system

Abstract

We provide the system. [Solution] Means of collecting past conversation data, means for preprocessing the collected conversation data, A means for training a generative AI model using the preprocessed conversation data, A means for analyzing an incoming message and generating multiple reply candidates using the generative AI model, A means of sending the generated reply candidates to the user's device, A means of sending the reply selected by the user, A system that includes this.

Description

Technical Field

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

Background Art

[0002] Patent Document 1 discloses a persona chatbot control method performed by at least one processor, including steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to an explanation of a 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

Summary of the Invention

Problems to be Solved by the Invention

[0004] In conventional messaging applications, users need to consider appropriate replies for each individual message, especially when dealing with a large number of messages for business purposes, which poses a significant burden. Also, in operations that require advanced communication skills such as love - based solicitation in the host industry, it is necessary to improve efficiency by reusing past effective conversations, but it is difficult to do so manually.

Means for Solving the Problems

[0005] The present invention solves the above-mentioned problems by providing a communication support system that includes means for collecting past conversation data, means for preprocessing the collected conversation data, means for training a generative AI model using the preprocessed conversation data, means for analyzing received messages and generating multiple reply candidates using the generative AI model, means for sending the generated reply candidates to the user's terminal, and means for sending the reply selected by the user. With this system, users can easily select the optimal reply candidate learned from past effective conversations and communicate efficiently.

[0006] "Past conversation data" refers to the collection of historical message exchanges that have taken place between users in a communication application.

[0007] A "generative AI model" is an artificial intelligence model that learns from given data and generates appropriate output for new data.

[0008] "Preprocessing" refers to a series of processes that convert raw data into a usable format, including operations such as text cleanup and duplicate removal.

[0009] "Contextual analysis" is the process of understanding the content of a message and the surrounding information in order to derive an appropriate response.

[0010] "Suggested replies" are a selection of appropriate reply messages generated by a generative AI model based on user input.

[0011] A "terminal" is a device that a user uses directly as an interface, and includes devices with communication capabilities such as smartphones and tablets.

[0012] A "server" is a central computer system that performs intensive computational processing such as data collection, analysis, storage, and training of AI models. [Brief explanation of the drawing]

[0013] [Figure 1] This is a conceptual diagram showing an example of the configuration of a data processing system according to the first embodiment. [Figure 2] This is a conceptual diagram showing an example of the essential functions of a data processing device and a smart device according to the first embodiment. [Figure 3] This is a conceptual diagram showing an example of the configuration of a data processing system according to the second embodiment. [Figure 4] This is a conceptual diagram showing an example of the main functions of a data processing device and smart glasses according to the second embodiment. [Figure 5] This is a conceptual diagram showing an example of the configuration of a data processing system according to the third embodiment. [Figure 6] This is a conceptual diagram showing an example of the main functions of a data processing device and a headset-type terminal according to the third embodiment. [Figure 7] This is a conceptual diagram showing an example of the configuration of a data processing system according to the fourth embodiment. [Figure 8] This is a conceptual diagram showing an example of the main functions of a data processing device and a robot according to the fourth embodiment. [Figure 9] This shows an emotion map where multiple emotions are mapped. [Figure 10] This shows an emotion map where multiple emotions are mapped. [Figure 11] This is a sequence diagram showing the processing flow of the data processing system in Example 1. [Figure 12] This is a sequence diagram showing the processing flow of the data processing system in Application Example 1. [Figure 13] This is a sequence diagram showing the processing flow of the data processing system in Example 2, when an emotion engine is combined. [Figure 14] This is a sequence diagram showing the processing flow of the data processing system in Application Example 2, which combines an emotion engine. [Modes for carrying out the invention]

[0014] Hereinafter, an example of an embodiment of a system according to the technology of the present disclosure will be described with reference to the accompanying drawings.

[0015] First, the terms used in the following description will be explained.

[0016] In the following embodiments, a numbered processor (hereinafter simply referred to as "processor") may be a single arithmetic unit or a combination of multiple arithmetic units. Also, the processor may be a single type of arithmetic unit or a combination of multiple types of arithmetic units. Examples of arithmetic units include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), an APU (Accelerated Processing Unit), and the like.

[0017] In the following embodiments, a numbered RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a work memory by the processor.

[0018] In the following embodiments, a numbered storage is one or more non-volatile storage devices that store various programs and various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), or magnetic tapes, and the like.

[0019] In the following embodiments, the signed communication interface (I / F) is an interface that includes a communication processor and an antenna, etc. The communication interface manages communication between multiple computers. Examples of communication standards applicable to the communication interface include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0020] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." That is, "A and / or B" means that it may be A alone, or B alone, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" applies when expressing three or more things linked by "and / or."

[0021] [First Embodiment]

[0022] Figure 1 shows an example of the configuration of the data processing system 10 according to the first embodiment.

[0023] As shown in Figure 1, the 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.

[0024] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0025] The smart device 14 comprises a computer 36, a reception device 38, an output device 40, a camera 42, and a communication interface 44. The computer 36 comprises a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The reception device 38, output device 40, and camera 42 are also connected to the bus 52.

[0026] The reception device 38 is equipped with a touch panel 38A and a microphone 38B, etc., and receives user input. The touch panel 38A receives user input by detecting contact with an object (e.g., a pen or finger). The microphone 38B receives user input by detecting the user's voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and 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.

[0027] 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 perceptible to the user 20 (e.g., audio and / or text). The display 40A displays visible information such as text and images according to instructions from the processor 46. The speaker 40B outputs audio according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0028] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various types of information between processor 46 and processor 28 via network 54.

[0029] Figure 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0030] As shown in Figure 2, in the data processing device 12, a specific processing 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" related to the technology of this 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 according to the specific processing program 56 executed on the RAM 30.

[0031] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0032] In the smart device 14, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. The reception output program 60 is used in conjunction with a 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0033] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0034] System Overview

[0035] The system of this invention proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. This system is implemented using the user's terminal, a server, and a generative AI model.

[0036] Program Processing Overview

[0037] 1. Data collection and preprocessing

[0038] When a user allows the LINE app to collect conversation history data, the device collects past conversation data and sends it to the server. The server preprocesses the received data, performing noise reduction and data cleaning.

[0039] 2. Model training

[0040] The server uses pre-processed data to train generative AI models. Techniques such as sentiment analysis are used here to enable the generation of more natural and optimal responses.

[0041] 3. Processing real-time messages

[0042] When a user receives a new message on LINE, their device sends the message to a server. The server analyzes the context of the received message and generates multiple reply options from a generative AI model based on past conversation data.

[0043] 4. Suggestions for reply options

[0044] The server sends the generated reply options to the terminal, which then displays them to the user. The user selects the most appropriate reply from the displayed options.

[0045] 5. Send a reply

[0046] When a user selects a reply option, the device sends that selected message to the recipient on LINE. This allows users to respond quickly and appropriately.

[0047] Specific example

[0048] The following is a specific example of this system.

[0049] Conversation example

[0050] (Message from a friend): "How have you been lately?"

[0051] When a user receives this message, the device sends it to the server in real time. The server analyzes the context of "How are you doing lately?" and generates possible replies based on past conversation data, such as the following:

[0052] "fine and you?"

[0053] "I'm a little busy, but I'm managing somehow."

[0054] "I recently found a new hobby."

[0055] The server sends these reply options to the device, and the device suggests them to the user. If the user selects "I'm fine! How about you?", the device sends that message to the friend.

[0056] In this way, users can easily provide appropriate responses, enabling efficient and effective conversations, especially in business-related situations or those requiring high communication skills.

[0057] The following describes the processing flow.

[0058] Step 1:

[0059] The user grants the LINE app permission to collect conversation history data.

[0060] Step 2:

[0061] The device retrieves past conversation data from the LINE app.

[0062] Step 3:

[0063] The device sends the acquired conversation data to the server.

[0064] Step 4:

[0065] The server saves the received conversation data to a database.

[0066] Step 5:

[0067] The server preprocesses the conversation data. This preprocessing includes text cleanup, noise reduction, and duplicate removal.

[0068] Step 6:

[0069] The server uses pre-processed data to train generative AI models. Here, techniques such as sentiment analysis are used in conjunction to improve the accuracy of the training.

[0070] Step 7:

[0071] The user receives a new message on LINE.

[0072] Step 8:

[0073] The terminal sends the received message to the server.

[0074] Step 9:

[0075] The server analyzes incoming messages to understand their context. In doing so, it compares them with past conversation data to determine the appropriate context.

[0076] Step 10:

[0077] Based on the analyzed contextual information, the server uses a generative AI model to generate multiple response options.

[0078] Step 11:

[0079] The server sends the generated reply options to the terminal.

[0080] Step 12:

[0081] The device displays reply suggestions to the user.

[0082] Step 13:

[0083] The user selects the most suitable reply from the displayed options.

[0084] Step 14:

[0085] The device sends the reply message selected by the user to the recipient on LINE.

[0086] This series of steps allows users to efficiently provide appropriate responses.

[0087] (Example 1)

[0088] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0089] Modern messaging applications lack sufficient means for users to respond quickly and appropriately. In particular, few existing systems generate responses that consider context based on past conversation data, requiring users to manually formulate replies, which can reduce communication efficiency. Furthermore, technologies that incorporate sentiment analysis to achieve more natural conversations are also insufficient.

[0090] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0091] In this invention, the server includes means for allowing the user to collect conversation history data, means for collecting past conversation data from the terminal and sending it to the server, means for preprocessing the received conversation data to remove noise and clean the data, means for training a generative AI model using the preprocessed conversation data, means for training the generative AI model by incorporating sentiment analysis, means for receiving a new message, means for sending the received message to the server, means for performing contextual analysis of the message and inputting a prompt sentence into the generative AI model to generate multiple reply candidates, means for sending the generated reply candidates to the user's terminal, means for the user to select from the displayed reply candidates, and means for sending the selected reply to the recipient of the messaging application. This enables the user to make quick and appropriate replies.

[0092] A "user" refers to a person who uses a messaging application to have a conversation.

[0093] "Device" refers to electronic devices such as smartphones and computers used by users.

[0094] A "server" refers to a computing system that receives data sent from a terminal and processes it.

[0095] "Conversation history data" refers to records of conversations that have taken place on messaging applications in the past.

[0096] "Preprocessing" refers to the process of preparing received data into an analyzable format. Specifically, this includes noise reduction and data cleaning.

[0097] A "generative AI model" refers to a model that uses artificial intelligence to generate output in response to a specific input. An example is a text generation model.

[0098] "Sentiment analysis" refers to the technique of identifying and classifying the emotions contained within text.

[0099] "New messages" refers to the most recent messages a user receives in their messaging application.

[0100] "Contextual analysis" refers to analytical methods used to understand the meaning and underlying intentions of text.

[0101] A "prompt" refers to the document or question that is input into a generative AI model.

[0102] "Suggested replies" refers to multiple response options generated by the server using a generated AI model.

[0103] A "messaging application" refers to software that users use to send and receive messages.

[0104] Modes for carrying out the invention

[0105] This invention relates to a system that suggests appropriate replies to messages received by a user in a messaging application, based on past conversation data. This system is implemented using the user's terminal, a server, and a generative AI model.

[0106] Data collection and preprocessing

[0107] When a user allows a messaging application to collect conversation history data, the device collects past conversation data and sends it to the server. The server preprocesses the received conversation data, performing noise reduction and data cleaning. Specifically, it uses libraries such as Pandas and NumPy to clean up the data. For example, it removes unnecessary information such as newline characters and emojis.

[0108] Model training

[0109] The server trains generative AI models using preprocessed data. Specifically, it uses the Hugging Face Transformers library to train generative AI models such as BERT and GPT-3®. In addition, the BERT model is used for sentiment analysis to label emotions within text. For example, the model is trained based on past conversation data regarding the message "How are you doing lately?".

[0110] Processing real-time messages

[0111] When a user receives a new message, the device sends it to the server. The server uses the spacy library to analyze the context of the message and inputs prompts into a generative AI model to generate an appropriate reply. For example, in response to a message like "How are you doing?", the server analyzes the context from past conversation data and generates an appropriate reply.

[0112] Suggested replies

[0113] The server sends several generated reply options to the device. The device then displays the reply options in a user-friendly interface using JavaScript® or Swift. For example, it might suggest the following reply options to the user:

[0114] "fine and you?"

[0115] "I'm a little busy, but I'm managing somehow."

[0116] "I recently found a new hobby."

[0117] Send a reply

[0118] The user selects one of the displayed reply options. The device then sends the selected reply to the recipient in the messaging application. This is done using the messaging application's API.

[0119] This specific process allows users to leverage generative AI models and sentiment analysis to provide quick and appropriate responses. This enables efficient and effective conversations, especially in business-oriented situations or those requiring high-level communication skills.

[0120] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0121] Step 1: Data collection and preprocessing

[0122] The user grants permission for the messaging application to collect conversation history data. The device collects the conversation history data permitted by the user and sends it to the server. The server preprocesses the received conversation data, performing noise reduction and data cleaning. Specifically, it uses the Pandas library to remove unnecessary line breaks and emojis and format the data.

[0123] Input: User-authorized conversation history data

[0124] Output: Pre-processed clean data

[0125] Step 2: Model Training

[0126] The server trains a generative AI model using pre-processed conversational data. Here, the Hugging Face Transformers library is used to train models such as BERT and GPT-3. The BERT model is also used for sentiment analysis, classifying the sentiment of each message. For example, for a message like "How are you?", conversations with similar contexts are extracted from past conversational data and used for training.

[0127] Input: Preprocessed conversation data

[0128] Output: Trained generative AI model

[0129] Step 3: Receiving real-time messages

[0130] When a user receives a new message in their messaging application, the device sends this new message to the server. The message includes the sender, recipient, and message content.

[0131] Input: New incoming message

[0132] Output: Message data sent to the server

[0133] Step 4: Contextual analysis and generation of reply candidates

[0134] The server performs contextual analysis on newly received messages. This analysis uses the spacy library to understand the meaning of the message. It then generates an appropriate prompt and inputs it into a generative AI model to produce multiple response options. For example, in response to the message "How are you?", based on past conversation data and sentiment analysis results, it generates response options such as: "I'm fine! How about you?", "I'm a little busy, but I'm managing," and "I recently found a new hobby."

[0135] Input: Newly received message

[0136] Output: Multiple response candidates generated

[0137] Step 5: Suggesting possible replies

[0138] The server sends the generated reply suggestions to the device. The device then displays the reply suggestions to the user using a simple and user-friendly interface based on JavaScript or Swift.

[0139] Input: Multiple generated reply suggestions

[0140] Output: Reply suggestions displayed to the user

[0141] Step 6: Select and send your reply.

[0142] The user selects the most suitable reply from the displayed options. The device then sends the selected reply to the recipient in the messaging application. This is done using the messaging application's API.

[0143] Input: User's selected reply

[0144] Output: Reply message sent to the recipient

[0145] (Application Example 1)

[0146] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart device 14 will be referred to as the "terminal."

[0147] Traditional food delivery services often struggle to respond quickly and accurately to user messages regarding order confirmations, order changes, and complaints. This frequently results in a poor user experience, potentially leading to decreased service usage and satisfaction. Furthermore, in business environments requiring frequent communication, greater efficiency is needed.

[0148] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0149] In this invention, the server includes means for collecting past conversation data, means for preprocessing the collected conversation data, means for training a generative AI model using the preprocessed conversation data, means for analyzing received messages and generating multiple reply candidates using the generative AI model, means for sending the generated reply candidates to the user's terminal, means for sending the reply selected by the user, and means for suggesting appropriate reply candidates to order confirmation messages, order change request messages, or complaint handling messages when the user is using an online delivery service. This enables users to make quick and appropriate replies, improving the user experience and increasing the efficiency of the delivery service.

[0150] "Past conversation data" refers to records of communications a user has had in the past, and this includes various formats such as text messages, voice messages, and chat logs.

[0151] "Preprocessing" refers to a series of processes that remove noise from collected conversation data, convert it to the required data format, and prepare it for analysis.

[0152] A "generative AI model" is a model that uses AI algorithms to generate new text data, and can also generate appropriate answers and replies based on the user's past conversation data.

[0153] "Suggested replies" refer to multiple possible responses to an incoming message, allowing the user to choose from among them.

[0154] "Online delivery service" refers to any service that allows users to order food and goods via the internet and have them delivered to their homes or workplaces.

[0155] An "order confirmation message" refers to a message sent to a user after they have placed an order for a product or service, to confirm the details of that order.

[0156] An "order change request message" refers to a message sent when a user wants to make changes to an order they have already placed.

[0157] A "complaint response message" refers to a message sent by a user when they are dissatisfied with a product or service or have a problem with it, and appropriate action is taken based on this message.

[0158] The present invention provides a system that suggests appropriate responses to various messages received by a user while using an online delivery service. This system includes the user's terminal, a server, and a generative AI model.

[0159] Specific System Procedures

[0160] The server first collects past conversation data from the user's device. This data includes text messages and chat logs, and once collected, it undergoes preprocessing such as noise reduction. The preprocessed data is then used to train a generative AI model. By incorporating sentiment analysis techniques at this stage, it becomes possible to generate more natural and appropriate responses.

[0161] When a user receives a new message via the delivery service, the device sends the message to the server in real time. The server analyzes the received message and generates multiple reply options using a generative AI model. The generated reply options are sent to the device and displayed to the user. The user selects the appropriate reply from the displayed options, and the selected reply is sent to the recipient.

[0162] Hardware and software

[0163] The hardware used includes the user's smartphone or other device, as well as a server. The software used includes scripts for data collection and preprocessing, generative AI models (e.g., OpenAI® GPT-3), and sentiment analysis libraries (e.g., TextBlob).

[0164] As a concrete example, consider a case where a user receives an "order confirmation message" while using a food delivery service. The following response options are generated in response to this message.

[0165] Example of a prompt

[0166] Past conversation history: "Order confirmation: Yes, please." "Delivery delay: Excuse me, is it delayed?"

[0167] Current message: "We will arrive shortly."

[0168] Possible replies to generate:

[0169] 1. "Understood, thank you."

[0170] 2. "How much longer until we arrive?"

[0171] In this way, users can respond quickly and accurately, improving their experience with online delivery services. Furthermore, increased efficiency can be expected in the work environment.

[0172] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0173] Step 1:

[0174] The user authorizes the collection of past conversation data through their device. The device, upon receiving permission, collects past chat logs and message history from LINE and other messaging applications. The collected data is then sent directly to the server. The input is past conversation data, and the output is the collected data. Specifically, this involves sending past message history to the server in text format.

[0175] Step 2:

[0176] The server preprocesses the conversation data received from the terminal. Preprocessing includes removing noise and filtering out unnecessary information from the data. The input is the collected data, and the output is the preprocessed, cleaned data. Specifically, this involves performing text analysis on the data and removing meaningless information and error messages.

[0177] Step 3:

[0178] The server trains a generative AI model using preprocessed data. Sentiment analysis techniques are also used to learn the emotional nuances of the data. The input is preprocessed data, and the output is the trained generative AI model. Specifically, this involves performing sentiment analysis using a Python library (e.g., TextBlob) and training a generative AI model such as OpenAI GPT-3.

