0863] Step 3: Train the model

[0864] The server trains a generative AI model (e.g., GPT-3) using the cleansed email data. Specifically, it splits the data into batches and trains the model over multiple epochs. The trained model learns the writing style and tone of each user.

[0865] Input: Cleansed email data (JSON format)

[0866] Output: A trained generative AI model

[0867] Step 4: Process incoming email

[0868] The user checks the received email and presses the message creation button if a reply is necessary. The terminal extracts the contents of the opened email and sends them to the server.

[0869] Input: Received email content (text format)

[0870] Output: Email content sent to the server (text format)

[0871] Step 5: Recognize emotions

[0872] The server sends the received email content to an emotion recognition engine to analyze the user's emotions. The recognized emotion data is classified into categories such as "joy," "sadness," and "anger," and is used in the copy generation process.

[0873] Input: Received email content (text format)

[0874] Output: Recognized emotion data (categorical format)

[0875] Step 6: Generate copy

[0876] The server provides a prompt to the generative AI model based on the content of the received email and the emotion data, and generates a draft. Example prompt: "Analyze the content of the received email and generate a polite reply that includes an apology. Based on the results of the emotion engine, please create a draft that takes the user's emotions into consideration."

[0877] Input: received email content (text format), recognized emotion data (category format), prompt text

[0878] Output: Generated draft (text format)

[0879] Step 7: Distributing the copy

[0880] The server delivers the generated text and emotion recognition results to the terminal, which then displays them to the user.

[0881] Input: Generated text (text format), emotion recognition results (category format)

[0882] Output: Text and emotion recognition results delivered to the device (text format)

[0883] Step 8: Edit and submit your copy

[0884] The user checks the draft displayed on the device and edits it as necessary. After that, the user presses the send button, and the edited and confirmed draft is sent from the device via the server. It is finally sent as an email.

[0885] Input: Generated draft (text format)

[0886] Output: Final email draft sent (email format)

[0887] (Application example 2)

[0888] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0889] While automated systems for generating reply proposals in electronic communications can significantly reduce time and effort, conventional systems often do not take emotion recognition into account, which can result in inappropriate replies. Furthermore, while there is a growing need for fast and appropriate replies to user comments in content distribution services, current methods suffer from slow response times. The present invention aims to solve these problems by generating fast and appropriate replies while taking user emotions into account.

[0890] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0891] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, and means for training an artificial intelligence model that learns each user's writing style using the preprocessed data. This makes it possible to generate an optimal writing style for each user based on past data. Also, in this invention, the server includes means for generating a response draft based on received content and emotion recognition results, and means for automatically analyzing comments and generating reply drafts in a content distribution service. This makes it possible to take user emotions into consideration and generate appropriate and prompt reply drafts in a content distribution service.

[0892] "Historical communications data" is a collection of previous emails, messages, and other communications records.

[0893] "Means of collection" refers to the functions and processes for importing past communication data into the server.

[0894] "Preprocessing and cleansing means" refers to functions and processes for removing unnecessary information from collected data and preparing the data.

[0895] An "artificial intelligence model" is a computer program that uses machine learning techniques to analyze data and is trained to perform specific tasks.

[0896] A "training method" is a function or process that uses historical communication data to teach an artificial intelligence model a particular writing style or pattern.

[0897] The "response draft generation button" is an interface element that the user can operate to start the process of automatically generating a reply draft.

[0898] A "user terminal" is a device operated by a user for sending e-mails and other communications.

[0899] "Received content" refers to the content of messages or emails received on the user's terminal.

[0900] "Emotion recognition results" are information obtained by inferring and analyzing a user's emotions from the content of comments and messages.

[0901] "Comment analysis" is the process of analyzing received comments to understand their content and generate appropriate responses.

[0902] A "means for automatically generating reply drafts" is a function or process for analyzing a comment or message and automatically generating a reply to that comment or message.

[0903] "Delivery means" refers to the function or process for transmitting the generated reply draft to the user terminal.

[0904] "Edit and confirm" is the process by which a user corrects an automatically generated draft and confirms that it should be sent.

[0905] The "means of transmitting communication" is the function or process for transmitting the revised or finalized text.

[0906] MODE FOR CARRYING OUT THE INVENTION

[0907] The present invention is a system for automatically generating reply proposals to comments in a content distribution service. Here, the specific method and operation of the system will be described in detail.

[0908] Server Processing

[0909] The server collects and preprocesses past communication data. During the data collection process, past comment data is retrieved through an API endpoint and saved in JSON format. The collected data is cleansed to remove unnecessary information. The preprocessed data is then used to train a generative AI model that learns each user's writing style. The artificial intelligence technology used here uses OpenAI's API.

[0910] Emotion Recognition and Sentence Generation

[0911] When the server receives a new comment from a user, it analyzes the content and performs emotion recognition. This emotion recognition is performed based on the text extracted from the user's comment, using OpenAI's emotion analysis model. The analysis results are reflected in the generation of reply proposals. An appropriate reply proposal is generated based on the received content and the emotion recognition results. In this process, a prompt sentence is used to instruct the generative AI model to generate a proposal.

[0912] Terminal handling

[0913] The device communicates with the server in response to user operations. When a user receives a comment and replies to it, the device sends the comment content to the server. When the server returns a reply draft and emotion recognition results, the device displays them to the user. The user can check the displayed draft and edit it as necessary. The edited and confirmed draft is sent from the device to the server, and a reply to the comment is finally sent.

[0914] User operations

[0915] When a user receives a comment on a content distribution service and needs to reply to that comment, they press the response draft button. This operation sends the comment content from the device to the server. The server generates a reply draft based on the comment content and emotion recognition results and returns it to the user's device. The user can then check the draft and edit it as necessary. Finally, by pressing the send button, the edited draft is finalized and sent back to the comment poster.

[0916] Specific examples

[0917] For example, if a user receives a comment such as "This news was really helpful!", the device sends the comment to the server, which then uses the following prompt to instruct the generative AI model to generate a sentence:

[0918] Example prompt sentence:

[0919] A user says 'This news was really helpful!'. You could reply with (emotion: joy):

[0920] The generated message, "Thank you for your feedback! We look forward to working with you in the future," is displayed on the user's device. The user can review this message, revise it as necessary, and then send it. This allows for an appropriate and prompt reply that takes the user's feelings into consideration.

[0921] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0922] Step 1:

[0923] The server collects past communication data.

[0924] Input: Historical comment data retrieved from an API endpoint.

[0925] How it works: The server collects data through the API and stores it in JSON format.

[0926] Output: Archived historical comment data.

[0927] Step 2:

[0928] The server pre-processes and cleanses the collected communication data.

[0929] Input: Collected historical comment data.

[0930] How it works: The server strips out unnecessary information from the data, leaving only the necessary fields. Specifically, it extracts the comment content and the identity of the commenter.

[0931] Output: Cleansed communication data.

[0932] Step 3:

[0933] The server uses the preprocessed data to train a generative AI model.

[0934] Input: Cleansed communication data.

[0935] How it works: The server uses the preprocessed data to train a generative AI model that learns each user's writing style. It uses OpenAI's API to generate and train the AI ​​model.

[0936] Output: A fully trained generative AI model.

[0937] Step 4:

[0938] The user receives the comment and presses the response draft button.

[0939] Input: The comment content received by the user.

[0940] How it works: The user presses the button on their device to create a response draft and send it to the server.

[0941] Output: The comment sent.

[0942] Step 5:

[0943] The server analyzes the received content and recognizes the emotion.

[0944] Input: The comment submitted.

[0945] How it works: The server uses OpenAI's sentiment analysis model to analyze the content of comments and recognize their sentiment.

[0946] Output: Recognized emotion data.

[0947] Step 6:

[0948] The server generates a draft response based on the received content and the emotion recognition results.

[0949] Input: Received comment content and recognized emotion data.

[0950] How it works: The server uses the generative AI model to generate a prompt, sends it to the AI ​​model, and generates an appropriate reply.

[0951] Output: The generated response draft.

[0952] Step 7:

[0953] The server delivers the generated response draft to the user terminal.

[0954] Input: The generated response draft.

[0955] Operation: The server sends the generated draft to the user terminal.

[0956] Output: A draft response displayed on the user's terminal.

[0957] Step 8:

[0958] The user reviews the draft response and edits it if necessary.

[0959] Input: Proposed response displayed on the terminal.

[0960] Action: The user reviews the draft response and makes any necessary corrections.

[0961] Output: Edited draft response.

[0962] Step 9:

[0963] The user presses the send button to send the reply draft.

[0964] Input: Edited draft response.

[0965] Action: The user presses the send button and the confirmed response draft is sent via the server.

[0966] Output: Final draft response sent.

[0967] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0968] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0969] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0970] [Third embodiment]

[0971] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0972] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0973] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0974] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0975] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0976] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0977] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0978] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0979] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0980] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0981] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0982] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0983] The present invention relates to a system for automatically generating reply drafts for electronic communications, which system functions around three central parties: a server, a terminal, and a user.

[0984] 1. Configuration

[0985] server

[0986] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the AI ​​model and generating sentences. Specifically, the server implements the following functions:

[0987] Data collection function: Collects past communication data from the company's internal email system.

[0988] Data preprocessing function: Cleanses collected data and removes unnecessary information.

[0989] Model training function: The cleansed data is used to train an artificial intelligence model that learns each user's writing style.

[0990] Text generation function: Analyzes the content of received communications and generates text suggestions based on artificial intelligence models.

[0991] Terminal

[0992] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[0993] Interface provision function: Displays a response draft creation button. When the user presses this button, the draft generation process begins.

[0994] Data transmission function: Sends the contents of received emails to the server.

[0995] Draft display function: Displays the draft sent from the server.

[0996] Draft editing function: Provides an interface for users to modify and edit the generated draft.

[0997] Sending function: The confirmed email draft is sent via the server.

[0998] User

[0999] A user is a person who operates a system and uses it according to a specific procedure, specifically by performing the following steps:

[1000] Check email: Check the email you received.

[1001] How to use the Create a Draft button: Press the Create a Draft button when replying.

[1002] Review and edit the transcript: Review the auto-generated transcript and make any necessary corrections.

[1003] Send Email: Send the revised / edited email draft.

[1004] 2. Program Processing

[1005] Server Processing

[1006] The server first collects past communication data from the company's email system. The collected data is pre-processed and cleansed to remove unnecessary information. An artificial intelligence model is then trained to learn each user's writing style and perspective. Once trained, the model is used to generate copy.

[1007] Terminal handling

[1008] When the user presses the "Create a draft" button, the device extracts the contents of the open email and sends it to the server. When the generated draft is returned from the server, the device displays it to the user. After the user confirms, edits, and confirms the draft, the device finally sends it.

[1009] User operations

[1010] The user opens the received email and, if a reply is necessary, presses the Create Message button. The user checks the displayed message and edits the content as necessary. The email is then sent with the final message.

[1011] 3. Specific Examples

[1012] Data collection and preprocessing

[1013] The server collects email data from the company's internal email server for the past year.

[1014] The server cleanses the collected data, removing unnecessary and personal information.

[1015] Training the model

[1016] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[1017] Copy generation and sending

[1018] The user opens the received email and presses the draft button.

[1019] The terminal extracts the mail contents and transmits them to the server.

