System

Abstract

A system is provided.SOLUTION: A system includes means for collecting online conversation data from a user, means for analyzing the collected data to extract question and answer pairs, means for using the extracted question and answer pairs to fine-tune a natural language model, means for generating answers using the fine-tuned natural language model for questions from the user, and means for displaying the generated answers to the user.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] There is a problem in that it is difficult for people seeking advice to respond to others' consultations without time or location constraints. In particular, there are many situations where it is difficult to directly consult with superiors or senior colleagues, or to find appropriate time for consultations in the midst of busy daily work. In addition, it is practically difficult to frequently seek advice from people in important decision-making positions (e.g., CEOs), making it difficult to obtain effective advice. [Means for solving the problem]

[0005] The present invention includes a means for collecting online conversation data from users and a means for analyzing the collected data to extract question and answer pairs. This allows important conversation content to be organized and an appropriate dataset to be prepared. Then, by using the extracted question and answer pairs to fine-tune a natural language model, the system learns the response patterns of actual users and creates an avatar model that can provide highly accurate responses. Furthermore, by utilizing this avatar model, the system provides a means for generating answers to questions from users and displaying them to the users, enabling consultations to be provided regardless of location or time. The system also includes a means for collecting user feedback and continuously improving the model based on that feedback, and a means for anonymizing the data to ensure privacy protection for online conversation data. This provides an environment for safe and efficient consultations.

[0006] A "user" is a person who uses the system to seek advice or who provides online conversation data.

[0007] "Online conversation data" refers to chat logs and audio and text data from web conferences conducted online by users.

[0008] "Means of collection" refers to software or hardware used to obtain users' online conversation data.

[0009] The "analysis means" refers to software for analyzing the collected online conversation data and extracting question and answer pairs.

[0010] A "question and answer pair" is a set of information that includes a question and a corresponding answer.

[0011] A "natural language model" is a computer program that applies machine learning techniques to understand and generate human language.

[0012] "Fine-tuning" refers to the process of further training an existing natural language model with a specific dataset to optimize its performance for a specific task.

[0013] A "means for generating an answer to a question" is software that receives a question from a user as input and generates an appropriate answer using a natural language model.

[0014] "Means for displaying to the user" refers to an interface or software for displaying the generated answer on the user's terminal.

[0015] "Feedback" refers to information that allows users to provide evaluations and opinions on answers provided.

[0016] "Continuous improvement" refers to the process of retraining the model based on collected feedback to improve its performance.

[0017] "Privacy protection" refers to technologies and methods for preventing users' personal information and conversation content from being leaked to the outside.

[0018] "Anonymization methods" refers to software or technology that removes personally identifiable information from collected online conversation data and preserves the anonymity of the data. [Brief explanation of the drawings]

[0019] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4]FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0020] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0021] First, the terms used in the following description will be explained.

[0022] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0023] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0024] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0025] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0026] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0027] [First embodiment]

[0028] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0029] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0030] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0031] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0032] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0033] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0034] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0035] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0036] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0037] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0038] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0039] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0040] A system for implementing the present invention is configured by combining various software and hardware components for users to provide online conversation data and for analyzing that data.

[0041] Explanation of program processing

[0042] 1. Collecting online conversation data

[0043] What the device does: When a user has an online conversation or web conference, the device collects the conversation in real time and stores the collected data in text or audio format.

[0044] Example: An employee has a web conference with their manager about a particular project, and the content of the conference is collected.

[0045] 2. Data transmission and storage

[0046] What the device does: Collected data is encrypted and securely sent to the server.

[0047] Server behavior: The server stores the received data in a specific database with restricted access to ensure privacy.

[0048] 3. Data Analysis

[0049] Server operation: The server analyzes the received data and converts it into text using natural language processing (NLP) techniques. It then extracts important keywords and question / answer pairs from the converted text data.

[0050] Example: The server extracts questions and answers about "plans to enter new markets" from a conversation between a boss and an employee.