[0179] Step 4:

[0180] When a user receives a new message via LINE or another messaging app, the device sends the message to a server. The server analyzes the context of the received message and generates multiple reply options from a generative AI model based on past conversation data. The input is the newly received message, and the output is the generated reply options. Specifically, this involves text analysis of the received message and outputting appropriate reply options.

[0181] Step 5:

[0182] The server sends the generated reply options to the user's terminal. The terminal displays the reply options to the user. The user selects the most appropriate reply from the displayed options. The input is the generated reply options, and the output is the reply selected by the user. Specifically, this involves displaying a list of reply options in the terminal's user interface and accepting the user's selection.

[0183] Step 6:

[0184] The user sends their selected reply from their device to the server, which then forwards that selected reply to the LINE recipient. The input is the user-selected reply, and the output is the actual reply message sent. Specifically, this involves sending the selected message to the recipient via the LINE API.

[0185] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0186] System Overview

[0187] The present invention's system proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. In particular, by combining it with an emotion engine that recognizes the user's emotions, it enables the generation of more personalized replies. This system is implemented using the user's terminal, a server, a generative AI model, and an emotion engine.

[0188] Program Processing Overview

[0189] 1. Data collection and preprocessing

[0190] When a user allows the LINE app to collect conversation history data, the device collects past conversation data and sends it to a server. The server stores the received data in a database and performs preprocessing. Preprocessing includes text cleanup, noise reduction, and duplicate removal.

[0191] 2. Model training

[0192] The server uses pre-processed data to train a generative AI model. During this process, data based on the user's emotional state is also used, incorporating sentiment analysis into the training. This allows the AI ​​model to respond to fluctuations in the user's emotions.

[0193] 3. Processing real-time messages

[0194] When a user receives a new message on LINE, their device sends the message to the server. The server performs contextual analysis of the received message and uses an emotion engine to recognize the user's emotional state. This is then compared with past conversation data to analyze the appropriate context.

[0195] 4. Generating reply suggestions

[0196] The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the user's emotional state. This ensures that the response options also match the user's emotions.

[0197] 5. Suggestions for reply options

[0198] The server sends generated reply options to the terminal, which then suggests them to the user. The user selects the most suitable reply from the displayed options.

[0199] 6. Send a reply

[0200] When a user selects a reply option, the device sends that selected message to the recipient on LINE. This allows users to respond quickly and appropriately.

[0201] Specific example

[0202] The following is a specific example of this system.

[0203] Conversation example

[0204] (Message from a friend): "How have you been lately?"

[0205] When a user receives this message, the device sends it to the server in real time. The server analyzes the context of "How are you doing?" and uses an emotion engine to identify the user's current emotion. For example, if the server identifies the user's current emotion as "stressed," it matches this with past conversation data to generate potential replies such as the following:

[0206] "I'm a little busy, but I'm managing somehow."

[0207] "Work has been really tough lately, how about you?"

[0208] "I'm feeling a bit stressed, but thank you for listening."

[0209] The server sends these response options to the device, and the device suggests them to the user. If the user selects "I'm a bit busy, but I'm managing," the device sends that message to the friend.

[0210] In this way, users can easily provide appropriate replies, enabling efficient and effective conversations, especially in business-related situations or those requiring high communication skills. Furthermore, a key feature is the ability to provide more personalized replies by considering the user's emotional state.

[0211] The following describes the processing flow.

[0212] System Overview

[0213] The system of this invention proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. In particular, by combining it with an emotion engine that recognizes the user's emotions, it enables the generation of more personalized replies. This system is implemented using the user's terminal, a server, a generative AI model, and an emotion engine.

[0214] Program Processing Overview

[0215] Step 1:

[0216] The user grants the LINE app permission to collect conversation history data.

[0217] Step 2:

[0218] The device retrieves past conversation data from the LINE app.

[0219] Step 3:

[0220] The device sends the acquired conversation data to the server.

[0221] Step 4:

[0222] The server saves the received conversation data to a database.

[0223] Step 5:

[0224] The server preprocesses the conversation data. Preprocessing includes text cleanup, noise reduction, and duplicate removal.

[0225] Step 6:

[0226] The server uses pre-processed data to train generative AI models. It also incorporates sentiment analysis into the training process, using data based on the user's emotional state.

[0227] Step 7:

[0228] The user receives a new message on LINE.

[0229] Step 8:

[0230] The terminal sends the received message to the server.

[0231] Step 9:

[0232] The server analyzes incoming messages to understand their context. During this process, it also recognizes the user's emotional state using an emotion engine.

[0233] Step 10:

[0234] The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the user's emotional state.

[0235] Step 11:

[0236] The server sends the generated reply options to the terminal.

[0237] Step 12:

[0238] The device displays reply suggestions to the user.

[0239] Step 13:

[0240] The user selects the most suitable reply from the displayed options.

[0241] Step 14:

[0242] The device sends the reply message selected by the user to the recipient on LINE.

[0243] Specific example

[0244] Conversation example

[0245] (Message from a friend): "How have you been lately?"

[0246] Step 1:

[0247] The user receives this message.

[0248] Step 2:

[0249] The device sends the message to the server in real time.

[0250] Step 3:

[0251] The server analyzes the context of the question "How are you doing lately?" and uses its emotion engine to identify the user's current emotional state. For example, it might identify the user's current emotion as "feeling stressed."

[0252] Step 4:

[0253] The server generates response options based on past conversation data and emotional states analyzed by the emotion engine, such as the following:

[0254] "I'm a little busy, but I'm managing somehow."

[0255] "Work has been really tough lately, how about you?"

[0256] "I'm feeling a bit stressed, but thank you for listening."

[0257] Step 5:

[0258] The server sends these suggested replies to the terminal.

[0259] Step 6:

[0260] The device suggests and displays reply options to the user.

[0261] Step 7:

[0262] The user selects, "I'm a little busy, but I'm managing."

[0263] Step 8:

[0264] The device sends the selected message to the friend.

[0265] This system not only automatically suggests appropriate replies but also takes into account the user's emotional state, enabling more personalized communication. Its effectiveness is particularly evident in business settings and situations requiring advanced communication skills.

[0266] (Example 2)

[0267] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the smart device 14 as the "terminal".

[0268] Traditional messaging applications have struggled to generate appropriate replies to received messages. In particular, existing technologies were insufficient to generate personalized replies that took into account the user's emotional state. As a result, users had to manually compose replies, hindering efficient and effective communication.

[0269] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0270] In this invention, the server includes means for collecting past conversation data, means for preprocessing the collected conversation data, means for training a generative AI model using the preprocessed conversation data, means for analyzing an incoming message and generating multiple reply candidates using the generative AI model, means for sending the generated reply candidates to the user's terminal, means for sending the reply selected by the user, means for using an emotion engine to recognize the user's emotional state, and means for generating reply candidates based on contextual information of the incoming message and the user's emotional state. This enables the generation of personalized replies that take the user's emotions into consideration, allowing for efficient, rapid, and appropriate replies.

[0271] "Past conversation data" refers to information that includes the history of messages a user has sent in the past.

[0272] "Preprocessing" refers to the process of removing noise from received data, cleaning up text, and removing duplicates.

[0273] A "generative AI model" is an artificial intelligence model that is trained based on collected data and generates appropriate outputs (responses) in response to given inputs.

[0274] An "emotion engine" is a tool or software used to analyze and identify a user's emotional state.

[0275] "Contextual information" refers to information that includes meaning derived from the content of a message and the flow of past conversations.

[0276] "Suggested replies" refer to multiple response options proposed by the generative AI model.

[0277] A "server" is a computer system used for tasks such as data collection, preprocessing, analysis, and training of AI models.

[0278] The "user's terminal" refers to a device (such as a smartphone, tablet, PC, etc.) used by the user, which has the function of receiving and displaying reply candidates from the server.

[0279] The "message application" is a general term for application software for the user to send and receive text messages.

[0280] Mode for Implementing the Invention

[0281] The system of the present invention proposes an appropriate reply based on past conversation data for a message received by the user in a message application such as LINE. In particular, this system enables the generation of more personalized replies by combining an emotion engine that recognizes the user's emotions. The mode for implementing this invention will be described in detail below.

[0282] The system is realized by using the user's terminal, server, generation AI model, and emotion engine. As specific hardware, the user's terminal includes smartphones, tablets, PCs, etc. A high-performance computer system (for example, a server machine equipped with NVIDIA GPU) is used for the server. Database management systems such as MySQL (registered trademark) and PostgreSQL are used for the database.

[0283] Data Collection and Pretreatment

[0284] When the user permits the collection of conversation history data in the LINE application, the terminal collects past conversation data from the LINE application and transmits the data to the server. The server saves the received conversation data in the database. Then, as pretreatment, text cleaning, noise removal, and duplication deletion are performed. By this pretreatment, the quality of the data is improved and it becomes suitable for the training of the generation AI model.

[0285] Model Training

[0286] The server uses the pre - processed data and the sentiment engine to train a generative AI model (e.g., GPT - 3 or BERT). In this training process, data based on the user's emotional state is also used. For sentiment analysis, a sentiment engine (e.g., IBM Watson (registered trademark), Azure (registered trademark) Text Analytics) is utilized. Thus, the generative AI model can also handle the user's emotional fluctuations.

[0287] Real - time message processing

[0288] When the user receives a new message on LINE, the terminal sends the message to the server in real - time. The server performs context analysis on the received message using natural language processing techniques (e.g., spaCy or NLTK). Furthermore, a sentiment engine is used to recognize the user's emotional state.

[0289] Generation of reply candidates

[0290] Based on the analyzed context information and the user's emotional state, the server uses the generative AI model to generate multiple reply candidates. The generated reply candidates will be content suitable for the context and the user's emotions.

[0291] Proposal and selection of reply candidates

[0292] The server sends the generated reply candidates to the terminal, and the terminal proposes them to the user. The user selects the most appropriate one from the displayed multiple reply candidates.

[0293] Sending of the reply

[0294] When the user selects a reply candidate, the terminal sends the selected message to the LINE counterpart. Thus, the user can quickly make an efficient and appropriate reply.

[0295] Specific example

[0296] The following are specific examples of this system.

[0297] (Message from a friend): "How are you recently?"

[0298] When the user receives this message, the terminal sends the message to the server in real time. The server analyzes the context of "How are you recently?" and uses the emotion engine to identify the current user's emotion. For example, if the current emotion of the user is identified as "feeling stressed", the following reply candidates are generated by matching with past conversation data.

[0299] "I'm a bit busy, but I'm managing somehow."

[0300] "Work has been tough lately. How about you?"

[0301] "I've been building up a little stress, but thanks for listening."

[0302] The server sends these reply candidates to the terminal, and the terminal proposes them to the user. If the user selects "I'm a bit busy, but I'm managing somehow.", the terminal sends that message to the friend. In this way, the user can reply quickly and appropriately.

[0303] The flow of specific processing in Example 2 will be described using FIG. 13.

[0304] Step 1:

[0305] The user permits the collection of conversation history data in the LINE app.

[0306] Input: User's permission setting

[0307] Output: Instruction to collect conversation history data

[0308] Specific operation: The user permits the collection of conversation data in the LINE settings screen.

[0309] Step 2:

[0310] The device retrieves past conversation data from the LINE app and sends that data to the server.

[0311] Input: Conversation history data

[0312] Output: Sending conversation data to the server

[0313] Specific operation: A background process on the device extracts LINE conversation history data and sends it to the server via a secure protocol (e.g., HTTPS).

[0314] Step 3:

[0315] The server saves the received conversation data to the database.

[0316] Input: Conversation data

[0317] Output: Data stored in the database

[0318] Specific operation: The server receives the conversation data and inserts it into the database in a specific format (e.g., JSON format).

[0319] Step 4:

[0320] The server performs preprocessing on the data it has stored.

[0321] Input: Conversation data stored in the database

[0322] Output: Preprocessed data

[0323] Specific actions: The server performs the following preprocessing:

[0324] Text cleanup: Remove unnecessary symbols and spaces.

[0325] Noise reduction: Removes spam messages and meaningless data.

[0326] Remove duplicates: If the same message is recorded multiple times, it will be merged into one.

[0327] Step 5:

[0328] The server uses pre-processed data to train the generated AI model.

[0329] Input: Preprocessed data

[0330] Output: Trained Generative AI Model

[0331] Specific operation: The server uses high-performance computing resources (e.g., NVIDIA GPU) to train a generative AI model (e.g., GPT-3). During this process, sentiment analysis data is also incorporated into the model using an emotion engine.

[0332] Step 6:

[0333] The user receives a new message on LINE.

[0334] Input: New message

[0335] Output: New message displayed on the terminal

[0336] Specific action: The LINE app receives a new message and displays it on the device.

[0337] Step 7:

[0338] The terminal sends the received message to the server.

[0339] Input: Received message

[0340] Output: Sending a message to the server

[0341] Specific operation: The device forwards messages received from the LINE app to the server, and the server receives them.

[0342] Step 8:

[0343] The server analyzes the context of the received message and uses an emotion engine to recognize the user's emotional state.

[0344] Input: Received message

[0345] Output: Contextual information, emotional state

[0346] Specific operation: The server uses natural language processing technology (e.g., spaCy) to analyze the context of the message and uses an emotion engine to recognize the user's emotional state.

[0347] Step 9:

[0348] Based on the analyzed contextual information and the user's emotional state, the server uses a generative AI model to generate multiple response options.

[0349] Input: Contextual information, emotional state, trained generative AI model

[0350] Output: Reply candidates

[0351] Specific operation: The server inputs a prompt message into the AI ​​model, which then generates response candidates that take into account contextual information and emotional state.

[0352] Step 10:

[0353] The server sends the generated response options to the terminal, which then suggests them to the user.

[0354] Input: Reply suggestions

[0355] Output: Sending reply suggestions to the terminal, presenting them to the user.

[0356] Specific operation: The server sends suggested replies to the terminal, and the terminal displays them to the user.

[0357] Step 11:

[0358] When a user selects a reply option, the device sends that selected message to the recipient via the LINE app through the server.

[0359] Input: Selected reply options

[0360] Output: Message sent to the recipient via the LINE app

[0361] Specific operation: The user selects a reply option on the screen, and the device sends that message to the recipient via the LINE app through the server.

[0362] (Application Example 2)

[0363] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as a "server" and the smart device 14 as a "terminal".

[0364] In physical stores, effective communication between staff and customers is crucial for improving customer satisfaction. However, responding quickly and appropriately to customer requests and questions is not easy. In particular, being able to respond in a way that is sensitive to the customer's emotional state would further enhance the customer experience, but there is a lack of effective tools to achieve this.

[0365] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting past dialogue data, means for pre-processing the collected dialogue data, means for training a generative AI model using the pre-processed dialogue data, means for analyzing received messages and generating a plurality of response candidates using the generative AI model, means for transmitting the generated response candidates to the user's display device, means for transmitting the response selected by the user, and means for comparing and analyzing past dialogue history using an emotion engine that identifies the customer's emotional state. As a result, staff can provide quick and appropriate responses to customer requests and questions, and in particular, personalized responses that are in line with the customer's emotions become possible.

[0366] "Dialogue data" refers to a record of conversations that take place between a user and another person.

[0367] "Preprocessing" is the process of removing unnecessary information from collected data and converting it into a usable format.

[0368] A "generative AI model" is an artificial intelligence model that learns patterns based on collected data and performs specific tasks.

[0369] "Received messages" are messages that a user receives from other people.

[0370] "Response candidates" are multiple reply options proposed by a generative AI model.

[0371] A "display device" is a device that a user uses to view response options and other information.

[0372] An "emotion engine" is a system that analyzes and identifies emotions from text, audio, and other sources.

[0373] "Contextual analysis" is the process of understanding the content and background of a message and extracting information to generate an appropriate response.

[0374] The "Concierge Assistant" system of this invention is a tool that supports communication between staff and customers in physical stores. This system operates in the following steps:

[0375] Hardware and software to be used

[0376] Hardware: Tablets for customer service (e.g., iPad (registered trademark))

[0377] software:

[0378] Message management application

[0379] Emotion engine (e.g., IBM Watson Tone Analyzer)

[0380] Generative AI models (e.g., GPT-4(registered trademark))

[0381] System Implementation Procedures

[0382] 1. Data collection and preprocessing

[0383] The server collects past conversation data from customer service tablets. The collected data is pre-processed on the server. Pre-processing includes text cleanup, noise reduction, and duplicate removal.

[0384] 2. Training of generative AI models

[0385] The server uses pre-processed data to train a generative AI model. During this process, an emotion engine is used to consider the customer's emotional state, resulting in a model that can adapt to individual customer emotional fluctuations.

[0386] 3. Processing real-time messages

[0387] When a customer service tablet receives a new question or request from a customer, it sends the content to the server. The server contextualizes the received message and uses an emotion engine to identify the customer's emotional state.

[0388] 4. Generating response candidates

[0389] The server uses a generative AI model to generate multiple response candidates based on the analyzed contextual information and the customer's emotional state.

[0390] 5. Proposal of response candidates

[0391] The server sends the generated response options to a customer service tablet, which then displays them to the staff. The staff then selects the most appropriate response option.

[0392] 6. Sending a response

[0393] When a staff member selects a response option, the customer service tablet displays that message to the customer.

[0394] Specific example

[0395] For example, if a customer asks, "How do I use this product?", the customer service tablet sends the question to the server. The server analyzes the context of "How do I use this product?" and identifies the customer's emotion through an emotion engine (e.g., IBM Watson Tone Analyzer). If the customer is identified as "interested," a generative AI model (e.g., GPT-4) generates response candidates such as the following.

[0396] "It would be best if you press this button first, and then do this next."

[0397] "We also have an easy-to-understand demo video; would you like to watch it?"

[0398] "If you'd like to try it out, we can help you."

[0399] If a staff member selects "We also have an easy-to-understand demo video; would you like to watch it?" from the suggested responses, the customer service tablet will display this message to the customer.

[0400] Example of a prompt

[0401] "Based on past interaction history and the customer's emotional state, please propose the most appropriate response."

[0402] This system will enable in-store staff to provide customers with quick and personalized service, which is expected to improve customer satisfaction.

[0403] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0404] Step 1:

[0405] Data collection and preprocessing

[0406] Input: Past conversation data collected by the user using a customer service tablet.

[0407] Processing: The terminal sends past conversation data to the server. The server stores the received data in a database and performs preprocessing such as data cleanup, noise reduction, and duplicate data removal.

[0408] Output: Pre-processed, clean dialogue data.

[0409] Step 2:

[0410] Training of generative AI models

[0411] Input: Preprocessed dialogue data and sentiment analysis data from the sentiment engine.

[0412] Processing: The server trains a generative AI model using pre-processed data. During this process, an emotion engine is used to incorporate emotional information contained in the dialogue data, building a model that can respond to changes in the user's emotions.

[0413] Output: A pre-trained generative AI model capable of responding to the user's emotional state.

[0414] Step 3:

[0415] Processing real-time messages

[0416] Input: A message from a new customer that the user receives on their customer service tablet.

[0417] Processing: The terminal sends the received message to the server. The server analyzes the context of the received message and uses an emotion engine to identify the customer's emotional state.

[0418] Output: Analyzed contextual information and emotional state information.