[1020] The server generates a draft based on the received content and returns it to the terminal.

[1021] The terminal displays the generated draft to the user.

[1022] The user checks and edits the draft, and once it is confirmed, presses the send button.

[1023] The terminal transmits the confirmed email draft to the server, and finally transmits the email to the destination.

[1024] This system significantly reduces the user's workload and enables quick and accurate generation and sending of response emails.

[1025] The processing flow will be explained below.

[1026] Step 1:

[1027] The server collects past communication data from the company's email system, primarily email data from the past year, and collects it separately for each user.

[1028] Step 2:

[1029] The server stores the collected email data in a database, along with metadata such as the sender / receiver, date and time, and subject of the email.

[1030] Step 3:

[1031] The server performs spam filtering and personal information cleansing of the stored data, removing unnecessary and privacy-sensitive information.

[1032] Step 4:

[1033] The server then categorizes and tags the cleansed data, assigning tags to each email based on its content (e.g., formal, casual, urgent, etc.) to make the data easier to organize and analyze.

[1034] Step 5:

[1035] The server uses the cleansed and tagged data to train an artificial intelligence model, which uses natural language processing techniques to learn each user's writing style and vocabulary.

[1036] Step 6:

[1037] The terminal displays a message drafting button when the user opens the email client. When the user selects an email to reply to and presses the message drafting button, the terminal proceeds to the next step.

[1038] Step 7:

[1039] When the user presses the "Create a draft" button, the device extracts the contents of the open email, and this extracted data is used as key information for generating the draft.

[1040] Step 8:

[1041] The terminal sends the extracted email content to the server. The sent data includes the full text of the received email and metadata.

[1042] Step 9:

[1043] The server analyzes the content of the received email and generates drafts using an artificial intelligence model based on the corresponding user's writing style.

[1044] Step 10:

[1045] The server sends the generated message back to the terminal, which contains the necessary information for the reply message.

[1046] Step 11:

[1047] The terminal displays the received draft to the user, who can then check the content and make corrections or edits as necessary.

[1048] Step 12:

[1049] The user checks and edits the draft and finally confirms it. After confirming, they press the send button to proceed to the next step.

[1050] Step 13:

[1051] The terminal sends the finalized email draft to the server. The sent data includes the final email content.

[1052] Step 14:

[1053] The server sends the email to the specified recipient based on the final draft it received. After sending, it is saved in the database as a sending history.

[1054] By following these steps, users can easily and efficiently create and send correspondence drafts.

[1055] Example 1

[1056] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1057] In conventional electronic communication, creating replies is often done manually, resulting in a significant workload. Furthermore, some users have difficulty maintaining a consistent writing style and formality, resulting in inconsistencies in the quality of corporate communications. In these circumstances, there is a need for efficient and consistent replies. Furthermore, there is a need for a means to effectively utilize past communication data and automatically generate replies that reflect a user's unique writing style. To address this issue, the present invention provides a system for automatically generating efficient and consistent reply drafts.

[1058] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1059] In this invention, the server includes means for collecting historical communication data, means for pre-processing and cleansing the historical communication data, and means for using the pre-processed data to train a generative AI model that learns each user's writing style, thereby enabling efficient and consistent automated generation of reply drafts.

[1060] "Data collection" is the process of obtaining historical communication data from a particular system or server.

[1061] "Cleansing" is the process of removing unnecessary or incorrect information from collected data and arranging it into an appropriate format.

[1062] A "generative AI model" is an artificial intelligence model that learns the unique characteristics of users, such as their writing style and position, and automatically generates optimal reply drafts.

[1063] The "response drafting button" is a terminal interface element that the user operates to initiate the email reply drafting process.

[1064] "Reply generation" is the process of using a generative AI model to create appropriate reply suggestions based on collected and pre-processed data.

[1065] "Data transmission" is the process of sending and receiving extracted email content and generated drafts between the server and the terminal.

[1066] "Draft editing" is the process in which the user reviews the generated draft and makes corrections or edits as necessary.

[1067] "Communication sending" is the process of sending the finalized email draft to the intended recipient.

[1068] The present invention relates to a system for automatically generating reply drafts for electronic communications, which system functions around three main entities: a server, a terminal, and a user.

[1069] Server configuration and functions

[1070] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the generative AI model and generating sentences. Specifically, it implements the following functions:

[1071] Data collection function: The server collects historical email data from the company's mail system. For example, it periodically retrieves data for the past year using a Linux cron job or a Python script.

[1072] Data preprocessing function: The server uses a data cleansing tool (e.g., the Pandas library) to remove unnecessary information and personal information from the collected data.

[1073] Model training function: The server uses the cleansed data to train a generative AI model (e.g., GPT-3 or BERT model). This model learns each user's writing style and perspective and is used to automatically generate email reply suggestions.

[1074] Text generation function: The server generates an appropriate text based on the content of the received email and returns it to the terminal.

[1075] Device configuration and functions

[1076] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[1077] Interface provision function: The terminal displays a response draft creation button. When the user presses this button, the draft generation process begins.

[1078] Email content extraction function: When the user presses a button, the device extracts the contents of the open email and sends them to the server.

[1079] Draft reception function: Receives drafts generated from the server.

[1080] Text display function: The terminal displays the received text to the user, for example, using a GUI framework (JavaFX or Electron).

[1081] Copy editing function: Provides an interface (such as a rich text editor) for users to review and edit the copy.

[1082] Final sending function: The user resends the finalized draft to the server and finally sends the email.

[1083] User operation procedure

[1084] The user uses the system by following these steps:

[1085] 1. Check your email: Open your received email.

[1086] 2. How to use the Create a Draft button: Press the Create a Draft button when replying.

[1087] 3. Review and edit the transcript: Review the auto-generated transcript and make any necessary corrections.

[1088] 4. Send the email: Send the revised and edited email.

[1089] Specific examples

[1090] Data collection and preprocessing

[1091] The server collects email data from the company's internal email server for the past year.

[1092] The server cleanses the collected data, removing unnecessary and personal information.

[1093] Training the model

[1094] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[1095] Copy generation and sending

[1096] The user opens the received email and presses the draft button.

[1097] The terminal extracts the mail contents and transmits them to the server.

[1098] The server generates a draft based on the received content and returns it to the terminal.

[1099] The terminal displays the generated draft to the user.

[1100] The user checks and edits the draft, and once it is confirmed, presses the send button.

[1101] The terminal transmits the finalized email draft to the server, and finally transmits the email to the destination.

[1102] Prompt Sentence Examples

[1103] "I received an email inquiring about the status of a recent project. Please draft a response."

[1104] This system significantly reduces the user's workload and enables quick and accurate generation and sending of response emails.

[1105] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1106] Step 1: Collect historical communication data

[1107] The server collects email data from your company's mail server for the past year, using a Python script and the mail server's IMAP protocol to retrieve all emails for a specified period.

[1108] Input: Email server account information, period (e.g., past year)

[1109] Output: Captured email data (subject, body, sender / receiver information, etc.)

[1110] Step 2: Preprocess the data

[1111] The server cleanses the retrieved email data using the Pandas library, specifically by removing unnecessary header information and personal information and converting the data into an appropriate format.

[1112] Input: Captured email data

[1113] Output: Cleansed data (well-formatted data with unnecessary information removed)

[1114] Step 3: Training the generative AI model

[1115] The server uses the cleansed data to train a generative AI model (e.g., GPT-3 or BERT), applying machine learning algorithms to learn each user's writing style and vocabulary.

[1116] Input: Cleansed data

[1117] Output: A trained generative AI model

[1118] Step 4: Begin the response drafting process

[1119] The user presses the drafting button displayed on the terminal's email client, which starts the drafting process.

[1120] Input: User button presses

[1121] Output: Signal to start the copywriting process

[1122] Step 5: Extract the contents of the received email

[1123] The terminal extracts the content of the received email that the user currently has open, specifically, the subject, body, and sender information of the email.

[1124] Input: Incoming emails that the user has opened

[1125] Output: Extracted email content (subject, body, sender information)

[1126] Step 6: Send the email content to the server

[1127] The terminal sends the extracted email content to the server as an HTTP POST request.

[1128] Input: Extracted email content

[1129] Output: Request sent to server

[1130] Step 7: Server-generated copy

[1131] The server creates a prompt based on the content of the received email and generates a draft using a generative AI model. Specifically, it analyzes the content of the email, generates a prompt for the generative AI model, and then creates a reply draft based on that.

[1132] Input: Email content (subject, body, sender information)

[1133] Output: Generated reply draft

[1134] Step 8: Receiving and viewing the generated draft

[1135] The terminal displays the generated draft received from the server to the user, specifically, using a GUI framework to display the draft in an easy-to-read format.

[1136] Input: Generated reply

[1137] Output: The text displayed on the screen

[1138] Step 9: Review and edit your copy

[1139] The user checks the displayed draft and corrects or edits it as necessary. Specifically, the user edits the content using a rich text editor.

[1140] Input: Displayed text

[1141] Output: Edited draft

[1142] Step 10: Submit the final draft

[1143] When the user has finished editing, he / she presses the send button to confirm the final draft, which causes the device to resend the confirmed draft to the server and finally send the email.

[1144] Input: Edited text, click the send button

[1145] Output: Send the finalized draft and email it to the destination

[1146] (Application example 1)

[1147] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1148] Creating replies in conventional electronic communications requires a significant amount of time and effort from the user, making it difficult to respond quickly and accurately, especially during busy periods or when a large number of inquiries are received. Furthermore, it is difficult to maintain consistency in replies for each requester based on the user's writing style and past communication data. Furthermore, with the demand for rapid responses to inquiries, such as after-sales service, there is a lack of technology to efficiently generate response drafts.

[1149] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1150] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, and means for using the preprocessed data to train an artificial intelligence model that learns each user's writing style, thereby enabling means for automatically analyzing the content of the inquiry, means for generating a reply draft based on the content of the inquiry, means for displaying the generated reply draft to the user in an editable form, and means for automatically sending the finalized reply.

[1151] "Past communication data" refers to data on electronic communications that a user has conducted in the past.

[1152] "Cleansing" is a process of removing unnecessary information from collected data to make the data clean.

[1153] An "artificial intelligence model" is an algorithm that uses machine learning techniques to learn a user's writing style and specific response patterns.

[1154] The "draft creation button" is an interface that allows the user to automatically generate a reply draft.

[1155] A "user terminal" is an electronic device operated by a user that has the means to communicate with a server and send and receive data.

[1156] A "draft response" is a draft reply that is automatically generated based on the content of the received communication.

[1157] "Automatic analysis of inquiry content" is a technology that understands the content of received inquiries and identifies appropriate categories and subjects.

[1158] "Reply draft generation" is the process of automatically creating an appropriate reply based on the analyzed inquiry content.

[1159] "Editable Draft Display" is a function that presents the generated reply draft to the user, allowing the user to modify the text as needed.

[1160] "Auto-send" is a function that allows the system to automatically send a confirmed reply.

[1161] The present invention relates to a system for automatically generating reply drafts in electronic communications, which functions around three main entities: a server, a terminal, and a user.

[1162] 1. Configuration

[1163] server

[1164] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the generative AI model and generating text suggestions. Specifically, it implements the following functions:

[1165] Data collection function: The server collects past communication data from the company's internal communication systems.