[0051] 4. Training the model

[0052] What the server does: It uses the extracted question and answer pairs to fine-tune an existing natural language model. Specifically, the fine-tuned model learns the user's unique answer patterns.

[0053] Example: The server uses the extracted data to train a natural language model to learn the speaking style of your boss.

[0054] 5. Provision of consultation system

[0055] User action: The user enters a question through the system's interface.

[0056] What the terminal does: Sends the entered question to the server.

[0057] What the server does: Analyzes the received question, uses a fine-tuned natural language model to generate the best answer, and sends this answer to the device.

[0058] Terminal behavior: Display the generated answer to the user.

[0059] For example: An employee asks, "How can I prepare for my next meeting?" and is provided with the best answer to that question.

[0060] 6. Feedback and model improvement

[0061] User Action: The user enters feedback on the answer provided.

[0062] Device Action: Sends feedback to the server.

[0063] Server action: The server analyzes the feedback and uses the results to retune the model and improve performance.

[0064] 7. Data anonymization

[0065] Device behavior: The data collected is anonymized to protect your privacy. This process removes any personally identifiable information, allowing you to use the system with confidence.

[0066] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized, so personal information will be protected.

[0067] With the above program and processing flow, the present invention is a system that allows users to consult with others and receive accurate answers regardless of time or place. This system is particularly useful for users who find it difficult to consult with superiors or seniors directly, and provides an environment where consultations can be conducted efficiently and safely.

[0068] The processing flow will be explained below.

[0069] Step 1:

[0070] Users can hold online conversations or web conferences, and the device can collect and store the conversations in real time in text or audio format.

[0071] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[0072] Step 2:

[0073] The device encrypts the collected data and sends it securely to a server, which then stores it in a database with access restrictions to ensure privacy.

[0074] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[0075] Step 3:

[0076] The server analyzes the received data and converts it into text using natural language processing (NLP) techniques, then extracts important keywords and question / answer pairs from the converted text data.

[0077] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[0078] Step 4:

[0079] The server uses the extracted question and answer pairs to fine-tune an existing natural language model (e.g., GPT-3®), learning user-specific answer patterns in the process.

[0080] Example: The server feeds the extracted dataset into a natural language model to learn the speaking style of your boss.

[0081] Step 5:

[0082] The user inputs a question through the system interface, and the terminal sends the user's question to the server.

[0083] Example: An employee types, "How can I prepare for my next meeting?"

[0084] Step 6:

[0085] The server analyzes the user's question, uses a fine-tuned natural language model to generate the best answer, and sends the generated answer to the device.

[0086] Example: The server invokes a natural language model to answer "how to prepare for my next meeting."

[0087] Step 7:

[0088] The terminal displays the answer sent from the server to the user. Example: An employee's terminal displays the answer "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[0089] Step 8:

[0090] The user enters feedback on the provided answer, and the device collects the feedback and sends it to the server.

[0091] Example: An employee gives feedback saying, "This answer was helpful."

[0092] Step 9:

[0093] The server analyzes the feedback and retunes the model based on the results to improve performance.

[0094] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[0095] Step 10:

[0096] The device anonymizes the data collected to protect privacy, a process that removes any personally identifiable information.

[0097] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[0098] Example 1

[0099] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0100] Conventional online conversation systems make it difficult for users to seek advice or obtain information in real time, and information sharing is hindered, especially in environments where it is difficult to ask questions directly to superiors or senior colleagues. Furthermore, the quality of answers is inconsistent, preventing users from accurately obtaining useful information. Furthermore, from the perspective of privacy protection, secure data collection and analysis are required. Therefore, there is a need for the development of a system that can effectively collect and analyze online conversation data, provide appropriate answers to users, and ensure privacy protection.