[0419] Step 4:

[0420] Generating response candidates

[0421] Input: Analyzed contextual information and emotional state information.

[0422] Processing: The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the customer's emotional state.

[0423] Output: Multiple response candidates generated by the server.

[0424] Step 5:

[0425] Suggested response options

[0426] Input: Multiple response candidates generated by the server.

[0427] Processing: The server sends the generated response options to the customer service tablet, which then presents them to the staff. The staff selects the most appropriate response from the screen.

[0428] Output: Staff-selected response.

[0429] Step 6:

[0430] Sending a response

[0431] Input: Response selected by the staff.

[0432] Processing: The terminal sends the selected response message back to the customer.

[0433] Output: Response message sent to the customer.

[0434] 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 user input for the result of the specific processing. The control unit 46A transmits the audio data indicating 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.

[0435] Data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of data generation model 58 is ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0436] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart device 14.

[0437] [Second Embodiment]

[0438] Figure 3 shows an example of the configuration of the data processing system 210 according to the second embodiment.

[0439] As shown in Figure 3, the data processing system 210 includes a data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0440] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0441] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication interface 44. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, and camera 42 are also connected to the bus 52.

[0442] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0443] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0444] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0445] Figure 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Figure 4, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0446] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[0447] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0448] In the smart glasses 214, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0449] Next, the identification processing performed by the identification processing unit 290 of the data processing device 12 will be described. 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".

[0450] System Overview

[0451] The system of this invention proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. This system is implemented using the user's terminal, a server, and a generative AI model.

[0452] Program Processing Overview

[0453] 1. Data collection and preprocessing

[0454] When a user allows the LINE app to collect conversation history data, the device collects past conversation data and sends it to the server. The server preprocesses the received data, performing noise reduction and data cleaning.

[0455] 2. Model training

[0456] The server uses pre-processed data to train generative AI models. Techniques such as sentiment analysis are used here to enable the generation of more natural and optimal responses.

[0457] 3. Processing real-time messages

[0458] When a user receives a new message on LINE, their device sends the message to a server. The server analyzes the context of the received message and generates multiple reply options from a generative AI model based on past conversation data.

[0459] 4. Suggestions for reply options

[0460] The server sends the generated reply options to the terminal, which then displays them to the user. The user selects the most appropriate reply from the displayed options.

[0461] 5. Send a reply

[0462] When a user selects a reply option, the device sends that selected message to the recipient on LINE. This allows users to respond quickly and appropriately.

[0463] Specific example

[0464] The following is a specific example of this system.

[0465] Conversation example

[0466] (Message from a friend): "How have you been lately?"

[0467] When a user receives this message, the device sends it to the server in real time. The server analyzes the context of "How are you doing lately?" and generates possible replies based on past conversation data, such as the following:

[0468] "fine and you?"

[0469] "I'm a little busy, but I'm managing somehow."

[0470] "I recently found a new hobby."

[0471] The server sends these reply options to the device, and the device suggests them to the user. If the user selects "I'm fine! How about you?", the device sends that message to the friend.

[0472] In this way, users can easily provide appropriate responses, enabling efficient and effective conversations, especially in business-related situations or those requiring high communication skills.

[0473] The following describes the processing flow.

[0474] Step 1:

[0475] The user grants the LINE app permission to collect conversation history data.

[0476] Step 2:

[0477] The device retrieves past conversation data from the LINE app.

[0478] Step 3:

[0479] The device sends the acquired conversation data to the server.

[0480] Step 4:

[0481] The server saves the received conversation data to a database.

[0482] Step 5:

[0483] The server preprocesses the conversation data. This preprocessing includes text cleanup, noise reduction, and duplicate removal.

[0484] Step 6:

[0485] The server uses pre-processed data to train generative AI models. Here, techniques such as sentiment analysis are used in conjunction to improve the accuracy of the training.

[0486] Step 7:

[0487] The user receives a new message on LINE.

[0488] Step 8:

[0489] The terminal sends the received message to the server.

[0490] Step 9:

[0491] The server analyzes incoming messages to understand their context. In doing so, it compares them with past conversation data to determine the appropriate context.

[0492] Step 10:

[0493] Based on the analyzed contextual information, the server uses a generative AI model to generate multiple response options.

[0494] Step 11:

[0495] The server sends the generated reply options to the terminal.

[0496] Step 12:

[0497] The device displays reply suggestions to the user.

[0498] Step 13:

[0499] The user selects the most suitable reply from the displayed options.

[0500] Step 14:

[0501] The device sends the reply message selected by the user to the recipient on LINE.

[0502] This series of steps allows users to efficiently provide appropriate responses.

[0503] (Example 1)

[0504] Next, we will describe Example 1. 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".

[0505] Modern messaging applications lack sufficient means for users to respond quickly and appropriately. In particular, few existing systems generate responses that consider context based on past conversation data, requiring users to manually formulate replies, which can reduce communication efficiency. Furthermore, technologies that incorporate sentiment analysis to achieve more natural conversations are also insufficient.

[0506] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0507] In this invention, the server includes means for allowing the user to collect conversation history data, means for collecting past conversation data from the terminal and sending it to the server, means for preprocessing the received conversation data to remove noise and clean the data, means for training a generative AI model using the preprocessed conversation data, means for training the generative AI model by incorporating sentiment analysis, means for receiving a new message, means for sending the received message to the server, means for performing contextual analysis of the message and inputting a prompt sentence into the generative AI model to generate multiple reply candidates, means for sending the generated reply candidates to the user's terminal, means for the user to select from the displayed reply candidates, and means for sending the selected reply to the recipient of the messaging application. This enables the user to make quick and appropriate replies.

[0508] A "user" refers to a person who uses a messaging application to have a conversation.

[0509] "Device" refers to electronic devices such as smartphones and computers used by users.

[0510] A "server" refers to a computing system that receives data sent from a terminal and processes it.

[0511] "Conversation history data" refers to records of conversations that have taken place on messaging applications in the past.

[0512] "Preprocessing" refers to the process of preparing received data into an analyzable format. Specifically, this includes noise reduction and data cleaning.

[0513] A "generative AI model" refers to a model that uses artificial intelligence to generate output in response to a specific input. An example is a text generation model.

[0514] "Sentiment analysis" refers to the technique of identifying and classifying the emotions contained within text.

[0515] "New messages" refers to the most recent messages a user receives in their messaging application.

[0516] "Contextual analysis" refers to analytical methods used to understand the meaning and underlying intentions of text.

[0517] A "prompt" refers to the document or question that is input into a generative AI model.

[0518] "Suggested replies" refers to multiple response options generated by the server using a generated AI model.

[0519] A "messaging application" refers to software that users use to send and receive messages.

[0520] Modes for carrying out the invention

[0521] This invention relates to a system that suggests appropriate replies to messages received by a user in a messaging application, based on past conversation data. This system is implemented using the user's terminal, a server, and a generative AI model.

[0522] Data collection and preprocessing

[0523] When a user allows a messaging application to collect conversation history data, the device collects past conversation data and sends it to the server. The server preprocesses the received conversation data, performing noise reduction and data cleaning. Specifically, it uses libraries such as Pandas and NumPy to clean up the data. For example, it removes unnecessary information such as newline characters and emojis.

[0524] Model training

[0525] The server trains generative AI models using preprocessed data. Specifically, it uses the Hugging Face Transformers library to train generative AI models such as BERT and GPT-3. In addition, it uses the BERT model to label emotions within text for sentiment analysis. For example, it trains the model based on past conversation data regarding the message "How are you doing lately?".

[0526] Processing real-time messages

[0527] When a user receives a new message, the device sends it to the server. The server uses the spacy library to analyze the context of the message and inputs prompts into a generative AI model to generate an appropriate reply. For example, in response to a message like "How are you doing?", the server analyzes the context from past conversation data and generates an appropriate reply.

[0528] Suggested replies

[0529] The server sends several generated reply options to the device. The device then uses JavaScript or Swift to display the reply options in a user-friendly interface. For example, it might suggest the following reply options to the user:

[0530] "fine and you?"

[0531] "I'm a little busy, but I'm managing somehow."

[0532] "I recently found a new hobby."

[0533] Send a reply

[0534] The user selects one of the displayed reply options. The device then sends the selected reply to the recipient in the messaging application. This is done using the messaging application's API.

[0535] This specific process allows users to leverage generative AI models and sentiment analysis to provide quick and appropriate responses. This enables efficient and effective conversations, especially in business-oriented situations or those requiring high-level communication skills.

[0536] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0537] Step 1: Data collection and preprocessing

[0538] The user grants permission for the messaging application to collect conversation history data. The device collects the conversation history data permitted by the user and sends it to the server. The server preprocesses the received conversation data, performing noise reduction and data cleaning. Specifically, it uses the Pandas library to remove unnecessary line breaks and emojis and format the data.

[0539] Input: User-authorized conversation history data

[0540] Output: Pre-processed clean data

[0541] Step 2: Model Training

[0542] The server trains a generative AI model using pre-processed conversational data. Here, the Hugging Face Transformers library is used to train models such as BERT and GPT-3. The BERT model is also used for sentiment analysis, classifying the sentiment of each message. For example, for a message like "How are you?", conversations with similar contexts are extracted from past conversational data and used for training.

[0543] Input: Preprocessed conversation data

[0544] Output: Trained generative AI model

[0545] Step 3: Receiving real-time messages

[0546] When a user receives a new message in their messaging application, the device sends this new message to the server. The message includes the sender, recipient, and message content.

[0547] Input: New incoming message

[0548] Output: Message data sent to the server

[0549] Step 4: Contextual analysis and generation of reply candidates

[0550] The server performs contextual analysis on newly received messages. This analysis uses the spacy library to understand the meaning of the message. It then generates an appropriate prompt and inputs it into a generative AI model to produce multiple response options. For example, in response to the message "How are you?", based on past conversation data and sentiment analysis results, it generates response options such as: "I'm fine! How about you?", "I'm a little busy, but I'm managing," and "I recently found a new hobby."

[0551] Input: Newly received message

[0552] Output: Multiple response candidates generated

[0553] Step 5: Suggesting possible replies

[0554] The server sends the generated reply suggestions to the device. The device then displays the reply suggestions to the user using a simple and user-friendly interface based on JavaScript or Swift.

[0555] Input: Multiple generated reply suggestions

[0556] Output: Reply suggestions displayed to the user

[0557] Step 6: Select and send your reply.

[0558] The user selects the most suitable reply from the displayed options. The device then sends the selected reply to the recipient in the messaging application. This is done using the messaging application's API.

[0559] Input: User's selected reply

[0560] Output: Reply message sent to the recipient

[0561] (Application Example 1)

[0562] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0563] Traditional food delivery services often struggle to respond quickly and accurately to user messages regarding order confirmations, order changes, and complaints. This frequently results in a poor user experience, potentially leading to decreased service usage and satisfaction. Furthermore, in business environments requiring frequent communication, greater efficiency is needed.

[0564] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0565] In this invention, the server includes means for collecting past conversation data, means for preprocessing the collected conversation data, means for training a generative AI model using the preprocessed conversation data, means for analyzing received messages and generating multiple reply candidates using the generative AI model, means for sending the generated reply candidates to the user's terminal, means for sending the reply selected by the user, and means for suggesting appropriate reply candidates to order confirmation messages, order change request messages, or complaint handling messages when the user is using an online delivery service. This enables users to make quick and appropriate replies, improving the user experience and increasing the efficiency of the delivery service.

[0566] "Past conversation data" refers to records of communications a user has had in the past, and this includes various formats such as text messages, voice messages, and chat logs.

[0567] "Preprocessing" refers to a series of processes that remove noise from collected conversation data, convert it to the required data format, and prepare it for analysis.

[0568] A "generative AI model" is a model that uses AI algorithms to generate new text data, and can also generate appropriate answers and replies based on the user's past conversation data.

[0569] "Suggested replies" refer to multiple possible responses to an incoming message, allowing the user to choose from among them.

[0570] "Online delivery service" refers to any service that allows users to order food and goods via the internet and have them delivered to their homes or workplaces.

[0571] An "order confirmation message" refers to a message sent to a user after they have placed an order for a product or service, to confirm the details of that order.

[0572] An "order change request message" refers to a message sent when a user wants to make changes to an order they have already placed.

[0573] A "complaint response message" refers to a message sent by a user when they are dissatisfied with a product or service or have a problem with it, and appropriate action is taken based on this message.

[0574] The present invention provides a system that suggests appropriate responses to various messages received by a user while using an online delivery service. This system includes the user's terminal, a server, and a generative AI model.

[0575] Specific System Procedures

[0576] The server first collects past conversation data from the user's device. This data includes text messages and chat logs, and once collected, it undergoes preprocessing such as noise reduction. The preprocessed data is then used to train a generative AI model. By incorporating sentiment analysis techniques at this stage, it becomes possible to generate more natural and appropriate responses.

[0577] When a user receives a new message via the delivery service, the device sends the message to the server in real time. The server analyzes the received message and generates multiple reply options using a generative AI model. The generated reply options are sent to the device and displayed to the user. The user selects the appropriate reply from the displayed options, and the selected reply is sent to the recipient.

[0578] Hardware and software

[0579] The hardware used includes the user's smartphone or other device, as well as a server. The software includes scripts for data collection and preprocessing, generative AI models (e.g., OpenAI GPT-3), and sentiment analysis libraries (e.g., TextBlob).

[0580] As a concrete example, consider a case where a user receives an "order confirmation message" while using a food delivery service. The following response options are generated in response to this message.

[0581] Example of a prompt

[0582] Past conversation history: "Order confirmation: Yes, please." "Delivery delay: Excuse me, is it delayed?"

[0583] Current message: "We will arrive shortly."

[0584] Possible replies to generate:

[0585] 1. "Understood, thank you."

[0586] 2. "How much longer until we arrive?"

[0587] In this way, users can respond quickly and accurately, improving their experience with online delivery services. Furthermore, increased efficiency can be expected in the work environment.

[0588] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[0589] Step 1:

[0590] The user authorizes the collection of past conversation data through their device. The device, upon receiving permission, collects past chat logs and message history from LINE and other messaging applications. The collected data is then sent directly to the server. The input is past conversation data, and the output is the collected data. Specifically, this involves sending past message history to the server in text format.

[0591] Step 2:

[0592] The server preprocesses the conversation data received from the terminal. Preprocessing includes removing noise and filtering out unnecessary information from the data. The input is the collected data, and the output is the preprocessed, cleaned data. Specifically, this involves performing text analysis on the data and removing meaningless information and error messages.

[0593] Step 3:

[0594] The server trains a generative AI model using preprocessed data. Sentiment analysis techniques are also used to learn the emotional nuances of the data. The input is preprocessed data, and the output is the trained generative AI model. Specifically, this involves performing sentiment analysis using a Python library (e.g., TextBlob) and training a generative AI model such as OpenAI GPT-3.

[0595] Step 4:

[0596] When a user receives a new message via LINE or another messaging app, the device sends the message to a server. The server analyzes the context of the received message and generates multiple reply options from a generative AI model based on past conversation data. The input is the newly received message, and the output is the generated reply options. Specifically, this involves text analysis of the received message and outputting appropriate reply options.

[0597] Step 5:

[0598] The server sends the generated reply options to the user's terminal. The terminal displays the reply options to the user. The user selects the most appropriate reply from the displayed options. The input is the generated reply options, and the output is the reply selected by the user. Specifically, this involves displaying a list of reply options in the terminal's user interface and accepting the user's selection.

[0599] Step 6:

[0600] The user sends their selected reply from their device to the server, which then forwards that selected reply to the LINE recipient. The input is the user-selected reply, and the output is the actual reply message sent. Specifically, this involves sending the selected message to the recipient via the LINE API.

[0601] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[0602] System Overview

[0603] The present invention's system proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. In particular, by combining it with an emotion engine that recognizes the user's emotions, it enables the generation of more personalized replies. This system is implemented using the user's terminal, a server, a generative AI model, and an emotion engine.

[0604] Program Processing Overview

[0605] 1. Data collection and preprocessing

[0606] When a user allows the LINE app to collect conversation history data, the device collects past conversation data and sends it to a server. The server stores the received data in a database and performs preprocessing. Preprocessing includes text cleanup, noise reduction, and duplicate removal.

[0607] 2. Model training

[0608] The server uses pre-processed data to train a generative AI model. During this process, data based on the user's emotional state is also used, incorporating sentiment analysis into the training. This allows the AI ​​model to respond to fluctuations in the user's emotions.

[0609] 3. Processing real-time messages

[0610] When a user receives a new message on LINE, their device sends the message to the server. The server performs contextual analysis of the received message and uses an emotion engine to recognize the user's emotional state. This is then compared with past conversation data to analyze the appropriate context.

[0611] 4. Generating reply suggestions

[0612] The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the user's emotional state. This ensures that the response options also match the user's emotions.

[0613] 5. Suggestions for reply options

[0614] The server sends generated reply options to the terminal, which then suggests them to the user. The user selects the most suitable reply from the displayed options.

[0615] 6. Send a reply

[0616] When a user selects a reply option, the device sends that selected message to the recipient on LINE. This allows users to respond quickly and appropriately.

[0617] Specific example

[0618] The following is a specific example of this system.

[0619] Conversation example

[0620] (Message from a friend): "How have you been lately?"

[0621] When a user receives this message, the device sends it to the server in real time. The server analyzes the context of "How are you doing?" and uses an emotion engine to identify the user's current emotion. For example, if the server identifies the user's current emotion as "stressed," it matches this with past conversation data to generate potential replies such as the following:

[0622] "I'm a little busy, but I'm managing somehow."

[0623] "Work has been really tough lately, how about you?"

[0624] "I'm feeling a bit stressed, but thank you for listening."

[0625] The server sends these response options to the device, and the device suggests them to the user. If the user selects "I'm a bit busy, but I'm managing," the device sends that message to the friend.

[0626] In this way, users can easily provide appropriate replies, enabling efficient and effective conversations, especially in business-related situations or those requiring high communication skills. Furthermore, a key feature is the ability to provide more personalized replies by considering the user's emotional state.

[0627] The following describes the processing flow.

[0628] System Overview

[0629] The system of this invention proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. In particular, by combining it with an emotion engine that recognizes the user's emotions, it enables the generation of more personalized replies. This system is implemented using the user's terminal, a server, a generative AI model, and an emotion engine.

[0630] Program Processing Overview

[0631] Step 1:

[0632] The user grants the LINE app permission to collect conversation history data.

[0633] Step 2:

[0634] The device retrieves past conversation data from the LINE app.

[0635] Step 3:

[0636] The device sends the acquired conversation data to the server.

[0637] Step 4:

[0638] The server saves the received conversation data to a database.

[0639] Step 5:

[0640] The server preprocesses the conversation data. Preprocessing includes text cleanup, noise reduction, and duplicate removal.

[0641] Step 6:

[0642] The server uses pre-processed data to train generative AI models. It also incorporates sentiment analysis into the training process, using data based on the user's emotional state.

[0643] Step 7:

[0644] The user receives a new message on LINE.

[0645] Step 8:

[0646] The terminal sends the received message to the server.

[0647] Step 9:

[0648] The server analyzes incoming messages to understand their context. During this process, it also recognizes the user's emotional state using an emotion engine.

[0649] Step 10:

[0650] The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the user's emotional state.