[1166] Data preprocessing function: The server cleanses the collected data and removes unnecessary information.

[1167] Model training function: The server uses the cleansed data to train a generative AI model to learn each user's writing style.

[1168] Message generation function: The server analyzes the received communication content and generates a reply message based on the generative AI model.

[1169] Automatic analysis of inquiry content: The server automatically analyzes the inquiry content and identifies the appropriate category and subject.

[1170] Terminal

[1171] The terminal is a device operated by the user, which communicates with the server to send and receive data. The terminal has the following functions:

[1172] Interface provision function: The terminal displays a response draft creation button, which the user can press to start the draft generation process.

[1173] Data transmission function: The terminal transmits the received communication content to the server.

[1174] Draft display function: The terminal displays the draft sent from the server.

[1175] Draft editing function: The terminal provides an interface for users to modify and edit the generated draft.

[1176] Send function: Sends the final draft of the text that the user has confirmed.

[1177] User

[1178] A user is a person who operates the system and uses it according to a specific procedure, specifically by performing the following steps:

[1179] Confirming the inquiry: The user confirms the received inquiry email.

[1180] Operation of the Create Copy Button: When a user wants to reply, he or she presses the Create Copy button.

[1181] Review and edit the draft: The user reviews the automatically generated draft and corrects it if necessary.

[1182] Sending email: The user sends the corrected and edited reply.

[1183] 2. Hardware and Software Used

[1184] Hardware:

[1185] Server machine: A machine with a high-performance CPU and GPU

[1186] Device: Devices such as smartphones and PCs

[1187] software:

[1188] TensorFlow: Used for training and inferencing generative AI models

[1189] NLTK: Used to preprocess data for natural language processing

[1190] sklearn:Used for splitting data function

[1191] 3. Data processing and data calculation

[1192] The server preprocesses past communication data collected by the server and removes unnecessary information. Using this cleansed data, the server trains an AI model that learns each user's writing style. The generated AI model analyzes the content of received communications and automatically generates reply drafts. It also automatically analyzes the content of inquiries and provides appropriate reply drafts.

[1193] 4. Specific Examples

[1194] As an example of actual system operation, the following prompt statements can be used:

[1195] Inquiries regarding this purchase

[1196] "The product never arrived"

[1197] Please tell me the return procedure.

[1198] In this way, the present invention significantly reduces the user's workload and enables quick and accurate generation and transmission of response emails.

[1199] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1200] Step 1:

[1201] The server collects past communication data. Specifically, it collects email data from the company's internal communication system for the past year and stores it in a database. At this stage, the input is internal email data, and the output is the collected, unprocessed data.

[1202] Step 2:

[1203] The server preprocesses and cleanses the collected data. Specifically, it uses the natural language processing library NLTK to remove unnecessary words (stop words) and personal information from the collected data. At this stage, the collected raw data is the input, and the cleansed data is the output.

[1204] Step 3:

[1205] The server uses the preprocessed data to train a generative AI model. Specifically, it uses TensorFlow to segment the data and an LSTM model to learn each user's writing style. At this stage, the input is the cleansed data and the output is a trained AI model.

[1206] Step 4:

[1207] When a user receives an inquiry email, the terminal displays the received email and the user presses the button to create a response draft. The received email content is the input, and the content is sent to the server as the output.

[1208] Step 5:

[1209] The server analyzes the content of the received email and automatically generates a reply. Specifically, it analyzes the inquiry content and generates an appropriate reply using a trained AI model. At this stage, the received email content is the input, and the generated reply is the output.

[1210] Step 6:

[1211] The terminal displays the generated reply draft to the user. The user can then review this draft and make corrections or edits as necessary. For example, in response to an inquiry such as "The arrival of the product is delayed," the following reply is generated: "We apologize for any inconvenience caused by your order. We are currently checking the delivery status. Could you please wait a moment?"

[1212] Step 7:

[1213] When the user finally confirms the draft, the terminal sends this confirmed draft to the server, which then automatically sends the confirmed draft as a reply to the original query. At this stage, the confirmed draft is the input, and a reply email is sent to the user as the output.

[1214] The above steps significantly reduce the user's workload and enable quick and accurate generation and sending of response emails.

[1215] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1216] The present invention relates to a system for automatically generating reply proposals for electronic communications, in particular incorporating an emotion engine. The system functions around three parties: a server, a terminal, and a user.

[1217] 1. Configuration

[1218] server

[1219] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the AI ​​model and generating sentences. Specifically, the server implements the following functions:

[1220] Data collection function: Collects past communication data from the company's internal email system.

[1221] Data preprocessing function: Cleanses collected data and removes unnecessary information.

[1222] Model training function: The cleansed data is used to train an artificial intelligence model that learns each user's writing style.

[1223] Text generation function: Analyzes the content of received communications and generates text suggestions based on artificial intelligence models.

[1224] Emotion recognition: Recognizes user emotions using an emotion engine and reflects them in the generated text.

[1225] Terminal

[1226] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[1227] Interface provision function: Displays a response draft creation button and emotion recognition results. When the user presses this button, the draft generation process begins.

[1228] Data transmission function: Sends the contents of received emails to the server.

[1229] Text display function: Displays the text sent from the server and emotion recognition results.

[1230] Draft editing function: Provides an interface for users to modify and edit the generated draft.

[1231] Sending function: The confirmed email draft is sent via the server.

[1232] User

[1233] A user is a person who operates a system and uses it according to a specific procedure, specifically by performing the following steps:

[1234] Check email: Check the email you received.

[1235] How to use the Create a Draft button: Press the Create a Draft button when replying.

[1236] Review and edit the text and emotion recognition results: Review the automatically generated text and emotion recognition results and make corrections as necessary.

[1237] Send Email: Send the revised / edited email draft.

[1238] 2. Program Processing

[1239] Server Processing

[1240] The server first collects past communication data from the company's email system. The collected data is pre-processed and cleansed, and unnecessary information is removed. An artificial intelligence model is then trained to learn each user's writing style and perspective. Once trained, the model is used to generate text suggestions. It also uses an emotion engine to recognize user emotions and reflect them in the generated text.

[1241] Terminal handling

[1242] When the user presses the "Create a message" button, the device extracts the contents of the open email and sends it to the server. When the generated message and emotion recognition results are returned from the server, the device displays them to the user. After the user confirms, edits, and confirms the message and emotion recognition results, the device finally sends the message.

[1243] User operations

[1244] The user opens the received email and, if a reply is necessary, presses the "Create a message" button. The user checks the displayed message and emotion recognition results, and edits the content as necessary. The email is then sent with the finalized message.

[1245] 3. Specific Examples

[1246] Data collection and preprocessing

[1247] The server collects email data from the company's internal email server for the past year.

[1248] The server cleanses the collected data, removing unnecessary and personal information.

[1249] Training the model

[1250] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[1251] emotion recognition

[1252] The server uses an emotion engine to recognize the user's emotions, and this emotion data is reflected in the copy generation.

[1253] Copy generation and sending

[1254] The user opens the received email and presses the draft button.

[1255] The terminal extracts the mail contents and transmits them to the server.

[1256] The server generates a draft based on the received content and the emotion recognition results and sends it back to the terminal.

[1257] The terminal displays the generated draft and emotion recognition results to the user.

[1258] The user checks and edits the draft, and once it is confirmed, presses the send button.

[1259] The terminal transmits the confirmed email draft to the server, and finally transmits the email to the destination.

[1260] This system significantly reduces the user's workload and enables the system to generate and send response emails quickly and accurately. Furthermore, the introduction of an emotion engine allows the system to provide appropriate message suggestions based on the user's emotions, enabling more human-like communication.

[1261] The processing flow will be explained below.

[1262] Step 1:

[1263] The server collects past communication data from the company's internal email system, primarily email data from the past year, and collects it separately for each user.

[1264] Step 2:

[1265] The server stores the collected email data in a database, along with metadata such as the sender / receiver, date and time, and subject of the email.

[1266] Step 3:

[1267] The server cleanses the stored data, filtering out spam and removing unnecessary and personal information, and uses natural language processing technology to format the data.

[1268] Step 4:

[1269] The server categorizes and tags the cleansed data, assigning tags to each email based on its content (e.g., formal, casual, urgent, etc.) to make the data easier to organize and analyze.

[1270] Step 5:

[1271] The server uses the cleansed and tagged data to train an artificial intelligence model, which uses natural language processing techniques to learn each user's writing style and vocabulary, enabling personalized responses.

[1272] Step 6:

[1273] The server trains an emotion engine to recognize user emotions by analyzing keystrokes, mouse movements, and voice inputs, and incorporates this emotion information into the copy generation model.

[1274] Step 7:

[1275] When the user opens the email client, the device displays a message creation button. When the user selects the email to reply to and presses the message creation button, the device proceeds to the next step.

[1276] Step 8:

[1277] When the user presses the "Create a copy" button, the device extracts the contents of the open email, and this extracted data is used as key information for generating the copy.

[1278] Step 9:

[1279] The terminal transmits the extracted email content and emotion data generated by the emotion engine to the server. The transmitted data includes the full text of the received email, metadata, and emotion information.

[1280] Step 10:

[1281] The server analyzes the content and emotional information of the received email and generates a draft message using an AI model based on the corresponding user's writing style, taking into account the user's emotional state.

[1282] Step 11:

[1283] The server then sends the generated message back to the device, which includes the necessary information for the email reply and the recommended text from the emotion engine.

[1284] Step 12:

[1285] The device displays the received draft to the user, who can then check the content and make corrections or edits as necessary.

[1286] Step 13:

[1287] The user checks and edits the draft and finally confirms it. After confirming it, the user presses the submit button to proceed to the next step.

[1288] Step 14:

[1289] The terminal sends the finalized email draft to the server. The sent data includes the final email content.

[1290] Step 15:

[1291] The server sends the email to the specified recipient based on the final draft received. After sending, it is recorded in the database as a sending history.

[1292] Example 2

[1293] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1294] Modern business communication requires fast and appropriate responses to large volumes of emails. However, for users who receive emails frequently, coming up with an appropriate response for each email can be a significant burden. It is also important for email replies to reflect the user's emotions, which requires the user to incorporate their own feelings into the text. Furthermore, it is even more difficult to reply in a consistent style using past communication data. To address these challenges, a system is needed that automatically generates emotionally relevant response drafts and allows users to easily edit and send them.

[1295] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1296] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, means for training an artificial intelligence model that learns each user's writing style using the preprocessed data, means for reflecting emotion data obtained from an emotion recognition engine in message draft generation, and means for delivering the generated response drafts and emotion recognition results to the user terminal. This allows for the automatic generation of quick and appropriate email responses, and the user only needs to edit them, significantly reducing the burden on the user. Furthermore, by reflecting the emotion recognition results, more human-like communication can be achieved.

[1297] "Past communication data" refers to records of communication such as emails sent and received in the past, and includes the content of exchanges between users.

[1298] "Preprocessing" refers to the process of removing unnecessary information from collected data and preparing the data in a format that is easy to analyze.

[1299] "Cleansing" refers to the process of removing noise and unnecessary parts from data and extracting only the necessary information.

[1300] "Artificial intelligence model" refers to a program or algorithm that learns a user's writing style or specific patterns through machine learning or deep learning.

[1301] "Response drafting button" refers to a button that a user operates on the interface to cause the system to initiate automatic response drafting.