[0101] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0102] In this invention, the server includes means for collecting online communication data from users, means for encrypting the collected data and securely transmitting it to the server, means for analyzing the received data and converting it into text using natural language processing technology, means for extracting important keywords and question-and-answer pairs from the converted text data, means for fine-tuning an existing generative AI model using the extracted question-and-answer pairs, means for accepting questions from users and generating optimal answers using the fine-tuned generative AI model, and means for displaying the generated answers to the users. This makes it possible to effectively collect and analyze online conversation data and provide appropriate answers to users.

[0103] "Online communication data" refers to data related to communication such as conversations and text messages that users have online.

[0104] "Encryption" is the process of transforming data using a specific algorithm to protect it from unauthorized access.

[0105] A "server" is a computer that provides various services within a network, and refers to hardware that runs software with specific functions.

[0106] "Natural language processing technology" refers to technology for processing and analyzing human language using computers, and includes tokenization, part-of-speech tagging, named entity recognition, etc.

[0107] "Convert to text" refers to the process of converting non-text data such as audio or images into data in character string format.

[0108] "Keywords" are words or phrases that contain particularly important information in the data, and are elements that serve as clues for understanding the data.

[0109] A "question and answer pair" is a dataset that combines a user's question with an appropriate answer.

[0110] A "generative AI model" is an artificial intelligence model that generates new information based on given input data.

[0111] "Fine-tuning" refers to the process of optimizing an existing artificial intelligence model to suit a specific task or user needs, and involves additional training.

[0112] "Feedback" refers to evaluations and opinions on the information and answers provided by users, which are useful for improving the system.

[0113] "Anonymization" is the removal of personally identifiable information from data, and is a privacy-protecting technique.

[0114] MODE FOR CARRYING OUT THE INVENTION

[0115] The system for carrying out the present invention is configured by combining various software and hardware for users to provide online conversation data and for analyzing the data.

[0116] Hardware and Software Configuration

[0117] Device configuration

[0118] A terminal is a device used by users to conduct online conversations and web conferences, and can be a PC, smartphone, tablet, etc. The terminal has voice recognition software (e.g., Google® Speech-to-Text API) and encryption software (e.g., AES encryption) installed.

[0119] Server Configuration

[0120] The server is a computer used to analyze and store collected data and fine-tune the generative AI model. The following software is installed on the server:

[0121] Natural language processing technology (e.g., SpaCy)

[0122] Database system (e.g., MySQL (registered trademark))

[0123] Generative AI models (e.g., GPT-3)

[0124] A concrete example of the processing flow

[0125] 1. Collecting online conversation data

[0126] The device collects conversations in real time when users are having online conversations or web conferences. The conversations are captured as voice data and converted into text data using voice recognition software. For example, when an employee discusses "sales strategies for a new product" with their boss in a web conference, the voice data is converted into text.

[0127] 2. Data transmission and storage

[0128] The collected text data is encrypted and sent to the server using AES encryption technology and HTTPS protocol, where it is decrypted and stored in a MySQL database.

[0129] 3. Data Analysis

[0130] The server analyzes the text data using natural language processing technology (e.g., SpaCy) and extracts important keywords and question-and-answer pairs. For example, questions and answers related to "sales strategies for new products" are extracted. The results of this analysis are also stored in a database.

[0131] 4. Training the model

[0132] The server uses the extracted question-answer pairs to fine-tune a generative AI model (e.g., GPT-3). This process optimizes the model over multiple epochs and learns the user's unique answer patterns. For example, the model is tuned to learn the speaking style of a boss.

[0133] 5. Provision of consultation system

[0134] Users input questions into the system through a dedicated interface. The input question is encoded and sent to the server. The server analyzes the question and uses a fine-tuned generative AI model to generate the best answer, which is then sent to the device. The device then displays the answer to the user. For example, if an employee asks, "How should I prepare for my next meeting?", the best answer will be displayed.

[0135] 6. Feedback and model improvement

[0136] Users input feedback on the answers they provide through the interface, which is sent to the server for analysis, and the server uses the feedback to retune the generative AI model and improve its performance.