[0651] Step 11:

[0652] The server sends the generated reply options to the terminal.

[0653] Step 12:

[0654] The device displays reply suggestions to the user.

[0655] Step 13:

[0656] The user selects the most suitable reply from the displayed options.

[0657] Step 14:

[0658] The device sends the reply message selected by the user to the recipient on LINE.

[0659] Specific example

[0660] Conversation example

[0661] (Message from a friend): "How have you been lately?"

[0662] Step 1:

[0663] The user receives this message.

[0664] Step 2:

[0665] The device sends the message to the server in real time.

[0666] Step 3:

[0667] The server analyzes the context of the question "How are you doing lately?" and uses its emotion engine to identify the user's current emotional state. For example, it might identify the user's current emotion as "feeling stressed."

[0668] Step 4:

[0669] The server generates response options based on past conversation data and emotional states analyzed by the emotion engine, such as the following:

[0670] "I'm a little busy, but I'm managing somehow."

[0671] "Work has been really tough lately, how about you?"

[0672] "I'm feeling a bit stressed, but thank you for listening."

[0673] Step 5:

[0674] The server sends these suggested replies to the terminal.

[0675] Step 6:

[0676] The device suggests and displays reply options to the user.

[0677] Step 7:

[0678] The user selects, "I'm a little busy, but I'm managing."

[0679] Step 8:

[0680] The device sends the selected message to the friend.

[0681] This system not only automatically suggests appropriate replies but also takes into account the user's emotional state, enabling more personalized communication. Its effectiveness is particularly evident in business settings and situations requiring advanced communication skills.

[0682] (Example 2)

[0683] Next, we will describe Example 2. 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".

[0684] Traditional messaging applications have struggled to generate appropriate replies to received messages. In particular, existing technologies were insufficient to generate personalized replies that took into account the user's emotional state. As a result, users had to manually compose replies, hindering efficient and effective communication.

[0685] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[0686] In this invention, the server includes means for collecting past conversation data, means for preprocessing the collected conversation data, means for training a generative AI model using the preprocessed conversation data, means for analyzing an incoming message and generating multiple reply candidates using the generative AI model, means for sending the generated reply candidates to the user's terminal, means for sending the reply selected by the user, means for using an emotion engine to recognize the user's emotional state, and means for generating reply candidates based on contextual information of the incoming message and the user's emotional state. This enables the generation of personalized replies that take the user's emotions into consideration, allowing for efficient, rapid, and appropriate replies.

[0687] "Past conversation data" refers to information that includes the history of messages a user has sent in the past.

[0688] "Preprocessing" refers to the process of removing noise from received data, cleaning up text, and removing duplicates.

[0689] A "generative AI model" is an artificial intelligence model that is trained based on collected data and generates appropriate outputs (responses) in response to given inputs.

[0690] An "emotion engine" is a tool or software used to analyze and identify a user's emotional state.

[0691] "Contextual information" refers to information that includes meaning derived from the content of a message and the flow of past conversations.

[0692] "Suggested replies" refer to multiple response options proposed by the generative AI model.

[0693] A "server" is a computer system used for tasks such as data collection, preprocessing, analysis, and training of AI models.

[0694] A "user's device" refers to a device used by the user (such as a smartphone, tablet, or PC) that has the function of receiving and displaying reply suggestions from the server.

[0695] A "messaging application" is a general term for application software that allows users to send and receive text messages.

[0696] Modes for carrying out the invention

[0697] The present invention provides a system that suggests appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. In particular, this system enables the generation of more personalized replies by incorporating an emotion engine that recognizes the user's emotions. The embodiments for carrying out this invention are described in detail below.

[0698] The system is implemented using user devices, servers, generative AI models, and emotion engines. Specifically, user devices include smartphones, tablets, and PCs. High-performance computer systems (e.g., server machines with NVIDIA GPUs) are used for the servers. Database management systems such as MySQL and PostgreSQL are used for the databases.

[0699] Data collection and preprocessing

[0700] When a user allows the LINE app to collect conversation history data, the device collects past conversation data from the LINE app and sends it to a server. The server stores the received conversation data in a database. Subsequently, preprocessing is performed, including text cleanup, noise reduction, and duplicate removal. This preprocessing improves the quality of the data, making it suitable for training generative AI models.

[0701] Model training

[0702] The server uses pre-processed data and an emotion engine to train a generative AI model (e.g., GPT-3 or BERT). This training process also utilizes data based on the user's emotional state. Emotion analysis leverages an emotion engine (e.g., IBM Watson, Azure Text Analytics). In this way, the generative AI model becomes capable of responding to fluctuations in the user's emotions.

[0703] Processing real-time messages

[0704] When a user receives a new message on LINE, their device sends the message to the server in real time. The server analyzes the context of the received message using natural language processing techniques (e.g., spaCy or NLTK). Furthermore, it uses an emotion engine to recognize the user's emotional state.

[0705] Generating reply suggestions

[0706] The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the user's emotional state. The generated response options are appropriate to the context and the user's emotional state.

[0707] Suggestions and selection of reply options

[0708] The server sends generated reply options to the terminal, which then presents them to the user. The user selects the best option from the displayed multiple reply options.

[0709] Send a reply

[0710] When a user selects a reply option, the device sends that selected message to the recipient on LINE. This allows users to respond quickly and efficiently to appropriate messages.

[0711] Specific example

[0712] The following is a specific example of this system.

[0713] (Message from a friend): "How have you been lately?"

[0714] When a user receives this message, the device sends it to the server in real time. The server analyzes the context of "How are you doing?" and uses an emotion engine to identify the user's current emotion. For example, if the server identifies the user's current emotion as "stressed," it matches this with past conversation data to generate potential replies such as the following:

[0715] "I'm a little busy, but I'm managing somehow."

[0716] "Work has been really tough lately, how about you?"

[0717] "I'm feeling a bit stressed, but thank you for listening."

[0718] The server sends these reply options to the device, which then suggests them to the user. If the user selects "I'm a bit busy, but I'm managing," the device sends that message to their friend. This method allows users to reply quickly and appropriately.

[0719] The flow of the specific processing in Example 2 will be explained using Figure 13.

[0720] Step 1:

[0721] The user grants permission for the LINE app to collect conversation history data.

[0722] Input: User permission settings

[0723] Output: Instructions for collecting conversation history data

[0724] Specific action: The user grants permission for the collection of conversation data in the LINE settings screen.

[0725] Step 2:

[0726] The device retrieves past conversation data from the LINE app and sends that data to the server.

[0727] Input: Conversation history data

[0728] Output: Sending conversation data to the server

[0729] Specific operation: A background process on the device extracts LINE conversation history data and sends it to the server via a secure protocol (e.g., HTTPS).

[0730] Step 3:

[0731] The server saves the received conversation data to the database.

[0732] Input: Conversation data

[0733] Output: Data stored in the database

[0734] Specific operation: The server receives the conversation data and inserts it into the database in a specific format (e.g., JSON format).

[0735] Step 4:

[0736] The server performs preprocessing on the data it has stored.

[0737] Input: Conversation data stored in the database

[0738] Output: Preprocessed data

[0739] Specific actions: The server performs the following preprocessing:

[0740] Text cleanup: Remove unnecessary symbols and spaces.

[0741] Noise reduction: Removes spam messages and meaningless data.

[0742] Remove duplicates: If the same message is recorded multiple times, it will be merged into one.

[0743] Step 5:

[0744] The server uses pre-processed data to train the generated AI model.

[0745] Input: Preprocessed data

[0746] Output: Trained Generative AI Model

[0747] Specific operation: The server uses high-performance computing resources (e.g., NVIDIA GPU) to train a generative AI model (e.g., GPT-3). During this process, sentiment analysis data is also incorporated into the model using an emotion engine.

[0748] Step 6:

[0749] The user receives a new message on LINE.

[0750] Input: New message

[0751] Output: New message displayed on the terminal

[0752] Specific action: The LINE app receives a new message and displays it on the device.

[0753] Step 7:

[0754] The terminal sends the received message to the server.

[0755] Input: Received message

[0756] Output: Sending a message to the server

[0757] Specific operation: The device forwards messages received from the LINE app to the server, and the server receives them.

[0758] Step 8:

[0759] The server analyzes the context of the received message and uses an emotion engine to recognize the user's emotional state.

[0760] Input: Received message

[0761] Output: Contextual information, emotional state

[0762] Specific operation: The server uses natural language processing technology (e.g., spaCy) to analyze the context of the message and uses an emotion engine to recognize the user's emotional state.

[0763] Step 9:

[0764] Based on the analyzed contextual information and the user's emotional state, the server uses a generative AI model to generate multiple response options.

[0765] Input: Contextual information, emotional state, trained generative AI model

[0766] Output: Reply candidates

[0767] Specific operation: The server inputs a prompt message into the AI ​​model, which then generates response candidates that take into account contextual information and emotional state.

[0768] Step 10:

[0769] The server sends the generated response options to the terminal, which then suggests them to the user.

[0770] Input: Reply suggestions

[0771] Output: Sending reply suggestions to the terminal, presenting them to the user.

[0772] Specific operation: The server sends suggested replies to the terminal, and the terminal displays them to the user.

[0773] Step 11:

[0774] When a user selects a reply option, the device sends that selected message to the recipient via the LINE app through the server.

[0775] Input: Selected reply options

[0776] Output: Message sent to the recipient via the LINE app

[0777] Specific operation: The user selects a reply option on the screen, and the device sends that message to the recipient via the LINE app through the server.

[0778] (Application Example 2)

[0779] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server," and the smart glasses 214 will be referred to as the "terminal."

[0780] In physical stores, effective communication between staff and customers is crucial for improving customer satisfaction. However, responding quickly and appropriately to customer requests and questions is not easy. In particular, being able to respond in a way that is sensitive to the customer's emotional state would further enhance the customer experience, but there is a lack of effective tools to achieve this.

[0781] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting past dialogue data, means for pre-processing the collected dialogue data, means for training a generative AI model using the pre-processed dialogue data, means for analyzing received messages and generating a plurality of response candidates using the generative AI model, means for transmitting the generated response candidates to the user's display device, means for transmitting the response selected by the user, and means for comparing and analyzing past dialogue history using an emotion engine that identifies the customer's emotional state. As a result, staff can provide quick and appropriate responses to customer requests and questions, and in particular, personalized responses that are in line with the customer's emotions become possible.

[0782] "Dialogue data" refers to a record of conversations that take place between a user and another person.

[0783] "Preprocessing" is the process of removing unnecessary information from collected data and converting it into a usable format.

[0784] A "generative AI model" is an artificial intelligence model that learns patterns based on collected data and performs specific tasks.

[0785] "Received messages" are messages that a user receives from other people.

[0786] "Response candidates" are multiple reply options proposed by a generative AI model.

[0787] A "display device" is a device that a user uses to view response options and other information.

[0788] An "emotion engine" is a system that analyzes and identifies emotions from text, audio, and other sources.

[0789] "Contextual analysis" is the process of understanding the content and background of a message and extracting information to generate an appropriate response.

[0790] The "Concierge Assistant" system of this invention is a tool that supports communication between staff and customers in physical stores. This system operates in the following steps:

[0791] Hardware and software to be used

[0792] Hardware: Tablets for customer service (e.g., iPad)

[0793] software:

[0794] Message management application

[0795] Emotion engine (e.g., IBM Watson Tone Analyzer)

[0796] Generative AI models (e.g., GPT-4)

[0797] System Implementation Procedures

[0798] 1. Data collection and preprocessing

[0799] The server collects past conversation data from customer service tablets. The collected data is pre-processed on the server. Pre-processing includes text cleanup, noise reduction, and duplicate removal.

[0800] 2. Training of generative AI models

[0801] The server uses pre-processed data to train a generative AI model. During this process, an emotion engine is used to consider the customer's emotional state, resulting in a model that can adapt to individual customer emotional fluctuations.

[0802] 3. Processing real-time messages

[0803] When a customer service tablet receives a new question or request from a customer, it sends the content to the server. The server contextualizes the received message and uses an emotion engine to identify the customer's emotional state.

[0804] 4. Generating response candidates

[0805] The server uses a generative AI model to generate multiple response candidates based on the analyzed contextual information and the customer's emotional state.

[0806] 5. Proposal of response candidates

[0807] The server sends the generated response options to a customer service tablet, which then displays them to the staff. The staff then selects the most appropriate response option.

[0808] 6. Sending a response

[0809] When a staff member selects a response option, the customer service tablet displays that message to the customer.

[0810] Specific example

[0811] For example, if a customer asks, "How do I use this product?", the customer service tablet sends the question to the server. The server analyzes the context of "How do I use this product?" and identifies the customer's emotion through an emotion engine (e.g., IBM Watson Tone Analyzer). If the customer is identified as "interested," a generative AI model (e.g., GPT-4) generates response candidates such as the following.

[0812] "It would be best if you press this button first, and then do this next."

[0813] "We also have an easy-to-understand demo video; would you like to watch it?"

[0814] "If you'd like to try it out, we can help you."

[0815] If a staff member selects "We also have an easy-to-understand demo video; would you like to watch it?" from the suggested responses, the customer service tablet will display this message to the customer.

[0816] Example of a prompt

[0817] "Based on past interaction history and the customer's emotional state, please propose the most appropriate response."

[0818] This system will allow in-store staff to provide quick and personalized service to customers, which is expected to improve customer satisfaction.

[0819] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[0820] Step 1:

[0821] Data collection and preprocessing

[0822] Input: Past conversation data collected by the user using a customer service tablet.

[0823] Processing: The terminal sends past conversation data to the server. The server stores the received data in a database and performs preprocessing such as data cleanup, noise reduction, and duplicate data removal.

[0824] Output: Pre-processed, clean dialogue data.

[0825] Step 2:

[0826] Training of generative AI models

[0827] Input: Preprocessed dialogue data and sentiment analysis data from the sentiment engine.

[0828] Processing: The server trains a generative AI model using pre-processed data. During this process, it incorporates emotional information from the dialogue data using an emotion engine, building a model that can respond to changes in the user's emotions.

[0829] Output: A pre-trained generative AI model capable of responding to the user's emotional state.

[0830] Step 3:

[0831] Processing real-time messages

[0832] Input: A message from a new customer that the user receives on their customer service tablet.

[0833] Processing: The terminal sends the received message to the server. The server analyzes the context of the received message and uses an emotion engine to identify the customer's emotional state.

[0834] Output: Analyzed contextual information and emotional state information.

[0835] Step 4:

[0836] Generating response candidates

[0837] Input: Analyzed contextual information and emotional state information.

[0838] Processing: The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the customer's emotional state.

[0839] Output: Multiple response candidates generated by the server.

[0840] Step 5:

[0841] Suggested response options

[0842] Input: Multiple response candidates generated by the server.

[0843] Processing: The server sends the generated response options to the customer service tablet, which then presents them to the staff. The staff selects the most appropriate response from the screen.

[0844] Output: Staff-selected response.

[0845] Step 6:

[0846] Sending a response

[0847] Input: Response selected by the staff.

[0848] Processing: The terminal sends the selected response message back to the customer.

[0849] Output: Response message sent to the customer.

[0850] 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 user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[0851] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0852] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and the specific processing may also be performed by the smart glasses 214.

[0853] [Third Embodiment]

[0854] Figure 5 shows an example of the configuration of the data processing system 310 according to the third embodiment.

[0855] As shown in Figure 5, the data processing system 310 includes a data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0856] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0857] The headset terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and display 343 are also connected to the bus 52.

[0858] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[0859] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[0860] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[0861] Figure 6 shows an example of the main functions of the data processing device 12 and the headset terminal 314. As shown in Figure 6, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[0862] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[0863] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0864] In the headset terminal 314, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[0865] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the headset terminal 314 will be referred to as the "terminal".

[0866] System Overview

[0867] The system of this invention proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. This system is implemented using the user's terminal, a server, and a generative AI model.

[0868] Program Processing Overview

[0869] 1. Data collection and preprocessing

[0870] When a user allows the LINE app to collect conversation history data, the device collects past conversation data and sends it to the server. The server preprocesses the received data, performing noise reduction and data cleaning.

[0871] 2. Model training

[0872] The server uses pre-processed data to train generative AI models. Techniques such as sentiment analysis are used here to enable the generation of more natural and optimal responses.

[0873] 3. Processing real-time messages

[0874] When a user receives a new message on LINE, their device sends the message to a server. The server analyzes the context of the received message and generates multiple reply options from a generative AI model based on past conversation data.

[0875] 4. Suggestions for reply options

[0876] The server sends the generated reply options to the terminal, which then displays them to the user. The user selects the most appropriate reply from the displayed options.

[0877] 5. Send a reply

[0878] When a user selects a reply option, the device sends that selected message to the recipient on LINE. This allows users to respond quickly and appropriately.

[0879] Specific example

[0880] The following is a specific example of this system.

[0881] Conversation example

[0882] (Message from a friend): "How have you been lately?"

[0883] When a user receives this message, the device sends it to the server in real time. The server analyzes the context of "How are you doing lately?" and generates possible replies based on past conversation data, such as the following:

[0884] "fine and you?"

[0885] "I'm a little busy, but I'm managing somehow."

[0886] "I recently found a new hobby."

[0887] The server sends these reply options to the device, and the device suggests them to the user. If the user selects "I'm fine! How about you?", the device sends that message to the friend.

[0888] In this way, users can easily provide appropriate responses, enabling efficient and effective conversations, especially in business-related situations or those requiring high communication skills.

[0889] The following describes the processing flow.

[0890] Step 1:

[0891] The user grants the LINE app permission to collect conversation history data.

[0892] Step 2:

[0893] The device retrieves past conversation data from the LINE app.

[0894] Step 3:

[0895] The device sends the acquired conversation data to the server.

[0896] Step 4:

[0897] The server saves the received conversation data to a database.

[0898] Step 5:

[0899] The server preprocesses the conversation data. This preprocessing includes text cleanup, noise reduction, and duplicate removal.

[0900] Step 6:

[0901] The server uses pre-processed data to train generative AI models. Here, techniques such as sentiment analysis are used in conjunction to improve the accuracy of the training.

[0902] Step 7:

[0903] The user receives a new message on LINE.

[0904] Step 8:

[0905] The terminal sends the received message to the server.

[0906] Step 9:

[0907] The server analyzes incoming messages to understand their context. In doing so, it compares them with past conversation data to determine the appropriate context.

[0908] Step 10:

[0909] Based on the analyzed contextual information, the server uses a generative AI model to generate multiple response options.

[0910] Step 11:

[0911] The server sends the generated reply options to the terminal.

[0912] Step 12:

[0913] The device displays reply suggestions to the user.

[0914] Step 13:

[0915] The user selects the most suitable reply from the displayed options.

[0916] Step 14:

[0917] The device sends the reply message selected by the user to the recipient on LINE.

[0918] This series of steps allows users to efficiently provide appropriate responses.

[0919] (Example 1)

[0920] Next, we will describe Example 1. 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."

[0921] Modern messaging applications lack sufficient means for users to respond quickly and appropriately. In particular, few existing systems generate responses that consider context based on past conversation data, requiring users to manually formulate replies, which can reduce communication efficiency. Furthermore, technologies that incorporate sentiment analysis to achieve more natural conversations are also insufficient.