[1302] A "user terminal" is a device operated by a user, and refers to computer equipment such as a personal computer or smartphone.

[1303] An "emotion recognition engine" refers to an algorithm or software that analyzes and classifies user emotions from text data.

[1304] "Emotion data" refers to user emotion information extracted by an emotion recognition engine, and includes emotion categories such as "joy," "sadness," and "anger," for example.

[1305] "Message generation" refers to the process of using an artificial intelligence model to automatically create a message response based on the received content and sentiment data.

[1306] "Delivery" refers to sending the generated text and emotion recognition results to a user terminal via a network.

[1307] "Communication" refers to the act of sending and receiving e-mails, etc. over a network.

[1308] An embodiment of this invention provides a system that collects and cleans past communication data, trains an artificial intelligence model based on that data, generates drafts using an emotion recognition engine, and allows users to review, edit, and send them. The components of this system and the specific processing flow are described below.

[1309] Server Roles

[1310] The server plays a central role in the system, collecting data, preprocessing, training models, generating text, and recognizing emotions using the following hardware and software:

[1311] Hardware: A powerful server computer (e.g., a server with dual Xeon processors)

[1312] Software: Databases (e.g., MongoDB), generative AI models (e.g., GPT-3), emotion recognition engines (e.g., Microsoft Azure Cognitive Services)

[1313] Data collection and preprocessing

[1314] The server collects email data from the company's internal email system for the past year. This is achieved by using the email server's API to retrieve emails and store them in a local database. The collected data then goes through a cleansing process to remove unnecessary parts (such as email headers and signatures) and extract only the body of the email.

[1315] Training the model

[1316] The server uses the cleansed email data to train a generative AI model that learns the writing style and tone of each user's emails using large datasets and advanced machine learning algorithms.

[1317] emotion recognition

[1318] The server uses an emotion recognition engine to analyze the user's emotions. Specifically, it uses emotional data extracted from the content of the email (e.g., "joy," "sadness," "anger," etc.) to generate more appropriate text.

[1319] Copy generation and sending

[1320] When a user checks an incoming email and presses the message creation button if a reply is necessary, the device extracts the email content and sends it to the server. The server generates a message based on the received email content and the emotion recognition results. The generated message and emotion recognition results are sent back to the device and displayed to the user.

[1321] The user can check and edit the displayed message, correct it as appropriate, and then press the send button to send the final message. At this time, the device sends the message via the server.

[1322] Specific examples

[1323] For example, you can use prompts like the following to instruct a generative AI model to generate specific sentences:

[1324] "Analyze the content of the received email and generate a polite reply that includes an apology. Based on the results of the emotion engine, create a reply that takes into consideration the user's emotions."

[1325] Based on this specific prompt, the server can automatically generate appropriate reply ideas using a generative AI model, significantly reducing the user's effort. Furthermore, by incorporating emotion recognition, more human-like communication can be achieved.

[1326] In this way, the present invention helps users to respond to emails quickly and accurately, and improves the efficiency of email reply work.

[1327] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1328] Step 1: Collect data

[1329] The server collects email data from the company's internal email system for the past year. Specifically, it uses the email server's API to retrieve emails from the past year. The retrieved email data is stored in a local database (e.g., MongoDB) in JSON format.

[1330] Input: Email data from the internal email system (period: past year)

[1331] Output: Email data stored in a local database (JSON format)

[1332] Step 2: Preprocessing the data

[1333] The server cleanses the collected email data. Specifically, it uses regular expressions to remove unnecessary information such as email headers and signatures, and extracts only the body of the email. The preprocessed data is then saved in JSON format.

[1334] Input: Collected email data (JSON format)

[1335] Output: Cleansed email data (JSON format)

[1336] Step 3: Train the model

[1337] The server trains a generative AI model (e.g., GPT-3) using the cleansed email data. Specifically, it splits the data into batches and trains the model over multiple epochs. The trained model learns the writing style and tone of each user.

[1338] Input: Cleansed email data (JSON format)

[1339] Output: A trained generative AI model

[1340] Step 4: Process incoming email

[1341] The user checks the received email and presses the message creation button if a reply is necessary. The terminal extracts the contents of the opened email and sends them to the server.

[1342] Input: Received email content (text format)

[1343] Output: Email content sent to the server (text format)

[1344] Step 5: Recognize emotions

[1345] The server sends the received email content to an emotion recognition engine to analyze the user's emotions. The recognized emotion data is classified into categories such as "joy," "sadness," and "anger," and is used in the copy generation process.

[1346] Input: Received email content (text format)

[1347] Output: Recognized emotion data (categorical format)

[1348] Step 6: Generate copy

[1349] The server provides a prompt to the generative AI model based on the content of the received email and the emotion data, and generates a draft. Example prompt: "Analyze the content of the received email and generate a polite reply that includes an apology. Based on the results of the emotion engine, please create a draft that takes the user's emotions into consideration."

[1350] Input: received email content (text format), recognized emotion data (category format), prompt text

[1351] Output: Generated draft (text format)

[1352] Step 7: Distributing the copy

[1353] The server delivers the generated text and emotion recognition results to the terminal, which then displays them to the user.

[1354] Input: Generated text (text format), emotion recognition results (category format)

[1355] Output: Text and emotion recognition results delivered to the device (text format)

[1356] Step 8: Edit and submit your copy

[1357] The user checks the draft displayed on the device and edits it as necessary. After that, the user presses the send button, and the edited and confirmed draft is sent from the device via the server. It is finally sent as an email.

[1358] Input: Generated draft (text format)

[1359] Output: Final email draft sent (email format)

[1360] (Application example 2)

[1361] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1362] While automated systems for generating reply proposals in electronic communications can significantly reduce time and effort, conventional systems often do not take emotion recognition into account, which can result in inappropriate replies. Furthermore, while there is a growing need for fast and appropriate replies to user comments in content distribution services, current methods suffer from slow response times. The present invention aims to solve these problems by generating fast and appropriate replies while taking user emotions into account.

[1363] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1364] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, and means for training an artificial intelligence model that learns each user's writing style using the preprocessed data. This makes it possible to generate an optimal writing style for each user based on past data. Also, in this invention, the server includes means for generating a response draft based on received content and emotion recognition results, and means for automatically analyzing comments and generating reply drafts in a content distribution service. This makes it possible to take user emotions into consideration and generate appropriate and prompt reply drafts in a content distribution service.

[1365] "Historical communications data" is a collection of previous emails, messages, and other communications records.

[1366] "Means of collection" refers to the functions and processes for importing past communication data into the server.

[1367] "Preprocessing and cleansing means" refers to functions and processes for removing unnecessary information from collected data and preparing the data.

[1368] An "artificial intelligence model" is a computer program that uses machine learning techniques to analyze data and is trained to perform specific tasks.

[1369] A "training method" is a function or process that uses historical communication data to teach an artificial intelligence model a particular writing style or pattern.

[1370] The "response draft generation button" is an interface element that the user can operate to start the process of automatically generating a reply draft.

[1371] A "user terminal" is a device operated by a user for sending e-mails and other communications.

[1372] "Received content" refers to the content of messages or emails received on the user's terminal.

[1373] "Emotion recognition results" are information obtained by inferring and analyzing a user's emotions from the content of comments and messages.

[1374] "Comment analysis" is the process of analyzing received comments to understand their content and generate appropriate responses.

[1375] A "means for automatically generating reply drafts" is a function or process for analyzing a comment or message and automatically generating a reply to that comment or message.

[1376] "Delivery means" refers to the function or process for transmitting the generated reply draft to the user terminal.

[1377] "Edit and confirm" is the process by which a user corrects an automatically generated draft and confirms that it should be sent.

[1378] The "means of transmitting communication" is the function or process for transmitting the revised or finalized text.

[1379] MODE FOR CARRYING OUT THE INVENTION

[1380] The present invention is a system for automatically generating reply proposals to comments in a content distribution service. Here, the specific method and operation of the system will be described in detail.

[1381] Server Processing

[1382] The server collects and preprocesses past communication data. During the data collection process, past comment data is retrieved through an API endpoint and saved in JSON format. The collected data is cleansed to remove unnecessary information. The preprocessed data is then used to train a generative AI model that learns each user's writing style. The artificial intelligence technology used here uses OpenAI's API.

[1383] Emotion Recognition and Sentence Generation

[1384] When the server receives a new comment from a user, it analyzes the content and performs emotion recognition. This emotion recognition is performed based on the text extracted from the user's comment, using OpenAI's emotion analysis model. The analysis results are reflected in the generation of reply proposals. An appropriate reply proposal is generated based on the received content and the emotion recognition results. In this process, a prompt sentence is used to instruct the generative AI model to generate a proposal.

[1385] Terminal handling

[1386] The device communicates with the server in response to user operations. When a user receives a comment and replies to it, the device sends the comment content to the server. When the server returns a reply draft and emotion recognition results, the device displays them to the user. The user can check the displayed draft and edit it as necessary. The edited and confirmed draft is sent from the device to the server, and a reply to the comment is finally sent.

[1387] User operations

[1388] When a user receives a comment on a content distribution service and needs to reply to that comment, they press the response draft button. This operation sends the comment content from the device to the server. The server generates a reply draft based on the comment content and emotion recognition results and returns it to the user's device. The user can then check the draft and edit it as necessary. Finally, by pressing the send button, the edited draft is finalized and sent back to the comment poster.

[1389] Specific examples

[1390] For example, if a user receives a comment such as "This news was really helpful!", the device sends the comment to the server, which then uses the following prompt to instruct the generative AI model to generate a sentence:

[1391] Example prompt sentence:

[1392] A user says 'This news was really helpful!'. You could reply with (emotion: joy):

[1393] The generated message, "Thank you for your feedback! We look forward to working with you in the future," is displayed on the user's device. The user can review this message, revise it as necessary, and then send it. This allows for an appropriate and prompt reply that takes the user's feelings into consideration.

[1394] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1395] Step 1:

[1396] The server collects past communication data.

[1397] Input: Historical comment data retrieved from an API endpoint.

[1398] How it works: The server collects data through the API and stores it in JSON format.

[1399] Output: Archived historical comment data.

[1400] Step 2:

[1401] The server pre-processes and cleanses the collected communication data.

[1402] Input: Collected historical comment data.

[1403] How it works: The server strips out unnecessary information from the data, leaving only the necessary fields. Specifically, it extracts the comment content and the identity of the commenter.

[1404] Output: Cleansed communication data.

[1405] Step 3:

[1406] The server uses the preprocessed data to train a generative AI model.

[1407] Input: Cleansed communication data.

[1408] How it works: The server uses the preprocessed data to train a generative AI model that learns each user's writing style. It uses OpenAI's API to generate and train the AI ​​model.

[1409] Output: A fully trained generative AI model.

[1410] Step 4:

[1411] The user receives the comment and presses the response draft button.

[1412] Input: The comment content received by the user.

[1413] How it works: The user presses the button on their device to create a response draft and send it to the server.

[1414] Output: The comment sent.

[1415] Step 5:

[1416] The server analyzes the received content and recognizes the emotion.

[1417] Input: The comment submitted.

[1418] How it works: The server uses OpenAI's sentiment analysis model to analyze the content of comments and recognize their sentiment.

[1419] Output: Recognized emotion data.