[0137] 7. Data anonymization

[0138] The device anonymizes the collected data to protect privacy. This is a process that removes any personally identifiable information. For example, "consultation about interpersonal relationships at work" is anonymized to protect personal information.

[0139] Examples of prompt statements

[0140] Examples of prompts used in this system include:

[0141] "If you were to extract questions and answers about a sales strategy for a new product during a web conference with your boss and ask about preparations for the next meeting, what would your answer be?"

[0142] "Generate relevant advice from anonymized consultations about workplace relationships."

[0143] The above is an embodiment of the invention.

[0144] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0145] Step 1: Collect online conversation data

[0146] The device collects the content of conversations in real time when users are having online conversations or web conferences. The content of the conversation is captured as audio data and converted into text data using speech recognition software (e.g., Google Speech-to-Text API). Specifically, the collected audio data is input into the speech recognition software and output as text data. For example, when an employee discusses "sales strategies for a new product" with his or her boss in a web conference, the audio data is converted into text data.

[0147] Step 2: Send and store data

[0148] The terminal encrypts the collected text data (e.g., AES encryption). The encrypted data is sent to the server via the network. The server decrypts the received encrypted data and stores it in a database (e.g., MySQL). Specifically, the data is entered into a database system and stored as decrypted text data from its encrypted state. For example, the contents of a meeting may be encrypted and sent to the server, where it is decrypted and stored.

[0149] Step 3: Analyze the data

[0150] The server analyzes the stored data using natural language processing technology (e.g., SpaCy). Important keywords and question-and-answer pairs are extracted from the text data. Specifically, the text data is input into the natural language processing software, which then outputs the analyzed keywords and question-and-answer pairs. For example, questions and answers related to "sales strategies for new products" are extracted.

[0151] Step 4: Train the model

[0152] The server uses the extracted question-answer pairs to fine-tune a generative AI model (e.g., GPT-3). This fine-tuning allows the model to learn the user's unique answer patterns. Specifically, the question-answer pair data is input into the training interface of the generative AI model, and a fine-tuned model is output. For example, the model is adjusted to learn the speaking style of a boss.

[0153] Step 5: Providing a consultation system

[0154] A user inputs a question via a system interface (e.g., a web application). The device encodes the input question and sends it to the server. The server analyzes the received question and generates an optimal answer using a fine-tuned generative AI model. The answer is sent to the device, which then displays the generated answer to the user. Specifically, the system encodes the user input and sends it to the server, then decodes and displays the generated answer. For example, if an employee asks, "How should I prepare for the next meeting?", the optimal answer will be displayed.

[0155] Step 6: Feedback and model improvement

[0156] The user inputs feedback on the provided answer via the interface. The device encodes the feedback data and sends it to the server. The server analyzes the feedback and readjusts the generative AI model based on the results, improving its performance. Specifically, the feedback data is input into the analysis system, which outputs a readjusted model. For example, an evaluation of the appropriateness of the provided answer can be input to improve the quality of the model.

[0157] Step 7: Anonymize the data

[0158] The device anonymizes the collected data to protect privacy. This process removes any personally identifiable information. The server stores and analyzes the anonymized data to ensure the user's privacy. Specifically, the system filters personal information from the data and uses an anonymization tool to replace proper nouns with generic terms. For example, "consultation about interpersonal relationships at work" is anonymized, protecting personal information.

[0159] (Application example 1)

[0160] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0161] In today's information society, online communication is increasing. In this context, it is essential for companies and individuals to manage not only knowledge but also security risks. However, there is a lack of systems that can detect security risks in real time and respond immediately. The objective of this invention is to provide a system that collects and analyzes online conversation data, analyzes security risks in real time, and notifies users.