[0922] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[0923] In this invention, the server includes means for allowing the user to collect conversation history data, means for collecting past conversation data from the terminal and sending it to the server, means for preprocessing the received conversation data to remove noise and clean the data, means for training a generative AI model using the preprocessed conversation data, means for training the generative AI model by incorporating sentiment analysis, means for receiving a new message, means for sending the received message to the server, means for performing contextual analysis of the message and inputting a prompt sentence into the generative AI model to generate multiple reply candidates, means for sending the generated reply candidates to the user's terminal, means for the user to select from the displayed reply candidates, and means for sending the selected reply to the recipient of the messaging application. This enables the user to make quick and appropriate replies.

[0924] A "user" refers to a person who uses a messaging application to have a conversation.

[0925] "Device" refers to electronic devices such as smartphones and computers used by users.

[0926] A "server" refers to a computing system that receives data sent from a terminal and processes it.

[0927] "Conversation history data" refers to records of conversations that have taken place on messaging applications in the past.

[0928] "Preprocessing" refers to the process of preparing received data into an analyzable format. Specifically, this includes noise reduction and data cleaning.

[0929] A "generative AI model" refers to a model that uses artificial intelligence to generate output in response to a specific input. An example is a text generation model.

[0930] "Sentiment analysis" refers to the technique of identifying and classifying the emotions contained within text.

[0931] "New messages" refers to the most recent messages a user receives in their messaging application.

[0932] "Contextual analysis" refers to analytical methods used to understand the meaning and underlying intentions of text.

[0933] A "prompt" refers to the document or question that is input into a generative AI model.

[0934] "Suggested replies" refers to multiple response options generated by the server using a generated AI model.

[0935] A "messaging application" refers to software that users use to send and receive messages.

[0936] Modes for carrying out the invention

[0937] This invention relates to a system that suggests appropriate replies to messages received by a user in a messaging application, based on past conversation data. This system is implemented using the user's terminal, a server, and a generative AI model.

[0938] Data collection and preprocessing

[0939] When a user allows a messaging application to collect conversation history data, the device collects past conversation data and sends it to the server. The server preprocesses the received conversation data, performing noise reduction and data cleaning. Specifically, it uses libraries such as Pandas and NumPy to clean up the data. For example, it removes unnecessary information such as newline characters and emojis.

[0940] Model training

[0941] The server trains generative AI models using preprocessed data. Specifically, it uses the Hugging Face Transformers library to train generative AI models such as BERT and GPT-3. In addition, it uses the BERT model to label emotions within text for sentiment analysis. For example, it trains the model based on past conversation data regarding the message "How are you doing lately?".

[0942] Processing real-time messages

[0943] When a user receives a new message, the device sends it to the server. The server uses the spacy library to analyze the context of the message and inputs prompts into a generative AI model to generate an appropriate reply. For example, in response to a message like "How are you doing?", the server analyzes the context from past conversation data and generates an appropriate reply.

[0944] Suggested replies

[0945] The server sends several generated reply options to the device. The device then uses JavaScript or Swift to display the reply options in a user-friendly interface. For example, it might suggest the following reply options to the user:

[0946] "fine and you?"

[0947] "I'm a little busy, but I'm managing somehow."

[0948] "I recently found a new hobby."

[0949] Send a reply

[0950] The user selects one of the displayed reply options. The device then sends the selected reply to the recipient in the messaging application. This is done using the messaging application's API.

[0951] This specific process allows users to leverage generative AI models and sentiment analysis to provide quick and appropriate responses. This enables efficient and effective conversations, especially in business-oriented situations or those requiring high-level communication skills.

[0952] The flow of the specific processing in Example 1 will be explained using Figure 11.

[0953] Step 1: Data collection and preprocessing

[0954] The user grants permission for the messaging application to collect conversation history data. The device collects the conversation history data permitted by the user and sends it to the server. The server preprocesses the received conversation data, performing noise reduction and data cleaning. Specifically, it uses the Pandas library to remove unnecessary line breaks and emojis and format the data.

[0955] Input: User-authorized conversation history data

[0956] Output: Pre-processed clean data

[0957] Step 2: Model Training

[0958] The server trains a generative AI model using pre-processed conversational data. Here, the Hugging Face Transformers library is used to train models such as BERT and GPT-3. The BERT model is also used for sentiment analysis, classifying the sentiment of each message. For example, for a message like "How are you?", conversations with similar contexts are extracted from past conversational data and used for training.

[0959] Input: Preprocessed conversation data

[0960] Output: Trained generative AI model

[0961] Step 3: Receiving real-time messages

[0962] When a user receives a new message in their messaging application, the device sends this new message to the server. The message includes the sender, recipient, and message content.

[0963] Input: New incoming message

[0964] Output: Message data sent to the server

[0965] Step 4: Contextual analysis and generation of reply candidates

[0966] The server performs contextual analysis on newly received messages. This analysis uses the spacy library to understand the meaning of the message. It then generates an appropriate prompt and inputs it into a generative AI model to produce multiple response options. For example, in response to the message "How are you?", based on past conversation data and sentiment analysis results, it generates response options such as: "I'm fine! How about you?", "I'm a little busy, but I'm managing," and "I recently found a new hobby."

[0967] Input: Newly received message

[0968] Output: Multiple response candidates generated

[0969] Step 5: Suggesting possible replies

[0970] The server sends the generated reply suggestions to the device. The device then displays the reply suggestions to the user using a simple and user-friendly interface based on JavaScript or Swift.

[0971] Input: Multiple generated reply suggestions

[0972] Output: Reply suggestions displayed to the user

[0973] Step 6: Select and send your reply.

[0974] The user selects the most suitable reply from the displayed options. The device then sends the selected reply to the recipient in the messaging application. This is done using the messaging application's API.

[0975] Input: User's selected reply

[0976] Output: Reply message sent to the recipient

[0977] (Application Example 1)

[0978] Next, we will explain Application Example 1. In the following explanation, 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."

[0979] Traditional food delivery services often struggle to respond quickly and accurately to user messages regarding order confirmations, order changes, and complaints. This frequently results in a poor user experience, potentially leading to decreased service usage and satisfaction. Furthermore, in business environments requiring frequent communication, greater efficiency is needed.

[0980] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[0981] In this invention, the server includes means for collecting past conversation data, means for preprocessing the collected conversation data, means for training a generative AI model using the preprocessed conversation data, means for analyzing received messages and generating multiple reply candidates using the generative AI model, means for sending the generated reply candidates to the user's terminal, means for sending the reply selected by the user, and means for suggesting appropriate reply candidates to order confirmation messages, order change request messages, or complaint handling messages when the user is using an online delivery service. This enables users to make quick and appropriate replies, improving the user experience and increasing the efficiency of the delivery service.

[0982] "Past conversation data" refers to records of communications a user has had in the past, and this includes various formats such as text messages, voice messages, and chat logs.

[0983] "Preprocessing" refers to a series of processes that remove noise from collected conversation data, convert it to the required data format, and prepare it for analysis.

[0984] A "generative AI model" is a model that uses AI algorithms to generate new text data, and can also generate appropriate answers and replies based on the user's past conversation data.

[0985] "Suggested replies" refer to multiple possible responses to an incoming message, allowing the user to choose from among them.

[0986] "Online delivery service" refers to any service that allows users to order food and goods via the internet and have them delivered to their homes or workplaces.

[0987] An "order confirmation message" refers to a message sent to a user after they have placed an order for a product or service, to confirm the details of that order.

[0988] An "order change request message" refers to a message sent when a user wants to make changes to an order they have already placed.

[0989] A "complaint response message" refers to a message sent by a user when they are dissatisfied with a product or service or have a problem with it, and appropriate action is taken based on this message.

[0990] The present invention provides a system that suggests appropriate responses to various messages received by a user while using an online delivery service. This system includes the user's terminal, a server, and a generative AI model.

[0991] Specific System Procedures

[0992] The server first collects past conversation data from the user's device. This data includes text messages and chat logs, and once collected, it undergoes preprocessing such as noise reduction. The preprocessed data is then used to train a generative AI model. By incorporating sentiment analysis techniques at this stage, it becomes possible to generate more natural and appropriate responses.

[0993] When a user receives a new message via the delivery service, the device sends the message to the server in real time. The server analyzes the received message and generates multiple reply options using a generative AI model. The generated reply options are sent to the device and displayed to the user. The user selects the appropriate reply from the displayed options, and the selected reply is sent to the recipient.

[0994] Hardware and software

[0995] The hardware used includes the user's smartphone or other device, as well as a server. The software includes scripts for data collection and preprocessing, generative AI models (e.g., OpenAI GPT-3), and sentiment analysis libraries (e.g., TextBlob).

[0996] As a concrete example, consider a case where a user receives an "order confirmation message" while using a food delivery service. The following response options are generated in response to this message.

[0997] Example of a prompt

[0998] Past conversation history: "Order confirmation: Yes, please." "Delivery delay: Excuse me, is it delayed?"

[0999] Current message: "We will arrive shortly."

[1000] Possible replies to generate:

[1001] 1. "Understood, thank you."

[1002] 2. "How much longer until we arrive?"

[1003] In this way, users can respond quickly and accurately, improving their experience with online delivery services. Furthermore, increased efficiency can be expected in the work environment.

[1004] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[1005] Step 1:

[1006] The user authorizes the collection of past conversation data through their device. The device, upon receiving permission, collects past chat logs and message history from LINE and other messaging applications. The collected data is then sent directly to the server. The input is past conversation data, and the output is the collected data. Specifically, this involves sending past message history to the server in text format.

[1007] Step 2:

[1008] The server preprocesses the conversation data received from the terminal. Preprocessing includes removing noise and filtering out unnecessary information from the data. The input is the collected data, and the output is the preprocessed, cleaned data. Specifically, this involves performing text analysis on the data and removing meaningless information and error messages.

[1009] Step 3:

[1010] The server trains a generative AI model using preprocessed data. Sentiment analysis techniques are also used to learn the emotional nuances of the data. The input is preprocessed data, and the output is the trained generative AI model. Specifically, this involves performing sentiment analysis using a Python library (e.g., TextBlob) and training a generative AI model such as OpenAI GPT-3.

[1011] Step 4:

[1012] When a user receives a new message via LINE or another messaging app, the device sends the message to a server. The server analyzes the context of the received message and generates multiple reply options from a generative AI model based on past conversation data. The input is the newly received message, and the output is the generated reply options. Specifically, this involves text analysis of the received message and outputting appropriate reply options.

[1013] Step 5:

[1014] The server sends the generated reply options to the user's terminal. The terminal displays the reply options to the user. The user selects the most appropriate reply from the displayed options. The input is the generated reply options, and the output is the reply selected by the user. Specifically, this involves displaying a list of reply options in the terminal's user interface and accepting the user's selection.

[1015] Step 6:

[1016] The user sends their selected reply from their device to the server, which then forwards that selected reply to the LINE recipient. The input is the user-selected reply, and the output is the actual reply message sent. Specifically, this involves sending the selected message to the recipient via the LINE API.

[1017] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[1018] System Overview

[1019] The present invention's system proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. In particular, by combining it with an emotion engine that recognizes the user's emotions, it enables the generation of more personalized replies. This system is implemented using the user's terminal, a server, a generative AI model, and an emotion engine.

[1020] Program Processing Overview

[1021] 1. Data collection and preprocessing

[1022] When a user allows the LINE app to collect conversation history data, the device collects past conversation data and sends it to a server. The server stores the received data in a database and performs preprocessing. Preprocessing includes text cleanup, noise reduction, and duplicate removal.

[1023] 2. Model training

[1024] The server uses pre-processed data to train a generative AI model. During this process, data based on the user's emotional state is also used, incorporating sentiment analysis into the training. This allows the AI ​​model to respond to fluctuations in the user's emotions.

[1025] 3. Processing real-time messages

[1026] When a user receives a new message on LINE, their device sends the message to the server. The server performs contextual analysis of the received message and uses an emotion engine to recognize the user's emotional state. This is then compared with past conversation data to analyze the appropriate context.

[1027] 4. Generating reply suggestions

[1028] The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the user's emotional state. This ensures that the response options also match the user's emotions.

[1029] 5. Suggestions for reply options

[1030] The server sends generated reply options to the terminal, which then suggests them to the user. The user selects the most suitable reply from the displayed options.

[1031] 6. Send a reply

[1032] When a user selects a reply option, the device sends that selected message to the recipient on LINE. This allows users to respond quickly and appropriately.

[1033] Specific example

[1034] The following is a specific example of this system.

[1035] Conversation example

[1036] (Message from a friend): "How have you been lately?"

[1037] When a user receives this message, the device sends it to the server in real time. The server analyzes the context of "How are you doing?" and uses an emotion engine to identify the user's current emotion. For example, if the server identifies the user's current emotion as "stressed," it matches this with past conversation data to generate potential replies such as the following:

[1038] "I'm a little busy, but I'm managing somehow."

[1039] "Work has been really tough lately, how about you?"

[1040] "I'm feeling a bit stressed, but thank you for listening."

[1041] The server sends these response options to the device, and the device suggests them to the user. If the user selects "I'm a bit busy, but I'm managing," the device sends that message to the friend.

[1042] In this way, users can easily provide appropriate replies, enabling efficient and effective conversations, especially in business-related situations or those requiring high communication skills. Furthermore, a key feature is the ability to provide more personalized replies by considering the user's emotional state.

[1043] The following describes the processing flow.

[1044] System Overview

[1045] The system of this invention proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. In particular, by combining it with an emotion engine that recognizes the user's emotions, it enables the generation of more personalized replies. This system is implemented using the user's terminal, a server, a generative AI model, and an emotion engine.

[1046] Program Processing Overview

[1047] Step 1:

[1048] The user grants the LINE app permission to collect conversation history data.

[1049] Step 2:

[1050] The device retrieves past conversation data from the LINE app.

[1051] Step 3:

[1052] The device sends the acquired conversation data to the server.

[1053] Step 4:

[1054] The server saves the received conversation data to a database.

[1055] Step 5:

[1056] The server preprocesses the conversation data. Preprocessing includes text cleanup, noise reduction, and duplicate removal.

[1057] Step 6:

[1058] The server uses pre-processed data to train generative AI models. It also incorporates sentiment analysis into the training process, using data based on the user's emotional state.

[1059] Step 7:

[1060] The user receives a new message on LINE.

[1061] Step 8:

[1062] The terminal sends the received message to the server.

[1063] Step 9:

[1064] The server analyzes incoming messages to understand their context. During this process, it also recognizes the user's emotional state using an emotion engine.

[1065] Step 10:

[1066] The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the user's emotional state.

[1067] Step 11:

[1068] The server sends the generated reply options to the terminal.

[1069] Step 12:

[1070] The device displays reply suggestions to the user.

[1071] Step 13:

[1072] The user selects the most suitable reply from the displayed options.

[1073] Step 14:

[1074] The device sends the reply message selected by the user to the recipient on LINE.

[1075] Specific example

[1076] Conversation example

[1077] (Message from a friend): "How have you been lately?"

[1078] Step 1:

[1079] The user receives this message.

[1080] Step 2:

[1081] The device sends the message to the server in real time.

[1082] Step 3:

[1083] The server analyzes the context of the question "How are you doing lately?" and uses its emotion engine to identify the user's current emotional state. For example, it might identify the user's current emotion as "feeling stressed."

[1084] Step 4:

[1085] The server generates response options based on past conversation data and emotional states analyzed by the emotion engine, such as the following:

[1086] "I'm a little busy, but I'm managing somehow."

[1087] "Work has been really tough lately, how about you?"

[1088] "I'm feeling a bit stressed, but thank you for listening."

[1089] Step 5:

[1090] The server sends these suggested replies to the terminal.

[1091] Step 6:

[1092] The device suggests and displays reply options to the user.

[1093] Step 7:

[1094] The user selects, "I'm a little busy, but I'm managing."

[1095] Step 8:

[1096] The device sends the selected message to the friend.

[1097] This system not only automatically suggests appropriate replies but also takes into account the user's emotional state, enabling more personalized communication. Its effectiveness is particularly evident in business settings and situations requiring advanced communication skills.

[1098] (Example 2)

[1099] Next, we will describe Example 2. 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."

[1100] Traditional messaging applications have struggled to generate appropriate replies to received messages. In particular, existing technologies were insufficient to generate personalized replies that took into account the user's emotional state. As a result, users had to manually compose replies, hindering efficient and effective communication.

[1101] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[1102] In this invention, the server includes means for collecting past conversation data, means for preprocessing the collected conversation data, means for training a generative AI model using the preprocessed conversation data, means for analyzing an incoming message and generating multiple reply candidates using the generative AI model, means for sending the generated reply candidates to the user's terminal, means for sending the reply selected by the user, means for using an emotion engine to recognize the user's emotional state, and means for generating reply candidates based on contextual information of the incoming message and the user's emotional state. This enables the generation of personalized replies that take the user's emotions into consideration, allowing for efficient, rapid, and appropriate replies.

[1103] "Past conversation data" refers to information that includes the history of messages a user has sent in the past.

[1104] "Preprocessing" refers to the process of removing noise from received data, cleaning up text, and removing duplicates.

[1105] A "generative AI model" is an artificial intelligence model that is trained based on collected data and generates appropriate outputs (responses) in response to given inputs.

[1106] An "emotion engine" is a tool or software used to analyze and identify a user's emotional state.

[1107] "Contextual information" refers to information that includes meaning derived from the content of a message and the flow of past conversations.

[1108] "Suggested replies" refer to multiple response options proposed by the generative AI model.

[1109] A "server" is a computer system used for tasks such as data collection, preprocessing, analysis, and training of AI models.

[1110] A "user's device" refers to a device used by the user (such as a smartphone, tablet, or PC) that has the function of receiving and displaying reply suggestions from the server.

[1111] A "messaging application" is a general term for application software that allows users to send and receive text messages.

[1112] Modes for carrying out the invention

[1113] The present invention provides a system that suggests appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. In particular, this system enables the generation of more personalized replies by incorporating an emotion engine that recognizes the user's emotions. The embodiments for carrying out this invention are described in detail below.

[1114] The system is implemented using user devices, servers, generative AI models, and emotion engines. Specifically, user devices include smartphones, tablets, and PCs. High-performance computer systems (e.g., server machines with NVIDIA GPUs) are used for the servers. Database management systems such as MySQL and PostgreSQL are used for the databases.

[1115] Data collection and preprocessing

[1116] When a user allows the LINE app to collect conversation history data, the device collects past conversation data from the LINE app and sends it to a server. The server stores the received conversation data in a database. Subsequently, preprocessing is performed, including text cleanup, noise reduction, and duplicate removal. This preprocessing improves the quality of the data, making it suitable for training generative AI models.

[1117] Model training

[1118] The server uses pre-processed data and an emotion engine to train a generative AI model (e.g., GPT-3 or BERT). This training process also utilizes data based on the user's emotional state. Emotion analysis leverages an emotion engine (e.g., IBM Watson, Azure Text Analytics). In this way, the generative AI model becomes capable of responding to fluctuations in the user's emotions.

[1119] Processing real-time messages

[1120] When a user receives a new message on LINE, their device sends the message to the server in real time. The server analyzes the context of the received message using natural language processing techniques (e.g., spaCy or NLTK). Furthermore, it uses an emotion engine to recognize the user's emotional state.