[1420] Step 6:

[1421] The server generates a draft response based on the received content and the emotion recognition results.

[1422] Input: Received comment content and recognized emotion data.

[1423] How it works: The server uses the generative AI model to generate a prompt, sends it to the AI ​​model, and generates an appropriate reply.

[1424] Output: The generated response draft.

[1425] Step 7:

[1426] The server delivers the generated response draft to the user terminal.

[1427] Input: The generated response draft.

[1428] Operation: The server sends the generated draft to the user terminal.

[1429] Output: A draft response displayed on the user's terminal.

[1430] Step 8:

[1431] The user reviews the draft response and edits it if necessary.

[1432] Input: Proposed response displayed on the terminal.

[1433] Action: The user reviews the draft response and makes any necessary corrections.

[1434] Output: Edited draft response.

[1435] Step 9:

[1436] The user presses the send button to send the reply draft.

[1437] Input: Edited draft response.

[1438] Action: The user presses the send button and the confirmed response draft is sent via the server.

[1439] Output: Final draft response sent.

[1440] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1441] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1442] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1443] [Fourth embodiment]

[1444] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1445] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1446] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1447] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1448] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1449] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1450] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1451] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1452] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1453] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1454] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1455] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1456] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1457] The present invention relates to a system for automatically generating reply drafts for electronic communications, which system functions around three central parties: a server, a terminal, and a user.

[1458] 1. Configuration

[1459] server

[1460] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the AI ​​model and generating sentences. Specifically, the server implements the following functions:

[1461] Data collection function: Collects past communication data from the company's internal email system.

[1462] Data preprocessing function: Cleanses collected data and removes unnecessary information.

[1463] Model training function: The cleansed data is used to train an artificial intelligence model that learns each user's writing style.

[1464] Text generation function: Analyzes the content of received communications and generates text suggestions based on artificial intelligence models.

[1465] Terminal

[1466] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[1467] Interface provision function: Displays a response draft creation button. When the user presses this button, the draft generation process begins.

[1468] Data transmission function: Sends the contents of received emails to the server.

[1469] Draft display function: Displays the draft sent from the server.

[1470] Draft editing function: Provides an interface for users to modify and edit the generated draft.

[1471] Sending function: The confirmed email draft is sent via the server.

[1472] User

[1473] A user is a person who operates a system and uses it according to a specific procedure, specifically by performing the following steps:

[1474] Check email: Check the email you received.

[1475] How to use the Create a Draft button: Press the Create a Draft button when replying.

[1476] Review and edit the transcript: Review the auto-generated transcript and make any necessary corrections.

[1477] Send Email: Send the revised / edited email draft.

[1478] 2. Program Processing

[1479] Server Processing

[1480] The server first collects past communication data from the company's email system. The collected data is pre-processed and cleansed to remove unnecessary information. An artificial intelligence model is then trained to learn each user's writing style and perspective. Once trained, the model is used to generate copy.

[1481] Terminal handling

[1482] When the user presses the "Create a draft" button, the device extracts the contents of the open email and sends it to the server. When the generated draft is returned from the server, the device displays it to the user. After the user confirms, edits, and confirms the draft, the device finally sends it.

[1483] User operations

[1484] The user opens the received email and, if a reply is necessary, presses the Create Message button. The user checks the displayed message and edits the content as necessary. The email is then sent with the final message.

[1485] 3. Specific Examples

[1486] Data collection and preprocessing

[1487] The server collects email data from the company's internal email server for the past year.

[1488] The server cleanses the collected data, removing unnecessary and personal information.

[1489] Training the model

[1490] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[1491] Copy generation and sending

[1492] The user opens the received email and presses the draft button.

[1493] The terminal extracts the mail contents and transmits them to the server.

[1494] The server generates a draft based on the received content and returns it to the terminal.

[1495] The terminal displays the generated draft to the user.

[1496] The user checks and edits the draft, and once it is confirmed, presses the send button.

[1497] The terminal transmits the confirmed email draft to the server, and finally transmits the email to the destination.

[1498] This system significantly reduces the user's workload and enables quick and accurate generation and sending of response emails.

[1499] The processing flow will be explained below.

[1500] Step 1:

[1501] The server collects past communication data from the company's email system, primarily email data from the past year, and collects it separately for each user.

[1502] Step 2:

[1503] The server stores the collected email data in a database, along with metadata such as the sender / receiver, date and time, and subject of the email.

[1504] Step 3:

[1505] The server performs spam filtering and personal information cleansing of the stored data, removing unnecessary and privacy-sensitive information.

[1506] Step 4:

[1507] The server then categorizes and tags the cleansed data, assigning tags to each email based on its content (e.g., formal, casual, urgent, etc.) to make the data easier to organize and analyze.

[1508] Step 5:

[1509] The server uses the cleansed and tagged data to train an artificial intelligence model, which uses natural language processing techniques to learn each user's writing style and vocabulary.

[1510] Step 6:

[1511] The terminal displays a message drafting button when the user opens the email client. When the user selects an email to reply to and presses the message drafting button, the terminal proceeds to the next step.

[1512] Step 7:

[1513] When the user presses the "Create a draft" button, the device extracts the contents of the open email, and this extracted data is used as key information for generating the draft.

[1514] Step 8:

[1515] The terminal sends the extracted email content to the server. The sent data includes the full text of the received email and metadata.

[1516] Step 9:

[1517] The server analyzes the content of the received email and generates drafts using an artificial intelligence model based on the corresponding user's writing style.

[1518] Step 10:

[1519] The server sends the generated message back to the terminal, which contains the necessary information for the reply message.

[1520] Step 11:

[1521] The terminal displays the received draft to the user, who can then check the content and make corrections or edits as necessary.

[1522] Step 12:

[1523] The user checks and edits the draft and finally confirms it. After confirming, they press the send button to proceed to the next step.

[1524] Step 13:

[1525] The terminal sends the finalized email draft to the server. The sent data includes the final email content.

[1526] Step 14:

[1527] The server sends the email to the specified recipient based on the final draft it received. After sending, it is saved in the database as a sending history.

[1528] By following these steps, users can easily and efficiently create and send correspondence drafts.

[1529] Example 1

[1530] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1531] In conventional electronic communication, creating replies is often done manually, resulting in a significant workload. Furthermore, some users have difficulty maintaining a consistent writing style and formality, resulting in inconsistencies in the quality of corporate communications. In these circumstances, there is a need for efficient and consistent replies. Furthermore, there is a need for a means to effectively utilize past communication data and automatically generate replies that reflect a user's unique writing style. To address this issue, the present invention provides a system for automatically generating efficient and consistent reply drafts.

[1532] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1533] In this invention, the server includes means for collecting historical communication data, means for pre-processing and cleansing the historical communication data, and means for using the pre-processed data to train a generative AI model that learns each user's writing style, thereby enabling efficient and consistent automated generation of reply drafts.

[1534] "Data collection" is the process of obtaining historical communication data from a particular system or server.

[1535] "Cleansing" is the process of removing unnecessary or incorrect information from collected data and arranging it into an appropriate format.

[1536] A "generative AI model" is an artificial intelligence model that learns the unique characteristics of users, such as their writing style and position, and automatically generates optimal reply drafts.

[1537] The "response drafting button" is a terminal interface element that the user operates to initiate the email reply drafting process.

[1538] "Reply generation" is the process of using a generative AI model to create appropriate reply suggestions based on collected and pre-processed data.

[1539] "Data transmission" is the process of sending and receiving extracted email content and generated drafts between the server and the terminal.

[1540] "Draft editing" is the process in which the user reviews the generated draft and makes corrections or edits as necessary.

[1541] "Communication sending" is the process of sending the finalized email draft to the intended recipient.

[1542] The present invention relates to a system for automatically generating reply drafts for electronic communications, which system functions around three main entities: a server, a terminal, and a user.

[1543] Server configuration and functions

[1544] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the generative AI model and generating sentences. Specifically, it implements the following functions:

[1545] Data collection function: The server collects historical email data from the company's mail system. For example, it periodically retrieves data for the past year using a Linux cron job or a Python script.

[1546] Data preprocessing function: The server uses a data cleansing tool (e.g., the Pandas library) to remove unnecessary information and personal information from the collected data.

[1547] Model training function: The server uses the cleansed data to train a generative AI model (e.g., GPT-3 or BERT model). This model learns each user's writing style and perspective and is used to automatically generate email reply suggestions.

[1548] Text generation function: The server generates an appropriate text based on the content of the received email and returns it to the terminal.

[1549] Device configuration and functions

[1550] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[1551] Interface provision function: The terminal displays a response draft creation button. When the user presses this button, the draft generation process begins.

[1552] Email content extraction function: When the user presses a button, the device extracts the contents of the open email and sends them to the server.

[1553] Draft reception function: Receives drafts generated from the server.

[1554] Text display function: The terminal displays the received text to the user, for example, using a GUI framework (JavaFX or Electron).

[1555] Copy editing function: Provides an interface (such as a rich text editor) for users to review and edit the copy.

[1556] Final sending function: The user resends the finalized draft to the server and finally sends the email.

[1557] User operation procedure

[1558] The user uses the system by following these steps:

[1559] 1. Check your email: Open your received email.

[1560] 2. How to use the Create a Draft button: Press the Create a Draft button when replying.

[1561] 3. Review and edit the transcript: Review the auto-generated transcript and make any necessary corrections.

[1562] 4. Send the email: Send the revised and edited email.

[1563] Specific examples

[1564] Data collection and preprocessing

[1565] The server collects email data from the company's internal email server for the past year.

[1566] The server cleanses the collected data, removing unnecessary and personal information.

[1567] Training the model

[1568] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[1569] Copy generation and sending

[1570] The user opens the received email and presses the draft button.

[1571] The terminal extracts the mail contents and transmits them to the server.

[1572] The server generates a draft based on the received content and returns it to the terminal.

[1573] The terminal displays the generated draft to the user.

[1574] The user checks and edits the draft, and once it is confirmed, presses the send button.

[1575] The terminal transmits the finalized email draft to the server, and finally transmits the email to the destination.

[1576] Prompt Sentence Examples

[1577] "I received an email inquiring about the status of a recent project. Please draft a response."

[1578] This system significantly reduces the user's workload and enables quick and accurate generation and sending of response emails.

[1579] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1580] Step 1: Collect historical communication data

[1581] The server collects email data from your company's mail server for the past year, using a Python script and the mail server's IMAP protocol to retrieve all emails for a specified period.

[1582] Input: Email server account information, period (e.g., past year)

[1583] Output: Captured email data (subject, body, sender / receiver information, etc.)

[1584] Step 2: Preprocess the data

[1585] The server cleanses the retrieved email data using the Pandas library, specifically by removing unnecessary header information and personal information and converting the data into an appropriate format.

[1586] Input: Captured email data

[1587] Output: Cleansed data (well-formatted data with unnecessary information removed)

[1588] Step 3: Training the generative AI model

[1589] The server uses the cleansed data to train a generative AI model (e.g., GPT-3 or BERT), applying machine learning algorithms to learn each user's writing style and vocabulary.

[1590] Input: Cleansed data

[1591] Output: A trained generative AI model

[1592] Step 4: Begin the response drafting process

[1593] The user presses the drafting button displayed on the terminal's email client, which starts the drafting process.