[0162] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0163] In this invention, the server includes means for collecting online conversation data from users, means for analyzing the collected data to extract question and answer pairs, means for fine-tuning a natural language model using the extracted question and answer pairs, means for generating answers to questions from users using the fine-tuned natural language model, means for collecting online conversation data in real time and analyzing security risks, and means for notifying users of risk alerts based on the analysis results. This makes it possible to efficiently manage online conversation data, detect security risks in real time, and respond immediately.

[0164] "Means for collecting online conversation data from users" refers to a combination of hardware and software for collecting the content of online meetings and conversations conducted by users in real time and saving it in text or audio format.

[0165] "Means for analyzing collected data to extract question and answer pairs" refers to software processes that apply natural language processing techniques to collected online conversation data to identify and extract key question and answer pairs.

[0166] "Means for fine-tuning natural language models using extracted question and answer pairs" refers to techniques for incrementally training existing natural language models using extracted data to improve their accuracy and responsiveness in specific contexts.

[0167] "Means for generating answers to questions from users using a fine-tuned natural language model" refers to algorithms and processes for generating optimal answers to questions entered by users using a fine-tuned natural language model.

[0168] "Means for displaying generated answers to a user" refers to an interface and software for displaying answers generated by a natural language model on a user's device.

[0169] "Means for collecting online conversation data in real time and analyzing security risks" refers to a combination of hardware and software for collecting online conversation data in real time and analyzing the data to detect security risks.

[0170] "Means for notifying users of risk alerts based on the analysis results" refers to alert systems and software that immediately notify users of security risks detected from the analyzed data.

[0171] The system for implementing this invention is configured by combining various software and hardware for allowing users to collect online conversation data in real time using smartphones or other devices and analyze that data.

[0172] System Overview

[0173] Acquisition System:

[0174] The device collects the conversation content in real time when the user is having an online conversation or web conference. The collected data is saved in text or audio format. Specifically, the voice data is converted to text using the Google Web Speech API. The collected text data is securely transmitted from the user's device to a server.

[0175] Data transmission and storage:

[0176] The device encrypts the collected data and sends it securely to a server, which then stores it in a specific database with access restrictions to ensure privacy.

[0177] Data Analysis:

[0178] The server parses the incoming data and converts it into text using natural language processing (NLP) techniques. From the converted text data, question and answer pairs are extracted. This extracted data is used to fine-tune an existing natural language model.

[0179] Security Risk Analysis:

[0180] The collected online conversation data is monitored in real time and security risks are analyzed. Risks are determined based on specific trigger words (e.g., "password," "confidential," etc.). This is done using an NLP model.

[0181] Risk Alert Notification:

[0182] Based on the analysis results, users are immediately notified of any security risks that are discovered. Once the data has been analyzed, the server sends an alert to the user along with a risk assessment. This notification is sent via an interface such as a push notification on a smartphone.

[0183] User Interface:

[0184] Users input questions through the interface, and the best answer to that question is generated by a fine-tuned natural language model. The generated answer is displayed on the user's device. In addition, users can input feedback on the provided answer, which is sent to the server and used to refine the model.

[0185] Examples and prompts

[0186] For example, if a user is using a project management system and during a conversation they say, "The budget for the next project has been decided to be 5 million yen," the system will collect and analyze that conversation in real time to determine whether it contains confidential information.

[0187] An example prompt is:

[0188] Online conversation data: "The budget for the next project has been decided to be 5 million yen."

[0189] Prompt: "Is this conversation a security risk?"

[0190] This system allows users to instantly identify potential security risks that arise during conversations and take appropriate measures.

[0191] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0192] Step 1:

[0193] The device collects conversation content in real time when a user is engaged in online conversations or web conferences. It receives voice or text data as input and converts it into text using speech recognition software (e.g., Google Web Speech API). The output is text-format conversation data.

[0194] Step 2:

[0195] The device encrypts the collected text data using a secure protocol (e.g., HTTPS) and securely sends it to the server. It receives text conversation data as input, encrypts it, and then sends it to the server. The output is encrypted text data.