[1121] Generating reply suggestions

[1122] The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the user's emotional state. The generated response options are appropriate to the context and the user's emotional state.

[1123] Suggestions and selection of reply options

[1124] The server sends generated reply options to the terminal, which then presents them to the user. The user selects the best option from the displayed multiple reply options.

[1125] Send a reply

[1126] When a user selects a reply option, the device sends that selected message to the recipient on LINE. This allows users to respond quickly and efficiently to appropriate messages.

[1127] Specific example

[1128] The following is a specific example of this system.

[1129] (Message from a friend): "How have you been lately?"

[1130] When a user receives this message, the device sends it to the server in real time. The server analyzes the context of "How are you doing?" and uses an emotion engine to identify the user's current emotion. For example, if the server identifies the user's current emotion as "stressed," it matches this with past conversation data to generate potential replies such as the following:

[1131] "I'm a little busy, but I'm managing somehow."

[1132] "Work has been really tough lately, how about you?"

[1133] "I'm feeling a bit stressed, but thank you for listening."

[1134] The server sends these reply options to the device, which then suggests them to the user. If the user selects "I'm a bit busy, but I'm managing," the device sends that message to their friend. This method allows users to reply quickly and appropriately.

[1135] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1136] Step 1:

[1137] The user grants permission for the LINE app to collect conversation history data.

[1138] Input: User permission settings

[1139] Output: Instructions for collecting conversation history data

[1140] Specific action: The user grants permission for the collection of conversation data in the LINE settings screen.

[1141] Step 2:

[1142] The device retrieves past conversation data from the LINE app and sends that data to the server.

[1143] Input: Conversation history data

[1144] Output: Sending conversation data to the server

[1145] Specific operation: A background process on the device extracts LINE conversation history data and sends it to the server via a secure protocol (e.g., HTTPS).

[1146] Step 3:

[1147] The server saves the received conversation data to the database.

[1148] Input: Conversation data

[1149] Output: Data stored in the database

[1150] Specific operation: The server receives the conversation data and inserts it into the database in a specific format (e.g., JSON format).

[1151] Step 4:

[1152] The server performs preprocessing on the data it has stored.

[1153] Input: Conversation data stored in the database

[1154] Output: Preprocessed data

[1155] Specific actions: The server performs the following preprocessing:

[1156] Text cleanup: Remove unnecessary symbols and spaces.

[1157] Noise reduction: Removes spam messages and meaningless data.

[1158] Remove duplicates: If the same message is recorded multiple times, it will be merged into one.

[1159] Step 5:

[1160] The server uses pre-processed data to train the generated AI model.

[1161] Input: Preprocessed data

[1162] Output: Trained Generative AI Model

[1163] Specific operation: The server uses high-performance computing resources (e.g., NVIDIA GPU) to train a generative AI model (e.g., GPT-3). During this process, sentiment analysis data is also incorporated into the model using an emotion engine.

[1164] Step 6:

[1165] The user receives a new message on LINE.

[1166] Input: New message

[1167] Output: New message displayed on the terminal

[1168] Specific action: The LINE app receives a new message and displays it on the device.

[1169] Step 7:

[1170] The terminal sends the received message to the server.

[1171] Input: Received message

[1172] Output: Sending a message to the server

[1173] Specific operation: The device forwards messages received from the LINE app to the server, and the server receives them.

[1174] Step 8:

[1175] The server analyzes the context of the received message and uses an emotion engine to recognize the user's emotional state.

[1176] Input: Received message

[1177] Output: Contextual information, emotional state

[1178] Specific operation: The server uses natural language processing technology (e.g., spaCy) to analyze the context of the message and uses an emotion engine to recognize the user's emotional state.

[1179] Step 9:

[1180] Based on the analyzed contextual information and the user's emotional state, the server uses a generative AI model to generate multiple response options.

[1181] Input: Contextual information, emotional state, trained generative AI model

[1182] Output: Reply candidates

[1183] Specific operation: The server inputs a prompt message into the AI ​​model, which then generates response candidates that take into account contextual information and emotional state.

[1184] Step 10:

[1185] The server sends the generated response options to the terminal, which then suggests them to the user.

[1186] Input: Reply suggestions

[1187] Output: Sending reply suggestions to the terminal, presenting them to the user.

[1188] Specific operation: The server sends suggested replies to the terminal, and the terminal displays them to the user.

[1189] Step 11:

[1190] When a user selects a reply option, the device sends that selected message to the recipient via the LINE app through the server.

[1191] Input: Selected reply options

[1192] Output: Message sent to the recipient via the LINE app

[1193] Specific operation: The user selects a reply option on the screen, and the device sends that message to the recipient via the LINE app through the server.

[1194] (Application Example 2)

[1195] Next, we will explain application example 2. In the following explanation, 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."

[1196] In physical stores, effective communication between staff and customers is crucial for improving customer satisfaction. However, responding quickly and appropriately to customer requests and questions is not easy. In particular, being able to respond in a way that is sensitive to the customer's emotional state would further enhance the customer experience, but there is a lack of effective tools to achieve this.

[1197] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting past dialogue data, means for pre-processing the collected dialogue data, means for training a generative AI model using the pre-processed dialogue data, means for analyzing received messages and generating a plurality of response candidates using the generative AI model, means for transmitting the generated response candidates to the user's display device, means for transmitting the response selected by the user, and means for comparing and analyzing past dialogue history using an emotion engine that identifies the customer's emotional state. As a result, staff can provide quick and appropriate responses to customer requests and questions, and in particular, personalized responses that are in line with the customer's emotions become possible.

[1198] "Dialogue data" refers to a record of conversations that take place between a user and another person.

[1199] "Preprocessing" is the process of removing unnecessary information from collected data and converting it into a usable format.

[1200] A "generative AI model" is an artificial intelligence model that learns patterns based on collected data and performs specific tasks.

[1201] "Received messages" are messages that a user receives from other people.

[1202] "Response candidates" are multiple reply options proposed by a generative AI model.

[1203] A "display device" is a device that a user uses to view response options and other information.

[1204] An "emotion engine" is a system that analyzes and identifies emotions from text, audio, and other sources.

[1205] "Contextual analysis" is the process of understanding the content and background of a message and extracting information to generate an appropriate response.

[1206] The "Concierge Assistant" system of this invention is a tool that supports communication between staff and customers in physical stores. This system operates in the following steps:

[1207] Hardware and software to be used

[1208] Hardware: Tablets for customer service (e.g., iPad)

[1209] software:

[1210] Message management application

[1211] Emotion engine (e.g., IBM Watson Tone Analyzer)

[1212] Generative AI models (e.g., GPT-4)

[1213] System Implementation Procedures

[1214] 1. Data collection and preprocessing

[1215] The server collects past conversation data from customer service tablets. The collected data is pre-processed on the server. Pre-processing includes text cleanup, noise reduction, and duplicate removal.

[1216] 2. Training of generative AI models

[1217] The server uses pre-processed data to train a generative AI model. During this process, an emotion engine is used to consider the customer's emotional state, resulting in a model that can adapt to individual customer emotional fluctuations.

[1218] 3. Processing real-time messages

[1219] When a customer service tablet receives a new question or request from a customer, it sends the content to the server. The server contextualizes the received message and uses an emotion engine to identify the customer's emotional state.

[1220] 4. Generating response candidates

[1221] The server uses a generative AI model to generate multiple response candidates based on the analyzed contextual information and the customer's emotional state.

[1222] 5. Proposal of response candidates

[1223] The server sends the generated response options to a customer service tablet, which then displays them to the staff. The staff then selects the most appropriate response option.

[1224] 6. Sending a response

[1225] When a staff member selects a response option, the customer service tablet displays that message to the customer.

[1226] Specific example

[1227] For example, if a customer asks, "How do I use this product?", the customer service tablet sends the question to the server. The server analyzes the context of "How do I use this product?" and identifies the customer's emotion through an emotion engine (e.g., IBM Watson Tone Analyzer). If the customer is identified as "interested," a generative AI model (e.g., GPT-4) generates response candidates such as the following.

[1228] "It would be best if you press this button first, and then do this next."

[1229] "We also have an easy-to-understand demo video; would you like to watch it?"

[1230] "If you'd like to try it out, we can help you."

[1231] If a staff member selects "We also have an easy-to-understand demo video; would you like to watch it?" from the suggested responses, the customer service tablet will display this message to the customer.

[1232] Example of a prompt

[1233] "Based on past interaction history and the customer's emotional state, please propose the most appropriate response."

[1234] This system will allow in-store staff to provide quick and personalized service to customers, which is expected to improve customer satisfaction.

[1235] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1236] Step 1:

[1237] Data collection and preprocessing

[1238] Input: Past conversation data collected by the user using a customer service tablet.

[1239] Processing: The terminal sends past conversation data to the server. The server stores the received data in a database and performs preprocessing such as data cleanup, noise reduction, and duplicate data removal.

[1240] Output: Pre-processed, clean dialogue data.

[1241] Step 2:

[1242] Training of generative AI models

[1243] Input: Preprocessed dialogue data and sentiment analysis data from the sentiment engine.

[1244] Processing: The server trains a generative AI model using pre-processed data. During this process, it incorporates emotional information from the dialogue data using an emotion engine, building a model that can respond to changes in the user's emotions.

[1245] Output: A pre-trained generative AI model capable of responding to the user's emotional state.

[1246] Step 3:

[1247] Processing real-time messages

[1248] Input: A message from a new customer that the user receives on their customer service tablet.

[1249] Processing: The terminal sends the received message to the server. The server analyzes the context of the received message and uses an emotion engine to identify the customer's emotional state.

[1250] Output: Analyzed contextual information and emotional state information.

[1251] Step 4:

[1252] Generating response candidates

[1253] Input: Analyzed contextual information and emotional state information.

[1254] Processing: The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the customer's emotional state.

[1255] Output: Multiple response candidates generated by the server.

[1256] Step 5:

[1257] Suggested response options

[1258] Input: Multiple response candidates generated by the server.

[1259] Processing: The server sends the generated response options to the customer service tablet, which then presents them to the staff. The staff selects the most appropriate response from the screen.

[1260] Output: Staff-selected response.

[1261] Step 6:

[1262] Sending a response

[1263] Input: Response selected by the staff.

[1264] Processing: The terminal sends the selected response message back to the customer.

[1265] Output: Response message sent to the customer.

[1266] The specific processing unit 290 transmits the result of the specific processing to the headset terminal 314. In the headset terminal 314, the control unit 46A causes the speaker 240 and display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[1267] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1268] In the above embodiment, an example was given in which specific processing is performed by the data processing device 12, but the technology of this disclosure is not limited thereto, and specific processing may also be performed by the headset terminal 314.

[1269] [Fourth Embodiment]

[1270] Figure 7 shows an example of the configuration of the data processing system 410 according to the fourth embodiment.

[1271] As shown in Figure 7, the 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.

[1272] The data processing device 12 comprises a computer 22, a database 24, and a communication interface 26. The computer 22 is an example of a "computer" related to the technology of this disclosure. The computer 22 comprises a processor 28, RAM 30, and storage 32. The processor 28, RAM 30, and storage 32 are connected to a bus 34. The database 24 and the communication interface 26 are also connected to the bus 34. The communication interface 26 is connected to a network 54. An example of the network 54 is a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1273] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication interface 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and storage 50. The processor 46, RAM 48, and storage 50 are connected to a bus 52. The microphone 238, speaker 240, camera 42, and controlled object 443 are also connected to the bus 52.

[1274] The microphone 238 receives voice signals from the user 20 and receives instructions from the user 20. The microphone 238 captures the voice signals from the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio according to the instructions from the processor 46.

[1275] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an image sensor such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the area around the user 20 (for example, an imaging range defined by a field of view equivalent to the width of a typical healthy person's field of vision).

[1276] Communication interface 44 is connected to network 54. Communication interfaces 44 and 26 are responsible for the exchange of various information between processor 46 and processor 28 via network 54. The exchange of various information between processor 46 and processor 28 using communication interfaces 44 and 26 is performed in a secure manner.

[1277] The controlled object 443 includes a display device, LEDs in the eyes, and motors that drive 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 robot 414's emotions can be expressed by controlling these motors. Furthermore, the robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[1278] Figure 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Figure 8, the data processing device 12 performs specific processing using the processor 28. The storage 32 stores the specific processing program 56.

[1279] The specific processing program 56 is an example of a "program" relating to the technology of this 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.

[1280] The storage 32 stores the data generation model 58 and the emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1281] In robot 414, the processor 46 performs the reception output processing. The storage 50 stores the reception output program 60. 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 processing is realized by the processor 46 operating as a control unit 46A according to the reception output program 60 executed on the RAM 48.

[1282] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1283] System Overview

[1284] The system of this invention proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. This system is implemented using the user's terminal, a server, and a generative AI model.

[1285] Program Processing Overview

[1286] 1. Data collection and preprocessing

[1287] When a user allows the LINE app to collect conversation history data, the device collects past conversation data and sends it to the server. The server preprocesses the received data, performing noise reduction and data cleaning.

[1288] 2. Model training

[1289] The server uses pre-processed data to train generative AI models. Techniques such as sentiment analysis are used here to enable the generation of more natural and optimal responses.

[1290] 3. Processing real-time messages

[1291] When a user receives a new message on LINE, their device sends the message to a server. The server analyzes the context of the received message and generates multiple reply options from a generative AI model based on past conversation data.

[1292] 4. Suggestions for reply options

[1293] The server sends the generated reply options to the terminal, which then displays them to the user. The user selects the most appropriate reply from the displayed options.

[1294] 5. Send a reply

[1295] When a user selects a reply option, the device sends that selected message to the recipient on LINE. This allows users to respond quickly and appropriately.

[1296] Specific example

[1297] The following is a specific example of this system.

[1298] Conversation example

[1299] (Message from a friend): "How have you been lately?"

[1300] When a user receives this message, the device sends it to the server in real time. The server analyzes the context of "How are you doing lately?" and generates possible replies based on past conversation data, such as the following:

[1301] "fine and you?"

[1302] "I'm a little busy, but I'm managing somehow."

[1303] "I recently found a new hobby."

[1304] The server sends these reply options to the device, and the device suggests them to the user. If the user selects "I'm fine! How about you?", the device sends that message to the friend.

[1305] In this way, users can easily provide appropriate responses, enabling efficient and effective conversations, especially in business-related situations or those requiring high communication skills.

[1306] The following describes the processing flow.

[1307] Step 1:

[1308] The user grants the LINE app permission to collect conversation history data.

[1309] Step 2:

[1310] The device retrieves past conversation data from the LINE app.

[1311] Step 3:

[1312] The device sends the acquired conversation data to the server.

[1313] Step 4:

[1314] The server saves the received conversation data to a database.

[1315] Step 5:

[1316] The server preprocesses the conversation data. This preprocessing includes text cleanup, noise reduction, and duplicate removal.

[1317] Step 6:

[1318] The server uses pre-processed data to train generative AI models. Here, techniques such as sentiment analysis are used in conjunction to improve the accuracy of the training.

[1319] Step 7:

[1320] The user receives a new message on LINE.

[1321] Step 8:

[1322] The terminal sends the received message to the server.

[1323] Step 9:

[1324] The server analyzes incoming messages to understand their context. In doing so, it compares them with past conversation data to determine the appropriate context.

[1325] Step 10:

[1326] Based on the analyzed contextual information, the server uses a generative AI model to generate multiple response options.

[1327] Step 11:

[1328] The server sends the generated reply options to the terminal.

[1329] Step 12:

[1330] The device displays reply suggestions to the user.

[1331] Step 13:

[1332] The user selects the most suitable reply from the displayed options.

[1333] Step 14:

[1334] The device sends the reply message selected by the user to the recipient on LINE.

[1335] This series of steps allows users to efficiently provide appropriate responses.

[1336] (Example 1)

[1337] Next, we will describe Example 1. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1338] Modern messaging applications lack sufficient means for users to respond quickly and appropriately. In particular, few existing systems generate responses that consider context based on past conversation data, requiring users to manually formulate replies, which can reduce communication efficiency. Furthermore, technologies that incorporate sentiment analysis to achieve more natural conversations are also insufficient.

[1339] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 1 is realized by the following means.

[1340] In this invention, the server includes means for allowing the user to collect conversation history data, means for collecting past conversation data from the terminal and sending it to the server, means for preprocessing the received conversation data to remove noise and clean the data, means for training a generative AI model using the preprocessed conversation data, means for training the generative AI model by incorporating sentiment analysis, means for receiving a new message, means for sending the received message to the server, means for performing contextual analysis of the message and inputting a prompt sentence into the generative AI model to generate multiple reply candidates, means for sending the generated reply candidates to the user's terminal, means for the user to select from the displayed reply candidates, and means for sending the selected reply to the recipient of the messaging application. This enables the user to make quick and appropriate replies.

[1341] A "user" refers to a person who uses a messaging application to have a conversation.

[1342] "Device" refers to electronic devices such as smartphones and computers used by users.

[1343] A "server" refers to a computing system that receives data sent from a terminal and processes it.

[1344] "Conversation history data" refers to records of conversations that have taken place on messaging applications in the past.

[1345] "Preprocessing" refers to the process of preparing received data into an analyzable format. Specifically, this includes noise reduction and data cleaning.

[1346] A "generative AI model" refers to a model that uses artificial intelligence to generate output in response to a specific input. An example is a text generation model.

[1347] "Sentiment analysis" refers to the technique of identifying and classifying the emotions contained within text.

[1348] "New messages" refers to the most recent messages a user receives in their messaging application.

[1349] "Contextual analysis" refers to analytical methods used to understand the meaning and underlying intentions of text.

[1350] A "prompt" refers to the document or question that is input into a generative AI model.

[1351] "Suggested replies" refers to multiple response options generated by the server using a generated AI model.

[1352] A "messaging application" refers to software that users use to send and receive messages.

[1353] Modes for carrying out the invention

[1354] This invention relates to a system that suggests appropriate replies to messages received by a user in a messaging application, based on past conversation data. This system is implemented using the user's terminal, a server, and a generative AI model.

[1355] Data collection and preprocessing

[1356] When a user allows a messaging application to collect conversation history data, the device collects past conversation data and sends it to the server. The server preprocesses the received conversation data, performing noise reduction and data cleaning. Specifically, it uses libraries such as Pandas and NumPy to clean up the data. For example, it removes unnecessary information such as newline characters and emojis.

[1357] Model training

[1358] The server trains generative AI models using preprocessed data. Specifically, it uses the Hugging Face Transformers library to train generative AI models such as BERT and GPT-3. In addition, it uses the BERT model to label emotions within text for sentiment analysis. For example, it trains the model based on past conversation data regarding the message "How are you doing lately?".

[1359] Processing real-time messages

[1360] When a user receives a new message, the device sends it to the server. The server uses the spacy library to analyze the context of the message and inputs prompts into a generative AI model to generate an appropriate reply. For example, in response to a message like "How are you doing?", the server analyzes the context from past conversation data and generates an appropriate reply.

[1361] Suggested replies

[1362] The server sends several generated reply options to the device. The device then uses JavaScript or Swift to display the reply options in a user-friendly interface. For example, it might suggest the following reply options to the user:

[1363] "fine and you?"