[1594] Input: User button presses

[1595] Output: Signal to start the copywriting process

[1596] Step 5: Extract the contents of the received email

[1597] The terminal extracts the content of the received email that the user currently has open, specifically, the subject, body, and sender information of the email.

[1598] Input: Incoming emails that the user has opened

[1599] Output: Extracted email content (subject, body, sender information)

[1600] Step 6: Send the email content to the server

[1601] The terminal sends the extracted email content to the server as an HTTP POST request.

[1602] Input: Extracted email content

[1603] Output: Request sent to server

[1604] Step 7: Server-generated copy

[1605] The server creates a prompt based on the content of the received email and generates a draft using a generative AI model. Specifically, it analyzes the content of the email, generates a prompt for the generative AI model, and then creates a reply draft based on that.

[1606] Input: Email content (subject, body, sender information)

[1607] Output: Generated reply draft

[1608] Step 8: Receiving and viewing the generated draft

[1609] The terminal displays the generated draft received from the server to the user, specifically, using a GUI framework to display the draft in an easy-to-read format.

[1610] Input: Generated reply

[1611] Output: The text displayed on the screen

[1612] Step 9: Review and edit your copy

[1613] The user checks the displayed draft and corrects or edits it as necessary. Specifically, the user edits the content using a rich text editor.

[1614] Input: Displayed text

[1615] Output: Edited draft

[1616] Step 10: Submit the final draft

[1617] When the user has finished editing, he / she presses the send button to confirm the final draft, which causes the device to resend the confirmed draft to the server and finally send the email.

[1618] Input: Edited text, click the send button

[1619] Output: Send the finalized draft and email it to the destination

[1620] (Application example 1)

[1621] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1622] Creating replies in conventional electronic communications requires a significant amount of time and effort from the user, making it difficult to respond quickly and accurately, especially during busy periods or when a large number of inquiries are received. Furthermore, it is difficult to maintain consistency in replies for each requester based on the user's writing style and past communication data. Furthermore, with the demand for rapid responses to inquiries, such as after-sales service, there is a lack of technology to efficiently generate response drafts.

[1623] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1624] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, and means for using the preprocessed data to train an artificial intelligence model that learns each user's writing style, thereby enabling means for automatically analyzing the content of the inquiry, means for generating a reply draft based on the content of the inquiry, means for displaying the generated reply draft to the user in an editable form, and means for automatically sending the finalized reply.

[1625] "Past communication data" refers to data on electronic communications that a user has conducted in the past.

[1626] "Cleansing" is a process of removing unnecessary information from collected data to make the data clean.

[1627] An "artificial intelligence model" is an algorithm that uses machine learning techniques to learn a user's writing style and specific response patterns.

[1628] The "draft creation button" is an interface that allows the user to automatically generate a reply draft.

[1629] A "user terminal" is an electronic device operated by a user that has the means to communicate with a server and send and receive data.

[1630] A "draft response" is a draft reply that is automatically generated based on the content of the received communication.

[1631] "Automatic analysis of inquiry content" is a technology that understands the content of received inquiries and identifies appropriate categories and subjects.

[1632] "Reply draft generation" is the process of automatically creating an appropriate reply based on the analyzed inquiry content.

[1633] "Editable Draft Display" is a function that presents the generated reply draft to the user, allowing the user to modify the text as needed.

[1634] "Auto-send" is a function that allows the system to automatically send a confirmed reply.

[1635] The present invention relates to a system for automatically generating reply drafts in electronic communications, which functions around three main entities: a server, a terminal, and a user.

[1636] 1. Configuration

[1637] server

[1638] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the generative AI model and generating text suggestions. Specifically, it implements the following functions:

[1639] Data collection function: The server collects past communication data from the company's internal communication systems.

[1640] Data preprocessing function: The server cleanses the collected data and removes unnecessary information.

[1641] Model training function: The server uses the cleansed data to train a generative AI model to learn each user's writing style.

[1642] Message generation function: The server analyzes the received communication content and generates a reply message based on the generative AI model.

[1643] Automatic analysis of inquiry content: The server automatically analyzes the inquiry content and identifies the appropriate category and subject.

[1644] Terminal

[1645] The terminal is a device operated by the user, which communicates with the server to send and receive data. The terminal has the following functions:

[1646] Interface provision function: The terminal displays a response draft creation button, which the user can press to start the draft generation process.

[1647] Data transmission function: The terminal transmits the received communication content to the server.

[1648] Draft display function: The terminal displays the draft sent from the server.

[1649] Draft editing function: The terminal provides an interface for users to modify and edit the generated draft.

[1650] Send function: Sends the final draft of the text that the user has confirmed.

[1651] User

[1652] A user is a person who operates the system and uses it according to a specific procedure, specifically by performing the following steps:

[1653] Confirming the inquiry: The user confirms the received inquiry email.

[1654] Operation of the Create Copy Button: When a user wants to reply, he or she presses the Create Copy button.

[1655] Review and edit the draft: The user reviews the automatically generated draft and corrects it if necessary.

[1656] Sending email: The user sends the corrected and edited reply.

[1657] 2. Hardware and Software Used

[1658] Hardware:

[1659] Server machine: A machine with a high-performance CPU and GPU

[1660] Device: Devices such as smartphones and PCs

[1661] software:

[1662] TensorFlow: Used for training and inferencing generative AI models

[1663] NLTK: Used to preprocess data for natural language processing

[1664] sklearn:Used for splitting data function

[1665] 3. Data processing and data calculation

[1666] The server preprocesses past communication data collected by the server and removes unnecessary information. Using this cleansed data, the server trains an AI model that learns each user's writing style. The generated AI model analyzes the content of received communications and automatically generates reply drafts. It also automatically analyzes the content of inquiries and provides appropriate reply drafts.

[1667] 4. Specific Examples

[1668] As an example of actual system operation, the following prompt statements can be used:

[1669] Inquiries regarding this purchase

[1670] "The product never arrived"

[1671] Please tell me the return procedure.

[1672] In this way, the present invention significantly reduces the user's workload and enables quick and accurate generation and transmission of response emails.

[1673] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1674] Step 1:

[1675] The server collects past communication data. Specifically, it collects email data from the company's internal communication system for the past year and stores it in a database. At this stage, the input is internal email data, and the output is the collected, unprocessed data.

[1676] Step 2:

[1677] The server preprocesses and cleanses the collected data. Specifically, it uses the natural language processing library NLTK to remove unnecessary words (stop words) and personal information from the collected data. At this stage, the collected raw data is the input, and the cleansed data is the output.

[1678] Step 3:

[1679] The server uses the preprocessed data to train a generative AI model. Specifically, it uses TensorFlow to segment the data and an LSTM model to learn each user's writing style. At this stage, the input is the cleansed data and the output is a trained AI model.

[1680] Step 4:

[1681] When a user receives an inquiry email, the terminal displays the received email and the user presses the button to create a response draft. The received email content is the input, and the content is sent to the server as the output.

[1682] Step 5:

[1683] The server analyzes the content of the received email and automatically generates a reply. Specifically, it analyzes the inquiry content and generates an appropriate reply using a trained AI model. At this stage, the received email content is the input, and the generated reply is the output.

[1684] Step 6:

[1685] The terminal displays the generated reply draft to the user. The user can then review this draft and make corrections or edits as necessary. For example, in response to an inquiry such as "The arrival of the product is delayed," the following reply is generated: "We apologize for any inconvenience caused by your order. We are currently checking the delivery status. Could you please wait a moment?"

[1686] Step 7:

[1687] When the user finally confirms the draft, the terminal sends this confirmed draft to the server, which then automatically sends the confirmed draft as a reply to the original query. At this stage, the confirmed draft is the input, and a reply email is sent to the user as the output.

[1688] The above steps significantly reduce the user's workload and enable quick and accurate generation and sending of response emails.

[1689] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1690] The present invention relates to a system for automatically generating reply proposals for electronic communications, in particular incorporating an emotion engine. The system functions around three parties: a server, a terminal, and a user.

[1691] 1. Configuration

[1692] server

[1693] The server has the function of collecting and preprocessing communication data. It is also responsible for the main part of training the AI ​​model and generating sentences. Specifically, the server implements the following functions:

[1694] Data collection function: Collects past communication data from the company's internal email system.

[1695] Data preprocessing function: Cleanses collected data and removes unnecessary information.

[1696] Model training function: The cleansed data is used to train an artificial intelligence model that learns each user's writing style.

[1697] Text generation function: Analyzes the content of received communications and generates text suggestions based on artificial intelligence models.

[1698] Emotion recognition: Recognizes user emotions using an emotion engine and reflects them in the generated text.

[1699] Terminal

[1700] A terminal is a device operated by a user that communicates with a server to send and receive data. A terminal has the following functions:

[1701] Interface provision function: Displays a response draft creation button and emotion recognition results. When the user presses this button, the draft generation process begins.

[1702] Data transmission function: Sends the contents of received emails to the server.

[1703] Text display function: Displays the text sent from the server and emotion recognition results.

[1704] Draft editing function: Provides an interface for users to modify and edit the generated draft.

[1705] Sending function: The confirmed email draft is sent via the server.

[1706] User

[1707] A user is a person who operates a system and uses it according to a specific procedure, specifically by performing the following steps:

[1708] Check email: Check the email you received.

[1709] How to use the Create a Draft button: Press the Create a Draft button when replying.

[1710] Review and edit the text and emotion recognition results: Review the automatically generated text and emotion recognition results and make corrections as necessary.

[1711] Send Email: Send the revised / edited email draft.

[1712] 2. Program Processing

[1713] Server Processing

[1714] The server first collects past communication data from the company's email system. The collected data is pre-processed and cleansed, and unnecessary information is removed. An artificial intelligence model is then trained to learn each user's writing style and perspective. Once trained, the model is used to generate text suggestions. It also uses an emotion engine to recognize user emotions and reflect them in the generated text.

[1715] Terminal handling

[1716] When the user presses the "Create a message" button, the device extracts the contents of the open email and sends it to the server. When the generated message and emotion recognition results are returned from the server, the device displays them to the user. After the user confirms, edits, and confirms the message and emotion recognition results, the device finally sends the message.

[1717] User operations

[1718] The user opens the received email and, if a reply is necessary, presses the "Create a message" button. The user checks the displayed message and emotion recognition results, and edits the content as necessary. The email is then sent with the finalized message.

[1719] 3. Specific Examples

[1720] Data collection and preprocessing

[1721] The server collects email data from the company's internal email server for the past year.

[1722] The server cleanses the collected data, removing unnecessary and personal information.

[1723] Training the model

[1724] The server uses the cleansed data to train a generative AI model, which learns the user's writing style and perspective.

[1725] emotion recognition

[1726] The server uses an emotion engine to recognize the user's emotions, and this emotion data is reflected in the copy generation.

[1727] Copy generation and sending

[1728] The user opens the received email and presses the draft button.

[1729] The terminal extracts the mail contents and transmits them to the server.

[1730] The server generates a draft based on the received content and the emotion recognition results and sends it back to the terminal.

[1731] The terminal displays the generated draft and emotion recognition results to the user.

[1732] The user checks and edits the draft, and once it is confirmed, presses the send button.

[1733] The terminal transmits the confirmed email draft to the server, and finally transmits the email to the destination.