[0196] Step 3:

[0197] The server stores the received data in a specific database. It receives encrypted text data as input, decrypts it, and stores it in the database. The output is the text data stored in the database.

[0198] Step 4:

[0199] The server analyzes the stored data and extracts question and answer pairs using natural language processing (NLP) techniques. It receives text data as input and applies NLP algorithms to extract important keywords and question and answer pairs. The output is the extracted question and answer pairs.

[0200] Step 5:

[0201] The server uses the extracted question and answer pairs to fine-tune the natural language model. It takes the question and answer pairs as input and retrains the generative AI model. The output is a fine-tuned natural language model.

[0202] Step 6:

[0203] The terminal receives the user's question through the system interface, and sends it to the server in text format. The output is the question in text format.

[0204] Step 7:

[0205] The server analyzes the received question and generates the best answer using a fine-tuned natural language model. It takes the user's question as input and generates an answer using a generative AI model. The output is the generated answer in text format.

[0206] Step 8:

[0207] The terminal displays the generated answer to the user. It takes as input the generated textual answer and displays it in a user interface. The output is the answer that is displayed to the user.

[0208] Step 9:

[0209] The server monitors the collected online conversation data in real time and analyzes security risks. It receives the conversation data as input and evaluates the security risk using an NLP model. The output is the security risk assessment result.

[0210] Step 10:

[0211] The server notifies the user of a risk alert based on the analysis results. It receives the security risk assessment results as input, generates a message for the alert notification, and sends it to the terminal. The output is an alert message that is notified to the user.

[0212] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0213] The system for implementing this invention has the ability to take emotional state into account when generating answers by allowing users to provide online conversation data, analyzing that data, and combining it with an emotion engine that recognizes the user's emotions.

[0214] Explanation of program processing

[0215] 1. Collecting online conversation data

[0216] Device operation: When a user engages in online conversations or web conferences, the device collects the conversation content in real time and stores it in text or audio format.

[0217] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[0218] 2. Data transmission and storage

[0219] What the device does: Collected data is encrypted and sent securely to a server, which stores the data in a database with access restrictions to ensure privacy.

[0220] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[0221] 3. Data Analysis

[0222] Server operation: The server analyzes the received data and converts it into text using natural language processing (NLP) techniques. It then extracts important keywords and question / answer pairs from the converted text data.

[0223] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[0224] 4. Emotion analysis using an emotion engine

[0225] Server operation: The server analyzes the user's emotional state based on the collected data using an emotion engine. Based on the analysis results, it assigns emotion labels to the conversation data.

[0226] Example: The server assigns emotion labels such as "anxiety" or "excitement" based on the user's speech.

[0227] 5. Training the model

[0228] What the server does: It uses the extracted question-answer pairs and sentiment labels to fine-tune an existing natural language model, learning user-specific answer patterns along with their emotional states in the process.

[0229] Example: The server learns the speaking style and corresponding emotional state of a boss and trains a natural language model (e.g., GPT-3).

[0230] 6. Provision of consultation system

[0231] User action: The user inputs a question through the system interface. The terminal sends the user's question to the server.

[0232] Example: An employee types, "How can I prepare for my next meeting?"

[0233] 7. Answer Generation

[0234] Server operation: The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. The answer is generated based on the user's emotional state. The generated answer is then sent to the device.

[0235] Example: The server provides an "anxious" employee with a response such as, "To prepare for the next meeting, first check the agenda and gather any relevant materials. If you have any concerns, please check with your leader."

[0236] 8. View Answers

[0237] What the device does: The device displays the answer sent by the server to the user.

[0238] Example: The employee's device displays the following response: "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[0239] 9. Feedback and Model Improvement

[0240] User action: The user enters feedback on the provided answer. The device collects the feedback and sends it to the server.

[0241] Server action: The server analyzes the feedback and retunes the model based on the results to improve performance.

[0242] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[0243] 10. Data Anonymization

[0244] What the device does: The data collected is anonymized to protect your privacy. This process removes any personally identifiable information.