[1364] "I'm a little busy, but I'm managing somehow."

[1365] "I recently found a new hobby."

[1366] Send a reply

[1367] The user selects one of the displayed reply options. The device then sends the selected reply to the recipient in the messaging application. This is done using the messaging application's API.

[1368] This specific process allows users to leverage generative AI models and sentiment analysis to provide quick and appropriate responses. This enables efficient and effective conversations, especially in business-oriented situations or those requiring high-level communication skills.

[1369] The flow of the specific processing in Example 1 will be explained using Figure 11.

[1370] Step 1: Data collection and preprocessing

[1371] The user grants permission for the messaging application to collect conversation history data. The device collects the conversation history data permitted by the user and sends it to the server. The server preprocesses the received conversation data, performing noise reduction and data cleaning. Specifically, it uses the Pandas library to remove unnecessary line breaks and emojis and format the data.

[1372] Input: User-authorized conversation history data

[1373] Output: Pre-processed clean data

[1374] Step 2: Model Training

[1375] The server trains a generative AI model using pre-processed conversational data. Here, the Hugging Face Transformers library is used to train models such as BERT and GPT-3. The BERT model is also used for sentiment analysis, classifying the sentiment of each message. For example, for a message like "How are you?", conversations with similar contexts are extracted from past conversational data and used for training.

[1376] Input: Preprocessed conversation data

[1377] Output: Trained generative AI model

[1378] Step 3: Receiving real-time messages

[1379] When a user receives a new message in their messaging application, the device sends this new message to the server. The message includes the sender, recipient, and message content.

[1380] Input: New incoming message

[1381] Output: Message data sent to the server

[1382] Step 4: Contextual analysis and generation of reply candidates

[1383] The server performs contextual analysis on newly received messages. This analysis uses the spacy library to understand the meaning of the message. It then generates an appropriate prompt and inputs it into a generative AI model to produce multiple response options. For example, in response to the message "How are you?", based on past conversation data and sentiment analysis results, it generates response options such as: "I'm fine! How about you?", "I'm a little busy, but I'm managing," and "I recently found a new hobby."

[1384] Input: Newly received message

[1385] Output: Multiple response candidates generated

[1386] Step 5: Suggesting possible replies

[1387] The server sends the generated reply suggestions to the device. The device then displays the reply suggestions to the user using a simple and user-friendly interface based on JavaScript or Swift.

[1388] Input: Multiple generated reply suggestions

[1389] Output: Reply suggestions displayed to the user

[1390] Step 6: Select and send your reply.

[1391] The user selects the most suitable reply from the displayed options. The device then sends the selected reply to the recipient in the messaging application. This is done using the messaging application's API.

[1392] Input: User's selected reply

[1393] Output: Reply message sent to the recipient

[1394] (Application Example 1)

[1395] Next, we will explain Application Example 1. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1396] Traditional food delivery services often struggle to respond quickly and accurately to user messages regarding order confirmations, order changes, and complaints. This frequently results in a poor user experience, potentially leading to decreased service usage and satisfaction. Furthermore, in business environments requiring frequent communication, greater efficiency is needed.

[1397] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 1 is realized by the following means.

[1398] In this invention, the server includes means for collecting past conversation data, means for preprocessing the collected conversation data, means for training a generative AI model using the preprocessed conversation data, means for analyzing received messages and generating multiple reply candidates using the generative AI model, means for sending the generated reply candidates to the user's terminal, means for sending the reply selected by the user, and means for suggesting appropriate reply candidates to order confirmation messages, order change request messages, or complaint handling messages when the user is using an online delivery service. This enables users to make quick and appropriate replies, improving the user experience and increasing the efficiency of the delivery service.

[1399] "Past conversation data" refers to records of communications a user has had in the past, and this includes various formats such as text messages, voice messages, and chat logs.

[1400] "Preprocessing" refers to a series of processes that remove noise from collected conversation data, convert it to the required data format, and prepare it for analysis.

[1401] A "generative AI model" is a model that uses AI algorithms to generate new text data, and can also generate appropriate answers and replies based on the user's past conversation data.

[1402] "Suggested replies" refer to multiple possible responses to an incoming message, allowing the user to choose from among them.

[1403] "Online delivery service" refers to any service that allows users to order food and goods via the internet and have them delivered to their homes or workplaces.

[1404] An "order confirmation message" refers to a message sent to a user after they have placed an order for a product or service, to confirm the details of that order.

[1405] An "order change request message" refers to a message sent when a user wants to make changes to an order they have already placed.

[1406] A "complaint response message" refers to a message sent by a user when they are dissatisfied with a product or service or have a problem with it, and appropriate action is taken based on this message.

[1407] The present invention provides a system that suggests appropriate responses to various messages received by a user while using an online delivery service. This system includes the user's terminal, a server, and a generative AI model.

[1408] Specific System Procedures

[1409] The server first collects past conversation data from the user's device. This data includes text messages and chat logs, and once collected, it undergoes preprocessing such as noise reduction. The preprocessed data is then used to train a generative AI model. By incorporating sentiment analysis techniques at this stage, it becomes possible to generate more natural and appropriate responses.

[1410] When a user receives a new message via the delivery service, the device sends the message to the server in real time. The server analyzes the received message and generates multiple reply options using a generative AI model. The generated reply options are sent to the device and displayed to the user. The user selects the appropriate reply from the displayed options, and the selected reply is sent to the recipient.

[1411] Hardware and software

[1412] The hardware used includes the user's smartphone or other device, as well as a server. The software includes scripts for data collection and preprocessing, generative AI models (e.g., OpenAI GPT-3), and sentiment analysis libraries (e.g., TextBlob).

[1413] As a concrete example, consider a case where a user receives an "order confirmation message" while using a food delivery service. The following response options are generated in response to this message.

[1414] Example of a prompt

[1415] Past conversation history: "Order confirmation: Yes, please." "Delivery delay: Excuse me, is it delayed?"

[1416] Current message: "We will arrive shortly."

[1417] Possible replies to generate:

[1418] 1. "Understood, thank you."

[1419] 2. "How much longer until we arrive?"

[1420] In this way, users can respond quickly and accurately, improving their experience with online delivery services. Furthermore, increased efficiency can be expected in the work environment.

[1421] The flow of a specific process in Application Example 1 will be explained using Figure 12.

[1422] Step 1:

[1423] The user authorizes the collection of past conversation data through their device. The device, upon receiving permission, collects past chat logs and message history from LINE and other messaging applications. The collected data is then sent directly to the server. The input is past conversation data, and the output is the collected data. Specifically, this involves sending past message history to the server in text format.

[1424] Step 2:

[1425] The server preprocesses the conversation data received from the terminal. Preprocessing includes removing noise and filtering out unnecessary information from the data. The input is the collected data, and the output is the preprocessed, cleaned data. Specifically, this involves performing text analysis on the data and removing meaningless information and error messages.

[1426] Step 3:

[1427] The server trains a generative AI model using preprocessed data. Sentiment analysis techniques are also used to learn the emotional nuances of the data. The input is preprocessed data, and the output is the trained generative AI model. Specifically, this involves performing sentiment analysis using a Python library (e.g., TextBlob) and training a generative AI model such as OpenAI GPT-3.

[1428] Step 4:

[1429] When a user receives a new message via LINE or another messaging app, the device sends the message to a server. The server analyzes the context of the received message and generates multiple reply options from a generative AI model based on past conversation data. The input is the newly received message, and the output is the generated reply options. Specifically, this involves text analysis of the received message and outputting appropriate reply options.

[1430] Step 5:

[1431] The server sends the generated reply options to the user's terminal. The terminal displays the reply options to the user. The user selects the most appropriate reply from the displayed options. The input is the generated reply options, and the output is the reply selected by the user. Specifically, this involves displaying a list of reply options in the terminal's user interface and accepting the user's selection.

[1432] Step 6:

[1433] The user sends their selected reply from their device to the server, which then forwards that selected reply to the LINE recipient. The input is the user-selected reply, and the output is the actual reply message sent. Specifically, this involves sending the selected message to the recipient via the LINE API.

[1434] Furthermore, an emotion engine that estimates the user's emotions may be incorporated. That is, the identification processing unit 290 may use the emotion identification model 59 to estimate the user's emotions and perform identification processing using the user's emotions.

[1435] System Overview

[1436] The present invention's system proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. In particular, by combining it with an emotion engine that recognizes the user's emotions, it enables the generation of more personalized replies. This system is implemented using the user's terminal, a server, a generative AI model, and an emotion engine.

[1437] Program Processing Overview

[1438] 1. Data collection and preprocessing

[1439] When a user allows the LINE app to collect conversation history data, the device collects past conversation data and sends it to a server. The server stores the received data in a database and performs preprocessing. Preprocessing includes text cleanup, noise reduction, and duplicate removal.

[1440] 2. Model training

[1441] The server uses pre-processed data to train a generative AI model. During this process, data based on the user's emotional state is also used, incorporating sentiment analysis into the training. This allows the AI ​​model to respond to fluctuations in the user's emotions.

[1442] 3. Processing real-time messages

[1443] When a user receives a new message on LINE, their device sends the message to the server. The server performs contextual analysis of the received message and uses an emotion engine to recognize the user's emotional state. This is then compared with past conversation data to analyze the appropriate context.

[1444] 4. Generating reply suggestions

[1445] The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the user's emotional state. This ensures that the response options also match the user's emotions.

[1446] 5. Suggestions for reply options

[1447] The server sends generated reply options to the terminal, which then suggests them to the user. The user selects the most suitable reply from the displayed options.

[1448] 6. Send a reply

[1449] When a user selects a reply option, the device sends that selected message to the recipient on LINE. This allows users to respond quickly and appropriately.

[1450] Specific example

[1451] The following is a specific example of this system.

[1452] Conversation example

[1453] (Message from a friend): "How have you been lately?"

[1454] When a user receives this message, the device sends it to the server in real time. The server analyzes the context of "How are you doing?" and uses an emotion engine to identify the user's current emotion. For example, if the server identifies the user's current emotion as "stressed," it matches this with past conversation data to generate potential replies such as the following:

[1455] "I'm a little busy, but I'm managing somehow."

[1456] "Work has been really tough lately, how about you?"

[1457] "I'm feeling a bit stressed, but thank you for listening."

[1458] The server sends these response options to the device, and the device suggests them to the user. If the user selects "I'm a bit busy, but I'm managing," the device sends that message to the friend.

[1459] In this way, users can easily provide appropriate replies, enabling efficient and effective conversations, especially in business-related situations or those requiring high communication skills. Furthermore, a key feature is the ability to provide more personalized replies by considering the user's emotional state.

[1460] The following describes the processing flow.

[1461] System Overview

[1462] The system of this invention proposes appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. In particular, by combining it with an emotion engine that recognizes the user's emotions, it enables the generation of more personalized replies. This system is implemented using the user's terminal, a server, a generative AI model, and an emotion engine.

[1463] Program Processing Overview

[1464] Step 1:

[1465] The user grants the LINE app permission to collect conversation history data.

[1466] Step 2:

[1467] The device retrieves past conversation data from the LINE app.

[1468] Step 3:

[1469] The device sends the acquired conversation data to the server.

[1470] Step 4:

[1471] The server saves the received conversation data to a database.

[1472] Step 5:

[1473] The server preprocesses the conversation data. Preprocessing includes text cleanup, noise reduction, and duplicate removal.

[1474] Step 6:

[1475] The server uses pre-processed data to train generative AI models. It also incorporates sentiment analysis into the training process, using data based on the user's emotional state.

[1476] Step 7:

[1477] The user receives a new message on LINE.

[1478] Step 8:

[1479] The terminal sends the received message to the server.

[1480] Step 9:

[1481] The server analyzes incoming messages to understand their context. During this process, it also recognizes the user's emotional state using an emotion engine.

[1482] Step 10:

[1483] The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the user's emotional state.

[1484] Step 11:

[1485] The server sends the generated reply options to the terminal.

[1486] Step 12:

[1487] The device displays reply suggestions to the user.

[1488] Step 13:

[1489] The user selects the most suitable reply from the displayed options.

[1490] Step 14:

[1491] The device sends the reply message selected by the user to the recipient on LINE.

[1492] Specific example

[1493] Conversation example

[1494] (Message from a friend): "How have you been lately?"

[1495] Step 1:

[1496] The user receives this message.

[1497] Step 2:

[1498] The device sends the message to the server in real time.

[1499] Step 3:

[1500] The server analyzes the context of the question "How are you doing lately?" and uses its emotion engine to identify the user's current emotional state. For example, it might identify the user's current emotion as "feeling stressed."

[1501] Step 4:

[1502] The server generates response options based on past conversation data and emotional states analyzed by the emotion engine, such as the following:

[1503] "I'm a little busy, but I'm managing somehow."

[1504] "Work has been really tough lately, how about you?"

[1505] "I'm feeling a bit stressed, but thank you for listening."

[1506] Step 5:

[1507] The server sends these suggested replies to the terminal.

[1508] Step 6:

[1509] The device suggests and displays reply options to the user.

[1510] Step 7:

[1511] The user selects, "I'm a little busy, but I'm managing."

[1512] Step 8:

[1513] The device sends the selected message to the friend.

[1514] This system not only automatically suggests appropriate replies but also takes into account the user's emotional state, enabling more personalized communication. Its effectiveness is particularly evident in business settings and situations requiring advanced communication skills.

[1515] (Example 2)

[1516] Next, we will describe Example 2. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1517] Traditional messaging applications have struggled to generate appropriate replies to received messages. In particular, existing technologies were insufficient to generate personalized replies that took into account the user's emotional state. As a result, users had to manually compose replies, hindering efficient and effective communication.

[1518] The identification process performed by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means.

[1519] In this invention, the server includes means for collecting past conversation data, means for preprocessing the collected conversation data, means for training a generative AI model using the preprocessed conversation data, means for analyzing an incoming message and generating multiple reply candidates using the generative AI model, means for sending the generated reply candidates to the user's terminal, means for sending the reply selected by the user, means for using an emotion engine to recognize the user's emotional state, and means for generating reply candidates based on contextual information of the incoming message and the user's emotional state. This enables the generation of personalized replies that take the user's emotions into consideration, allowing for efficient, rapid, and appropriate replies.

[1520] "Past conversation data" refers to information that includes the history of messages a user has sent in the past.

[1521] "Preprocessing" refers to the process of removing noise from received data, cleaning up text, and removing duplicates.

[1522] A "generative AI model" is an artificial intelligence model that is trained based on collected data and generates appropriate outputs (responses) in response to given inputs.

[1523] An "emotion engine" is a tool or software used to analyze and identify a user's emotional state.

[1524] "Contextual information" refers to information that includes meaning derived from the content of a message and the flow of past conversations.

[1525] "Suggested replies" refer to multiple response options proposed by the generative AI model.

[1526] A "server" is a computer system used for tasks such as data collection, preprocessing, analysis, and training of AI models.

[1527] A "user's device" refers to a device used by the user (such as a smartphone, tablet, or PC) that has the function of receiving and displaying reply suggestions from the server.

[1528] A "messaging application" is a general term for application software that allows users to send and receive text messages.

[1529] Modes for carrying out the invention

[1530] The present invention provides a system that suggests appropriate replies to messages received by users in messaging applications such as LINE, based on past conversation data. In particular, this system enables the generation of more personalized replies by incorporating an emotion engine that recognizes the user's emotions. The embodiments for carrying out this invention are described in detail below.

[1531] The system is implemented using user devices, servers, generative AI models, and emotion engines. Specifically, user devices include smartphones, tablets, and PCs. High-performance computer systems (e.g., server machines with NVIDIA GPUs) are used for the servers. Database management systems such as MySQL and PostgreSQL are used for the databases.

[1532] Data collection and preprocessing

[1533] When a user allows the LINE app to collect conversation history data, the device collects past conversation data from the LINE app and sends it to a server. The server stores the received conversation data in a database. Subsequently, preprocessing is performed, including text cleanup, noise reduction, and duplicate removal. This preprocessing improves the quality of the data, making it suitable for training generative AI models.

[1534] Model training

[1535] The server uses pre-processed data and an emotion engine to train a generative AI model (e.g., GPT-3 or BERT). This training process also utilizes data based on the user's emotional state. Emotion analysis leverages an emotion engine (e.g., IBM Watson, Azure Text Analytics). In this way, the generative AI model becomes capable of responding to fluctuations in the user's emotions.

[1536] Processing real-time messages

[1537] When a user receives a new message on LINE, their device sends the message to the server in real time. The server analyzes the context of the received message using natural language processing techniques (e.g., spaCy or NLTK). Furthermore, it uses an emotion engine to recognize the user's emotional state.

[1538] Generating reply suggestions

[1539] The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the user's emotional state. The generated response options are appropriate to the context and the user's emotional state.

[1540] Suggestions and selection of reply options

[1541] The server sends generated reply options to the terminal, which then presents them to the user. The user selects the best option from the displayed multiple reply options.

[1542] Send a reply

[1543] When a user selects a reply option, the device sends that selected message to the recipient on LINE. This allows users to respond quickly and efficiently to appropriate messages.

[1544] Specific example

[1545] The following is a specific example of this system.

[1546] (Message from a friend): "How have you been lately?"

[1547] When a user receives this message, the device sends it to the server in real time. The server analyzes the context of "How are you doing?" and uses an emotion engine to identify the user's current emotion. For example, if the server identifies the user's current emotion as "stressed," it matches this with past conversation data to generate potential replies such as the following:

[1548] "I'm a little busy, but I'm managing somehow."

[1549] "Work has been really tough lately, how about you?"

[1550] "I'm feeling a bit stressed, but thank you for listening."

[1551] The server sends these reply options to the device, which then suggests them to the user. If the user selects "I'm a bit busy, but I'm managing," the device sends that message to their friend. This method allows users to reply quickly and appropriately.

[1552] The flow of the specific processing in Example 2 will be explained using Figure 13.

[1553] Step 1:

[1554] The user grants permission for the LINE app to collect conversation history data.

[1555] Input: User permission settings

[1556] Output: Instructions for collecting conversation history data

[1557] Specific action: The user grants permission for the collection of conversation data in the LINE settings screen.

[1558] Step 2:

[1559] The device retrieves past conversation data from the LINE app and sends that data to the server.

[1560] Input: Conversation history data

[1561] Output: Sending conversation data to the server

[1562] Specific operation: A background process on the device extracts LINE conversation history data and sends it to the server via a secure protocol (e.g., HTTPS).

[1563] Step 3:

[1564] The server saves the received conversation data to the database.

[1565] Input: Conversation data

[1566] Output: Data stored in the database

[1567] Specific operation: The server receives the conversation data and inserts it into the database in a specific format (e.g., JSON format).

[1568] Step 4:

[1569] The server performs preprocessing on the data it has stored.

[1570] Input: Conversation data stored in the database

[1571] Output: Preprocessed data

[1572] Specific actions: The server performs the following preprocessing:

[1573] Text cleanup: Remove unnecessary symbols and spaces.

[1574] Noise reduction: Removes spam messages and meaningless data.

[1575] Remove duplicates: If the same message is recorded multiple times, it will be merged into one.