[1734] This system significantly reduces the user's workload and enables the system to generate and send response emails quickly and accurately. Furthermore, the introduction of an emotion engine allows the system to provide appropriate message suggestions based on the user's emotions, enabling more human-like communication.

[1735] The processing flow will be explained below.

[1736] Step 1:

[1737] The server collects past communication data from the company's internal email system, primarily email data from the past year, and collects it separately for each user.

[1738] Step 2:

[1739] The server stores the collected email data in a database, along with metadata such as the sender / receiver, date and time, and subject of the email.

[1740] Step 3:

[1741] The server cleanses the stored data, filtering out spam and removing unnecessary and personal information, and uses natural language processing technology to format the data.

[1742] Step 4:

[1743] The server categorizes and tags the cleansed data, assigning tags to each email based on its content (e.g., formal, casual, urgent, etc.) to make the data easier to organize and analyze.

[1744] Step 5:

[1745] The server uses the cleansed and tagged data to train an artificial intelligence model, which uses natural language processing techniques to learn each user's writing style and vocabulary, enabling personalized responses.

[1746] Step 6:

[1747] The server trains an emotion engine to recognize user emotions by analyzing keystrokes, mouse movements, and voice inputs, and incorporates this emotion information into the copy generation model.

[1748] Step 7:

[1749] When the user opens the email client, the device displays a message creation button. When the user selects the email to reply to and presses the message creation button, the device proceeds to the next step.

[1750] Step 8:

[1751] When the user presses the "Create a copy" button, the device extracts the contents of the open email, and this extracted data is used as key information for generating the copy.

[1752] Step 9:

[1753] The terminal transmits the extracted email content and emotion data generated by the emotion engine to the server. The transmitted data includes the full text of the received email, metadata, and emotion information.

[1754] Step 10:

[1755] The server analyzes the content and emotional information of the received email and generates a draft message using an AI model based on the corresponding user's writing style, taking into account the user's emotional state.

[1756] Step 11:

[1757] The server then sends the generated message back to the device, which includes the necessary information for the email reply and the recommended text from the emotion engine.

[1758] Step 12:

[1759] The device displays the received draft to the user, who can then check the content and make corrections or edits as necessary.

[1760] Step 13:

[1761] The user checks and edits the draft and finally confirms it. After confirming it, the user presses the submit button to proceed to the next step.

[1762] Step 14:

[1763] The terminal sends the finalized email draft to the server. The sent data includes the final email content.

[1764] Step 15:

[1765] The server sends the email to the specified recipient based on the final draft received. After sending, it is recorded in the database as a sending history.

[1766] Example 2

[1767] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1768] Modern business communication requires fast and appropriate responses to large volumes of emails. However, for users who receive emails frequently, coming up with an appropriate response for each email can be a significant burden. It is also important for email replies to reflect the user's emotions, which requires the user to incorporate their own feelings into the text. Furthermore, it is even more difficult to reply in a consistent style using past communication data. To address these challenges, a system is needed that automatically generates emotionally relevant response drafts and allows users to easily edit and send them.

[1769] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1770] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, means for training an artificial intelligence model that learns each user's writing style using the preprocessed data, means for reflecting emotion data obtained from an emotion recognition engine in message draft generation, and means for delivering the generated response drafts and emotion recognition results to the user terminal. This allows for the automatic generation of quick and appropriate email responses, and the user only needs to edit them, significantly reducing the burden on the user. Furthermore, by reflecting the emotion recognition results, more human-like communication can be achieved.

[1771] "Past communication data" refers to records of communication such as emails sent and received in the past, and includes the content of exchanges between users.

[1772] "Preprocessing" refers to the process of removing unnecessary information from collected data and preparing the data in a format that is easy to analyze.

[1773] "Cleansing" refers to the process of removing noise and unnecessary parts from data and extracting only the necessary information.

[1774] "Artificial intelligence model" refers to a program or algorithm that learns a user's writing style or specific patterns through machine learning or deep learning.

[1775] "Response drafting button" refers to a button that a user operates on the interface to cause the system to initiate automatic response drafting.

[1776] A "user terminal" is a device operated by a user, and refers to computer equipment such as a personal computer or smartphone.

[1777] An "emotion recognition engine" refers to an algorithm or software that analyzes and classifies user emotions from text data.

[1778] "Emotion data" refers to user emotion information extracted by an emotion recognition engine, and includes emotion categories such as "joy," "sadness," and "anger," for example.

[1779] "Message generation" refers to the process of using an artificial intelligence model to automatically create a message response based on the received content and sentiment data.

[1780] "Delivery" refers to sending the generated text and emotion recognition results to a user terminal via a network.

[1781] "Communication" refers to the act of sending and receiving e-mails, etc. over a network.

[1782] An embodiment of this invention provides a system that collects and cleans past communication data, trains an artificial intelligence model based on that data, generates drafts using an emotion recognition engine, and allows users to review, edit, and send them. The components of this system and the specific processing flow are described below.

[1783] Server Roles

[1784] The server plays a central role in the system, collecting data, preprocessing, training models, generating text, and recognizing emotions using the following hardware and software:

[1785] Hardware: A powerful server computer (e.g., a server with dual Xeon processors)

[1786] Software: Databases (e.g., MongoDB), generative AI models (e.g., GPT-3), emotion recognition engines (e.g., Microsoft Azure Cognitive Services)

[1787] Data collection and preprocessing

[1788] The server collects email data from the company's internal email system for the past year. This is achieved by using the email server's API to retrieve emails and store them in a local database. The collected data then goes through a cleansing process to remove unnecessary parts (such as email headers and signatures) and extract only the body of the email.

[1789] Training the model

[1790] The server uses the cleansed email data to train a generative AI model that learns the writing style and tone of each user's emails using large datasets and advanced machine learning algorithms.

[1791] emotion recognition

[1792] The server uses an emotion recognition engine to analyze the user's emotions. Specifically, it uses emotional data extracted from the content of the email (e.g., "joy," "sadness," "anger," etc.) to generate more appropriate text.

[1793] Copy generation and sending

[1794] When a user checks an incoming email and presses the message creation button if a reply is necessary, the device extracts the email content and sends it to the server. The server generates a message based on the received email content and the emotion recognition results. The generated message and emotion recognition results are sent back to the device and displayed to the user.

[1795] The user can check and edit the displayed message, correct it as appropriate, and then press the send button to send the final message. At this time, the device sends the message via the server.

[1796] Specific examples

[1797] For example, you can use prompts like the following to instruct a generative AI model to generate specific sentences:

[1798] "Analyze the content of the received email and generate a polite reply that includes an apology. Based on the results of the emotion engine, create a reply that takes into consideration the user's emotions."

[1799] Based on this specific prompt, the server can automatically generate appropriate reply ideas using a generative AI model, significantly reducing the user's effort. Furthermore, by incorporating emotion recognition, more human-like communication can be achieved.

[1800] In this way, the present invention helps users to respond to emails quickly and accurately, and improves the efficiency of email reply work.

[1801] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1802] Step 1: Collect data

[1803] The server collects email data from the company's internal email system for the past year. Specifically, it uses the email server's API to retrieve emails from the past year. The retrieved email data is stored in a local database (e.g., MongoDB) in JSON format.

[1804] Input: Email data from the internal email system (period: past year)

[1805] Output: Email data stored in a local database (JSON format)

[1806] Step 2: Preprocessing the data

[1807] The server cleanses the collected email data. Specifically, it uses regular expressions to remove unnecessary information such as email headers and signatures, and extracts only the body of the email. The preprocessed data is then saved in JSON format.

[1808] Input: Collected email data (JSON format)

[1809] Output: Cleansed email data (JSON format)

[1810] Step 3: Train the model

[1811] The server trains a generative AI model (e.g., GPT-3) using the cleansed email data. Specifically, it splits the data into batches and trains the model over multiple epochs. The trained model learns the writing style and tone of each user.

[1812] Input: Cleansed email data (JSON format)

[1813] Output: A trained generative AI model

[1814] Step 4: Process incoming email

[1815] The user checks the received email and presses the message creation button if a reply is necessary. The terminal extracts the contents of the opened email and sends them to the server.

[1816] Input: Received email content (text format)

[1817] Output: Email content sent to the server (text format)

[1818] Step 5: Recognize emotions

[1819] The server sends the received email content to an emotion recognition engine to analyze the user's emotions. The recognized emotion data is classified into categories such as "joy," "sadness," and "anger," and is used in the copy generation process.

[1820] Input: Received email content (text format)

[1821] Output: Recognized emotion data (categorical format)

[1822] Step 6: Generate copy

[1823] The server provides a prompt to the generative AI model based on the content of the received email and the emotion data, and generates a draft. Example prompt: "Analyze the content of the received email and generate a polite reply that includes an apology. Based on the results of the emotion engine, please create a draft that takes the user's emotions into consideration."

[1824] Input: received email content (text format), recognized emotion data (category format), prompt text

[1825] Output: Generated draft (text format)

[1826] Step 7: Distributing the copy

[1827] The server delivers the generated text and emotion recognition results to the terminal, which then displays them to the user.

[1828] Input: Generated text (text format), emotion recognition results (category format)

[1829] Output: Text and emotion recognition results delivered to the device (text format)

[1830] Step 8: Edit and submit your copy

[1831] The user checks the draft displayed on the device and edits it as necessary. After that, the user presses the send button, and the edited and confirmed draft is sent from the device via the server. It is finally sent as an email.

[1832] Input: Generated draft (text format)

[1833] Output: Final email draft sent (email format)

[1834] (Application example 2)

[1835] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1836] While automated systems for generating reply proposals in electronic communications can significantly reduce time and effort, conventional systems often do not take emotion recognition into account, which can result in inappropriate replies. Furthermore, while there is a growing need for fast and appropriate replies to user comments in content distribution services, current methods suffer from slow response times. The present invention aims to solve these problems by generating fast and appropriate replies while taking user emotions into account.

[1837] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1838] In this invention, the server includes means for collecting past communication data, means for preprocessing and cleansing the past communication data, and means for training an artificial intelligence model that learns each user's writing style using the preprocessed data. This makes it possible to generate an optimal writing style for each user based on past data. Also, in this invention, the server includes means for generating a response draft based on received content and emotion recognition results, and means for automatically analyzing comments and generating reply drafts in a content distribution service. This makes it possible to take user emotions into consideration and generate appropriate and prompt reply drafts in a content distribution service.

[1839] "Historical communications data" is a collection of previous emails, messages, and other communications records.

[1840] "Means of collection" refers to the functions and processes for importing past communication data into the server.

[1841] "Preprocessing and cleansing means" refers to functions and processes for removing unnecessary information from collected data and preparing the data.

[1842] An "artificial intelligence model" is a computer program that uses machine learning techniques to analyze data and is trained to perform specific tasks.

[1843] A "training method" is a function or process that uses historical communication data to teach an artificial intelligence model a particular writing style or pattern.

[1844] The "response draft generation button" is an interface element that the user can operate to start the process of automatically generating a reply draft.

[1845] A "user terminal" is a device operated by a user for sending e-mails and other communications.

[1846] "Received content" refers to the content of messages or emails received on the user's terminal.

[1847] "Emotion recognition results" are information obtained by inferring and analyzing a user's emotions from the content of comments and messages.

[1848] "Comment analysis" is the process of analyzing received comments to understand their content and generate appropriate responses.

[1849] A "means for automatically generating reply drafts" is a function or process for analyzing a comment or message and automatically generating a reply to that comment or message.

[1850] "Delivery means" refers to the function or process for transmitting the generated reply draft to the user terminal.

[1851] "Edit and confirm" is the process by which a user corrects an automatically generated draft and confirms that it should be sent.

[1852] The "means of transmitting communication" is the function or process for transmitting the revised or finalized text.

[1853] MODE FOR CARRYING OUT THE INVENTION

[1854] The present invention is a system for automatically generating reply proposals to comments in a content distribution service. Here, the specific method and operation of the system will be described in detail.

[1855] Server Processing

[1856] The server collects and preprocesses past communication data. During the data collection process, past comment data is retrieved through an API endpoint and saved in JSON format. The collected data is cleansed to remove unnecessary information. The preprocessed data is then used to train a generative AI model that learns each user's writing style. The artificial intelligence technology used here uses OpenAI's API.

[1857] Emotion Recognition and Sentence Generation

[1858] When the server receives a new comment from a user, it analyzes the content and performs emotion recognition. This emotion recognition is performed based on the text extracted from the user's comment, using OpenAI's emotion analysis model. The analysis results are reflected in the generation of reply proposals. An appropriate reply proposal is generated based on the received content and the emotion recognition results. In this process, a prompt sentence is used to instruct the generative AI model to generate a proposal.

[1859] Terminal handling

[1860] The device communicates with the server in response to user operations. When a user receives a comment and replies to it, the device sends the comment content to the server. When the server returns a reply draft and emotion recognition results, the device displays them to the user. The user can check the displayed draft and edit it as necessary. The edited and confirmed draft is sent from the device to the server, and a reply to the comment is finally sent.

[1861] User operations

[1862] When a user receives a comment on a content distribution service and needs to reply to that comment, they press the response draft button. This operation sends the comment content from the device to the server. The server generates a reply draft based on the comment content and emotion recognition results and returns it to the user's device. The user can then check the draft and edit it as necessary. Finally, by pressing the send button, the edited draft is finalized and sent back to the comment poster.

[1863] Specific examples

[1864] For example, if a user receives a comment such as "This news was really helpful!", the device sends the comment to the server, which then uses the following prompt to instruct the generative AI model to generate a sentence:

[1865] Example prompt sentence:

[1866] A user says 'This news was really helpful!'. You could reply with (emotion: joy):

[1867] The generated message, "Thank you for your feedback! We look forward to working with you in the future," is displayed on the user's device. The user can review this message, revise it as necessary, and then send it. This allows for an appropriate and prompt reply that takes the user's feelings into consideration.

[1868] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1869] Step 1:

[1870] The server collects past communication data.

[1871] Input: Historical comment data retrieved from an API endpoint.

[1872] How it works: The server collects data through the API and stores it in JSON format.

[1873] Output: Archived historical comment data.

[1874] Step 2:

[1875] The server pre-processes and cleanses the collected communication data.

[1876] Input: Collected historical comment data.

[1877] How it works: The server strips out unnecessary information from the data, leaving only the necessary fields. Specifically, it extracts the comment content and the identity of the commenter.

[1878] Output: Cleansed communication data.

[1879] Step 3:

[1880] The server uses the preprocessed data to train a generative AI model.

[1881] Input: Cleansed communication data.

[1882] How it works: The server uses the preprocessed data to train a generative AI model that learns each user's writing style. It uses OpenAI's API to generate and train the AI ​​model.

[1883] Output: A fully trained generative AI model.

[1884] Step 4:

[1885] The user receives the comment and presses the response draft button.

[1886] Input: The comment content received by the user.

[1887] How it works: The user presses the button on their device to create a response draft and send it to the server.

[1888] Output: The comment sent.

[1889] Step 5:

[1890] The server analyzes the received content and recognizes the emotion.

[1891] Input: The comment submitted.

[1892] How it works: The server uses OpenAI's sentiment analysis model to analyze the content of comments and recognize their sentiment.

[1893] Output: Recognized emotion data.

[1894] Step 6:

[1895] The server generates a draft response based on the received content and the emotion recognition results.

[1896] Input: Received comment content and recognized emotion data.

[1897] How it works: The server uses the generative AI model to generate a prompt, sends it to the AI ​​model, and generates an appropriate reply.

[1898] Output: The generated response draft.

[1899] Step 7:

[1900] The server delivers the generated response draft to the user terminal.

[1901] Input: The generated response draft.

[1902] Operation: The server sends the generated draft to the user terminal.

[1903] Output: A draft response displayed on the user's terminal.

[1904] Step 8:

[1905] The user reviews the draft response and edits it if necessary.

[1906] Input: Proposed response displayed on the terminal.

[1907] Action: The user reviews the draft response and makes any necessary corrections.

[1908] Output: Edited draft response.

[1909] Step 9:

[1910] The user presses the send button to send the reply draft.

[1911] Input: Edited draft response.

[1912] Action: The user presses the send button and the confirmed response draft is sent via the server.

[1913] Output: Final draft response sent.

[1914] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1915] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1916] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1917] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1918] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1919] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1920] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1921] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1922] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1923] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1924] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1925] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1926] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1927] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1928] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1929] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1930] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1931] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1932] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1933] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1934] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1935] The following is further disclosed regarding the above embodiment.

[1936] (Claim 1)

[1937] a means for collecting historical communications data;

[1938] A means of pre-processing and cleansing historical communication data;

[1939] means for training an artificial intelligence model that learns each user's writing style using the preprocessed data;

[1940] a means for providing a user terminal on which a response draft creation button is arranged that can be operated by the user when responding;

[1941] means for acquiring the content of a communication received when a user presses a button and generating a response draft based on an artificial intelligence model;

[1942] means for delivering the generated response draft to a user terminal;

[1943] means for sending a communication when the user has edited and confirmed the draft response;

[1944] A system including:

[1945] (Claim 2)

[1946] 2. The system according to claim 1, wherein the means for generating a response draft reflects a specific writing style for each user based on past communication data.

[1947] (Claim 3)

[1948] 2. The system according to claim 1, wherein the data collection means and pre-processing means target communication data over the past year and classify and tag the data.

[1949] (Claim 4)

[1950] 2. The system according to claim 1, wherein the response draft generator uses natural language processing technology to analyze the content of the received communication and generate an appropriate response draft.

[1951] "Example 1"

[1952] (Claim 1)

[1953] a means for collecting historical communications data;

[1954] A means of pre-processing and cleansing historical communication data;

[1955] a means for training a generative AI model that learns each user's writing style using the preprocessed data; and

[1956] a means for providing a user terminal on which a response draft creation button is arranged that can be operated by the user when responding;

[1957] A means for acquiring the content of the communication received when the user presses the button and generating a response draft based on the generation AI model;

[1958] means for delivering the generated response draft to a user terminal;

[1959] means for sending a communication when the user has edited and confirmed the draft response;

[1960] A system including:

[1961] (Claim 2)

[1962] 2. The system according to claim 1, wherein the means for generating a response draft reflects a specific writing style for each user based on past communication data.

[1963] (Claim 3)

[1964] 2. The system according to claim 1, wherein the data collection means and pre-processing means target communication data over the past year and classify and tag the data.

[1965] "Application Example 1"

[1966] (Claim 1)

[1967] a means for collecting historical communications data;

[1968] A means of pre-processing and cleansing historical communication data;

[1969] means for training an artificial intelligence model that learns each user's writing style using the preprocessed data;

[1970] a means for providing a user terminal on which a response draft creation button is arranged that can be operated by the user when responding;

[1971] means for acquiring the content of a communication received when a user presses a button and generating a response draft based on an artificial intelligence model;

[1972] means for delivering the generated response draft to a user terminal;

[1973] means for sending a communication when the user has edited and confirmed the draft response;

[1974] A means for automatically analyzing the content of inquiries;

[1975] means for generating a reply draft based on the inquiry;

[1976] means for displaying the generated reply draft to the user in an editable form;

[1977] A means for automatically sending the confirmed reply;

[1978] A system including:

[1979] (Claim 2)

[1980] 2. The system according to claim 1, wherein the means for generating a response draft reflects a specific writing style for each user based on past communication data.

[1981] (Claim 3)

[1982] 2. The system according to claim 1, wherein the data collection means and pre-processing means target communication data over the past year and classify and tag the data.

[1983] "Example 2: Combining Emotion Engines"

[1984] (Claim 1)

[1985] a means for collecting historical communications data;

[1986] A means of pre-processing and cleansing historical communication data;

[1987] means for training an artificial intelligence model that learns each user's writing style using the preprocessed data;

[1988] a means for providing a user terminal on which a response draft creation button is arranged that can be operated by the user when responding;

[1989] means for acquiring the content of a communication received when a user presses a button and generating a response draft based on an artificial intelligence model;

[1990] A means for reflecting emotion data obtained from the emotion recognition engine in copy generation;

[1991] means for delivering the generated response draft and emotion recognition results to a user terminal;

[1992] means for sending a communication when the user has edited and confirmed the draft response;

[1993] A system including:

[1994] (Claim 2)

[1995] 2. The system according to claim 1, wherein the means for generating a response draft reflects a specific writing style for each user based on past communication data.

[1996] (Claim 3)

[1997] 2. The system according to claim 1, wherein the data collection means and pre-processing means target communication data over the past year and classify and tag the data.

[1998] "Application example 2 when combining emotion engines"

[1999] (Claim 1)

[2000] a means for collecting historical communications data;

[2001] A means of pre-processing and cleansing historical communication data;

[2002] means for training an artificial intelligence model that learns each user's writing style using the preprocessed data;

[2003] a means for providing a user terminal on which a response draft creation button is arranged that can be operated by the user when responding;

[2004] means for acquiring the content of a communication received when a user presses a button and generating a response draft based on an artificial intelligence model;

[2005] means for generating a response draft based on the received content and the emotion recognition result;

[2006] A means for automatically analyzing comments and generating reply proposals in a content distribution service;

[2007] means for delivering the generated response draft to a user terminal;

[2008] means for sending a communication when the user has edited and confirmed the draft response;

[2009] A system including:

[2010] (Claim 2)

[2011] 2. The system according to claim 1, wherein the means for generating a response draft reflects a specific writing style for each user based on past communication data and emotion recognition results.

[2012] (Claim 3)

[2013] 2. The system according to claim 1, wherein the data collection means and pre-processing means target communication data over the past year and classify and tag the data. [Explanation of symbols]

[2014] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. a means for collecting historical communications data; A means of pre-processing and cleansing historical communication data; means for training an artificial intelligence model that learns each user's writing style using the preprocessed data; a means for providing a user terminal on which a response draft creation button is arranged that can be operated by the user when responding; means for acquiring the content of a communication received when a user presses a button and generating a response draft based on an artificial intelligence model; means for delivering the generated response draft to a user terminal; means for sending a communication when the user has edited and confirmed the draft response; A system including:

2. 2. The system of claim 1, wherein the means for generating a response draft reflects a specific writing style for each user based on past communication data.

3. 2. The system according to claim 1, wherein said data collection means and pre-processing means target communication data over the past year and classify and tag the data.

4. 2. The system according to claim 1, wherein the response draft generator uses natural language processing technology to analyze the content of the received communication and generate an appropriate response draft.

Images