[1576] Step 5:

[1577] The server uses pre-processed data to train the generated AI model.

[1578] Input: Preprocessed data

[1579] Output: Trained Generative AI Model

[1580] Specific operation: The server uses high-performance computing resources (e.g., NVIDIA GPU) to train a generative AI model (e.g., GPT-3). During this process, sentiment analysis data is also incorporated into the model using an emotion engine.

[1581] Step 6:

[1582] The user receives a new message on LINE.

[1583] Input: New message

[1584] Output: New message displayed on the terminal

[1585] Specific action: The LINE app receives a new message and displays it on the device.

[1586] Step 7:

[1587] The terminal sends the received message to the server.

[1588] Input: Received message

[1589] Output: Sending a message to the server

[1590] Specific operation: The device forwards messages received from the LINE app to the server, and the server receives them.

[1591] Step 8:

[1592] The server analyzes the context of the received message and uses an emotion engine to recognize the user's emotional state.

[1593] Input: Received message

[1594] Output: Contextual information, emotional state

[1595] Specific operation: The server uses natural language processing technology (e.g., spaCy) to analyze the context of the message and uses an emotion engine to recognize the user's emotional state.

[1596] Step 9:

[1597] Based on the analyzed contextual information and the user's emotional state, the server uses a generative AI model to generate multiple response options.

[1598] Input: Contextual information, emotional state, trained generative AI model

[1599] Output: Reply candidates

[1600] Specific operation: The server inputs a prompt message into the AI ​​model, which then generates response candidates that take into account contextual information and emotional state.

[1601] Step 10:

[1602] The server sends the generated response options to the terminal, which then suggests them to the user.

[1603] Input: Reply suggestions

[1604] Output: Sending reply suggestions to the terminal, presenting them to the user.

[1605] Specific operation: The server sends suggested replies to the terminal, and the terminal displays them to the user.

[1606] Step 11:

[1607] When a user selects a reply option, the device sends that selected message to the recipient via the LINE app through the server.

[1608] Input: Selected reply options

[1609] Output: Message sent to the recipient via the LINE app

[1610] Specific operation: The user selects a reply option on the screen, and the device sends that message to the recipient via the LINE app through the server.

[1611] (Application Example 2)

[1612] Next, we will explain application example 2. In the following explanation, the data processing device 12 will be referred to as the "server" and the robot 414 as the "terminal".

[1613] In physical stores, effective communication between staff and customers is crucial for improving customer satisfaction. However, responding quickly and appropriately to customer requests and questions is not easy. In particular, being able to respond in a way that is sensitive to the customer's emotional state would further enhance the customer experience, but there is a lack of effective tools to achieve this.

[1614] The specific processing performed by the specific processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes means for collecting past dialogue data, means for pre-processing the collected dialogue data, means for training a generative AI model using the pre-processed dialogue data, means for analyzing received messages and generating a plurality of response candidates using the generative AI model, means for transmitting the generated response candidates to the user's display device, means for transmitting the response selected by the user, and means for comparing and analyzing past dialogue history using an emotion engine that identifies the customer's emotional state. As a result, staff can provide quick and appropriate responses to customer requests and questions, and in particular, personalized responses that are in line with the customer's emotions become possible.

[1615] "Dialogue data" refers to a record of conversations that take place between a user and another person.

[1616] "Preprocessing" is the process of removing unnecessary information from collected data and converting it into a usable format.

[1617] A "generative AI model" is an artificial intelligence model that learns patterns based on collected data and performs specific tasks.

[1618] "Received messages" are messages that a user receives from other people.

[1619] "Response candidates" are multiple reply options proposed by a generative AI model.

[1620] A "display device" is a device that a user uses to view response options and other information.

[1621] An "emotion engine" is a system that analyzes and identifies emotions from text, audio, and other sources.

[1622] "Contextual analysis" is the process of understanding the content and background of a message and extracting information to generate an appropriate response.

[1623] The "Concierge Assistant" system of this invention is a tool that supports communication between staff and customers in physical stores. This system operates in the following steps:

[1624] Hardware and software to be used

[1625] Hardware: Tablets for customer service (e.g., iPad)

[1626] software:

[1627] Message management application

[1628] Emotion engine (e.g., IBM Watson Tone Analyzer)

[1629] Generative AI models (e.g., GPT-4)

[1630] System Implementation Procedures

[1631] 1. Data collection and preprocessing

[1632] The server collects past conversation data from customer service tablets. The collected data is pre-processed on the server. Pre-processing includes text cleanup, noise reduction, and duplicate removal.

[1633] 2. Training of generative AI models

[1634] The server uses pre-processed data to train a generative AI model. During this process, an emotion engine is used to consider the customer's emotional state, resulting in a model that can adapt to individual customer emotional fluctuations.

[1635] 3. Processing real-time messages

[1636] When a customer service tablet receives a new question or request from a customer, it sends the content to the server. The server contextualizes the received message and uses an emotion engine to identify the customer's emotional state.

[1637] 4. Generating response candidates

[1638] The server uses a generative AI model to generate multiple response candidates based on the analyzed contextual information and the customer's emotional state.

[1639] 5. Proposal of response candidates

[1640] The server sends the generated response options to a customer service tablet, which then displays them to the staff. The staff then selects the most appropriate response option.

[1641] 6. Sending a response

[1642] When a staff member selects a response option, the customer service tablet displays that message to the customer.

[1643] Specific example

[1644] For example, if a customer asks, "How do I use this product?", the customer service tablet sends the question to the server. The server analyzes the context of "How do I use this product?" and identifies the customer's emotion through an emotion engine (e.g., IBM Watson Tone Analyzer). If the customer is identified as "interested," a generative AI model (e.g., GPT-4) generates response candidates such as the following.

[1645] "It would be best if you press this button first, and then do this next."

[1646] "We also have an easy-to-understand demo video; would you like to watch it?"

[1647] "If you'd like to try it out, we can help you."

[1648] If a staff member selects "We also have an easy-to-understand demo video; would you like to watch it?" from the suggested responses, the customer service tablet will display this message to the customer.

[1649] Example of a prompt

[1650] "Based on past interaction history and the customer's emotional state, please propose the most appropriate response."

[1651] This system will allow in-store staff to provide quick and personalized service to customers, which is expected to improve customer satisfaction.

[1652] The flow of a specific process in Application Example 2 will be explained using Figure 14.

[1653] Step 1:

[1654] Data collection and preprocessing

[1655] Input: Past conversation data collected by the user using a customer service tablet.

[1656] Processing: The terminal sends past conversation data to the server. The server stores the received data in a database and performs preprocessing such as data cleanup, noise reduction, and duplicate data removal.

[1657] Output: Pre-processed, clean dialogue data.

[1658] Step 2:

[1659] Training of generative AI models

[1660] Input: Preprocessed dialogue data and sentiment analysis data from the sentiment engine.

[1661] Processing: The server trains a generative AI model using pre-processed data. During this process, it incorporates emotional information from the dialogue data using an emotion engine, building a model that can respond to changes in the user's emotions.

[1662] Output: A pre-trained generative AI model capable of responding to the user's emotional state.

[1663] Step 3:

[1664] Processing real-time messages

[1665] Input: A message from a new customer that the user receives on their customer service tablet.

[1666] Processing: The terminal sends the received message to the server. The server analyzes the context of the received message and uses an emotion engine to identify the customer's emotional state.

[1667] Output: Analyzed contextual information and emotional state information.

[1668] Step 4:

[1669] Generating response candidates

[1670] Input: Analyzed contextual information and emotional state information.

[1671] Processing: The server uses a generative AI model to generate multiple response options based on the analyzed contextual information and the customer's emotional state.

[1672] Output: Multiple response candidates generated by the server.

[1673] Step 5:

[1674] Suggested response options

[1675] Input: Multiple response candidates generated by the server.

[1676] Processing: The server sends the generated response options to the customer service tablet, which then presents them to the staff. The staff selects the most appropriate response from the screen.

[1677] Output: Staff-selected response.

[1678] Step 6:

[1679] Sending a response

[1680] Input: Response selected by the staff.

[1681] Processing: The terminal sends the selected response message back to the customer.

[1682] Output: Response message sent to the customer.

[1683] 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 controlled object 443 to output the result of the specific processing. The microphone 238 acquires audio indicating user input for the result of the specific processing. The control unit 46A transmits the audio data indicating user input acquired by the microphone 238 to the data processing unit 12. In the data processing unit 12, the specific processing unit 290 acquires the audio data.

[1684] Data generation model 58 is a type of so-called generative AI (Artificial Intelligence). One example of data generation model 58 is ChatGPT (Internet search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search) <url: https: gemini.google.com ?hl="ja">Examples of generative AI include the following. The data generation model 58 is obtained by performing deep learning on a neural network. The data generation model 58 is input with prompts containing instructions, and with inference data such as audio data representing speech, text data representing text, and image data representing images. The data generation model 58 infers from the input inference data according to the instructions indicated by the prompts, and outputs the inference results in data formats such as audio data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1685] 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 this disclosure is not limited thereto, and the specific processing may also be performed by the robot 414.

[1686] Furthermore, the emotion identification model 59, acting 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 a specific mapping, which is an emotion map (see Figure 9). Similarly, the emotion identification model 59 may also determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1687] Figure 9 shows an emotion map 400 in which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. The closer to the center of the concentric circles, the more primitive the emotions are located. Further out of the concentric circles, emotions representing states and actions arising from mental states are located. Emotion is a concept that includes feelings and mental states. On the left side of the concentric circles, emotions that are generally generated from reactions occurring in the brain are located. On the right side of the concentric circles, emotions that are generally induced by situational judgment are located. Above and below the concentric circles, emotions that are generally generated from reactions occurring in the brain and induced by situational judgment are located. In addition, the emotion of "pleasure" is located on the upper side of the concentric circles, and the emotion of "displeasure" is located on the lower side. Thus, in the emotion map 400, multiple emotions are mapped based on the structure in which emotions arise, and emotions that are likely to occur simultaneously are mapped close together.

[1688] These emotions are distributed at the 3 o'clock position on the Emotion Map 400, and usually fluctuate between feelings of security and anxiety. In the right half of the Emotion Map 400, situational awareness takes precedence over internal feelings, resulting in a calm impression.

[1689] The inside of the Emotion Map 400 represents inner thoughts, while the outside represents actions. Therefore, the further you go from the outside of the Emotion Map 400, the more visible (expressed in actions) your emotions become.

[1690] Here, human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. Similarly, in robots, cars, motorcycles, etc., emotions can be created based on various balances, such as posture and battery level. When these balances deviate from the ideal, it results in discomfort, and when they approach the ideal, it results in pleasure. The emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on a system for analyzing brain physiological signals of speech emotion recognition and emotion, Tokushima University, doctoral dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map contains emotions belonging to a region called "response," where sensation is dominant. The right half of the emotion map contains emotions belonging to a region called "situation," where situational awareness is dominant.

[1691] The emotion map defines two emotions that promote learning. One is the emotion around the middle of the negative "repentance" and "reflection" on the situation side. In other words, it is 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 the emotion around the positive "desire" on the reaction side. In other words, it is when the robot has positive feelings such as "I want more" or "I want to know more."

[1692] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​representing each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple training data sets, which are combinations of user input and emotion values ​​representing each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions located close together have similar values, as shown in the emotion map 900 in Figure 10. Figure 10 shows an example where multiple emotions such as "reassured," "calm," and "confident" have similar emotion values.

[1693] The above description primarily focuses on the functions of the data processing device 12 in relation to this disclosure. However, the system related to this disclosure is not necessarily implemented on a server. The system related to this disclosure may be implemented as a general information processing system. This disclosure may be implemented, for example, as a software program that runs on a personal computer or as an application that runs on a smartphone. The method related to this disclosure may be provided to users in SaaS (Software as a Service) format.

[1694] In the above embodiment, an example was given in which a specific process is performed by a single computer 22. However, the technology of this disclosure is not limited thereto, and a distributed processing of the specific process may be performed by multiple computers, including computer 22. For example, a data generation model 58 may be provided in an external device of the data processing device 12, and the external device may generate data according to the input data.

[1695] In the above embodiment, an example was given in which the specific processing program 56 is stored in the storage 32, but the technology of this disclosure is not limited thereto. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-temporary storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-temporary storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[1696] 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.

[1697] Furthermore, it is not necessary to store the entirety 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 the entirety of the specific processing program 56 in the storage 32; it is acceptable to store only a portion of the specific processing program 56.

[1698] The following types of processors can be used as hardware resources to perform specific processing. Examples of processors include a CPU, a general-purpose processor that functions as a hardware resource to perform specific processing by executing software, i.e., a program. Other examples of processors include dedicated electrical circuits, such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application Specific Integrated Circuits), which have circuit configurations specifically designed to perform specific processing. All of these processors have built-in or connected memory, and all of them perform specific processing by using memory.

[1699] The hardware resource that performs a specific process may consist of one of these various processors, or it may consist of 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). Alternatively, the hardware resource that performs a specific process may consist of a single processor.

[1700] Examples of configurations using a single processor include, firstly, a configuration in which one or more CPUs and software are combined to form a single processor, and this processor functions as a hardware resource that performs a specific process. Secondly, there is a configuration using a processor that realizes the functions of the entire system, including multiple hardware resources that perform a specific process, on a single IC chip, as exemplified by SoCs (System-on-a-chip). In this way, a specific process is realized using one or more of the above types of processors as hardware resources.

[1701] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits that combine circuit elements such as semiconductor devices. Also, the specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps can be deleted, new steps added, or the processing order rearranged, as long as it does not deviate from the main purpose.

[1702] The descriptions and illustrations presented above are detailed explanations of the technical aspects of this disclosure and are merely examples of the technical aspects. For example, the above descriptions of the structure, function, operation, and effect are examples of the structure, function, operation, and effect of the technical aspects of this disclosure. Therefore, it goes without saying that you may delete unnecessary parts, add new elements, or replace elements in the descriptions and illustrations presented above, as long as you do not deviate from the essence of the technical aspects of this disclosure. Furthermore, in order to avoid confusion and facilitate understanding of the technical aspects of this disclosure, explanations of common technical knowledge and the like that do not require special explanation to enable the implementation of the technical aspects of this disclosure have been omitted from the descriptions and illustrations presented above.

[1703] All documents, patent applications, and technical standards described herein are incorporated by reference to the same extent as if each individual document, patent application, and technical standard were specifically and individually noted to be incorporated by reference.

[1704] The following is further disclosed regarding the embodiments described above.

[1705] (Claim 1)

[1706] Means of collecting past conversation data,

[1707] means for preprocessing the collected conversation data,

[1708] A means for training a generative AI model using the preprocessed conversation data,

[1709] A means for analyzing an incoming message and generating multiple reply candidates using the generative AI model,

[1710] A means of sending the generated reply candidates to the user's device,

[1711] A means of sending the reply selected by the user,

[1712] A system that includes this.

[1713] (Claim 2)

[1714] The system according to claim 1, further comprising means for analyzing context based on past conversation data.

[1715] (Claim 3)

[1716] The system according to claim 1, further comprising means for incorporating sentiment analysis into the training of a generative AI model.

[1717] "Example 1"

[1718] (Claim 1)

[1719] Means by which users can authorize the collection of conversation history data,

[1720] A means of collecting past conversation data from a terminal and sending it to a server,

[1721] A method for preprocessing received conversation data to remove noise and clean the data,

[1722] A means for training a generative AI model using preprocessed conversational data,

[1723] Methods for training generative AI models by incorporating sentiment analysis,

[1724] A means of receiving new messages,

[1725] A means of sending received messages to the server,

[1726] A means of performing contextual analysis of a message, inputting the prompt into an AI model to generate multiple reply candidates,

[1727] A means of sending the generated reply candidates to the user's device,

[1728] A means for the user to select from the displayed reply suggestions,

[1729] A means of sending the selected reply to the recipient in the messaging application,

[1730] A system that includes this.

[1731] (Claim 2)

[1732] The system according to claim 1, further comprising means for analyzing context based on past conversation data and generating reply candidates based on the context.

[1733] (Claim 3)

[1734] The system according to claim 1, further comprising means for incorporating sentiment analysis into the training of a generative AI model and for storing the trained model.

[1735] "Application Example 1"

[1736] (Claim 1)

[1737] Means of collecting past conversation data,

[1738] means for preprocessing the collected conversation data,

[1739] A means for training a generative AI model using the preprocessed conversation data,

[1740] A means for analyzing an incoming message and generating multiple reply candidates using the generative AI model,

[1741] A means of sending the generated reply candidates to the user's device,

[1742] A means of sending the reply selected by the user,

[1743] A means of suggesting appropriate response options to order confirmation messages, order change request messages, or complaint handling messages when a user is using an online delivery service,

[1744] A system that includes this.

[1745] (Claim 2)

[1746] The system according to claim 1, further comprising means for analyzing context based on past conversation data.

[1747] (Claim 3)

[1748] The system according to claim 1, further comprising means for incorporating sentiment analysis into the training of a generative AI model.

[1749] "Example 2 of combining an emotion engine"

[1750] (Claim 1)

[1751] Means of collecting past conversation data,

[1752] means for preprocessing the collected conversation data,

[1753] A means for training a generative AI model using the preprocessed conversation data,

[1754] A means for analyzing an incoming message and generating multiple reply candidates using the generation AI model,

[1755] A means of sending the generated reply candidates to the user's device,

[1756] A means of sending the reply selected by the user,

[1757] A means of using an emotion engine to recognize the user's emotional state,

[1758] A means of generating reply suggestions based on the contextual information of the received message and the user's emotional state,

[1759] A system that includes this.

[1760] (Claim 2)

[1761] The system according to claim 1, further comprising means for analyzing context based on past conversation data.

[1762] (Claim 3)

[1763] The system according to claim 1, further comprising means for incorporating sentiment analysis into the training of a generative AI model.

[1764] "Application example 2 when combining with an emotional engine"

[1765] (Claim 1)

[1766] Means for collecting past dialogue data,

[1767] Means for preprocessing the collected dialogue data,

[1768] A means for training a generative AI model using the pre-processed dialogue data,

[1769] A means for analyzing a received message and generating multiple response candidates using the generative AI model,

[1770] A means for transmitting the generated response candidates to the user's display device,

[1771] Means for sending the response selected by the user,

[1772] A means of using an emotion engine to identify the customer's emotional state, comparing and analyzing it with past conversation history,

[1773] A system that includes this.

[1774] (Claim 2)

[1775] The system according to claim 1, further comprising means for analyzing context based on past dialogue data.

[1776] (Claim 3)

[1777] The system according to claim 1, further comprising means for incorporating sentiment analysis into the training of a generative AI model. [Explanation of Symbols]

[1778] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Devices 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robots< / url:> < / url:> < / url:> < / url:>

Claims

1. Means of collecting past conversation data, means for preprocessing the collected conversation data, A means for training a generative AI model using the preprocessed conversation data, A means for analyzing an incoming message and generating multiple reply candidates using the generative AI model, A means of sending the generated reply candidates to the user's device, A means of sending the reply selected by the user, A system that includes this.

2. The system according to claim 1, further comprising means for analyzing context based on past conversation data.

3. The system according to claim 1, further comprising means for incorporating sentiment analysis into the training of a generative AI model.

Images