[0245] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[0246] With the above program and processing flow, the present invention is a system that allows users to consult with others and receive appropriate answers regardless of time or place. This system is particularly useful for users who find it difficult to consult directly with superiors or seniors, and provides an environment where consultations can be conducted efficiently and safely while taking into account their emotional state.

[0247] The processing flow will be explained below.

[0248] Step 1:

[0249] Users can hold online conversations or web conferences, and the device can collect and save the conversations in real time as text or audio.

[0250] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[0251] Step 2:

[0252] The device encrypts the collected data and sends it securely to a server, where it is stored in a specific database with access restrictions to ensure privacy.

[0253] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[0254] Step 3:

[0255] The server analyzes the received data and converts it into text using natural language processing (NLP) techniques, then extracts important keywords and question / answer pairs from the converted text data.

[0256] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[0257] Step 4:

[0258] The server analyzes the user's emotional state based on the collected data using an emotion engine, and assigns emotion labels to the conversation data based on the analysis results.

[0259] Example: The server assigns emotion labels such as "anxiety" or "excitement" based on the user's speech.

[0260] Step 5:

[0261] The server uses the extracted question and answer pairs and sentiment labels to fine-tune an existing natural language model, learning the user's unique answer patterns along with their emotional state in the process.

[0262] Example: The server learns the speaking style and corresponding emotional state of a boss and trains a natural language model.

[0263] Step 6:

[0264] The user inputs a question through the system interface, and the terminal sends the user's question to the server.

[0265] Example: An employee types, "How can I prepare for my next meeting?"

[0266] Step 7:

[0267] The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. The answer is generated based on the user's emotional state. The generated answer is then sent to the device.

[0268] Example: The server provides an "anxious" employee with a response such as, "To prepare for the next meeting, first check the agenda and gather any relevant materials. If you have any concerns, please check with your leader."

[0269] Step 8:

[0270] The terminal displays the answer sent from the server to the user.

[0271] Example: The employee's device displays the following response: "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[0272] Step 9:

[0273] The user enters feedback on the provided answer, and the device collects the feedback and sends it to the server.

[0274] Example: An employee gives feedback saying, "This answer was helpful."

[0275] Step 10:

[0276] The server analyzes the feedback and retunes the model based on the results to improve performance.

[0277] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[0278] Step 11:

[0279] The device anonymizes the data collected to protect privacy, a process that removes any personally identifiable information.

[0280] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[0281] Example 2

[0282] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0283] Current consultation systems that use online conversation data generate answers without considering the user's emotional state, which can result in the system failing to provide the most appropriate answer for the user. Furthermore, data privacy protection is insufficient, posing a risk of personal information leaks. The objective of this invention is to solve these problems and provide a system that generates appropriate answers that consider the user's emotional state and protects data privacy.

[0284] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0285] In this invention, the server includes terminal means for collecting online conversation data from users, terminal means for encrypting the collected data and transmitting it to the server in a secure manner, database means, means for the server to analyze the received data using natural language processing technology and extract question and answer pairs, means for the server to analyze the user's emotional state from the data using a sentiment analysis engine and assign sentiment labels, server means for fine-tuning a natural language model using the extracted question and answer pairs and sentiment labels, terminal means for accepting questions from users, means for the server to generate answers to the user's questions using the fine-tuned natural language model and sentiment analysis engine, and terminal means for displaying the generated answers to the users. This makes it possible to provide appropriate answers that take the user's emotional state into consideration and protect the privacy of the data.

[0286] "Terminal means" refers to electronic devices that users use to collect online conversation data and input questions.

[0287] "Server" refers to the computer system that receives, analyzes, and stores collected data.

[0288] "Encryption" is a technology that converts data using a specific algorithm to protect it from unauthorized access or tampering.

[0289] "Server means" refers to the hardware or software used by the server to perform a particular function.

[0290] "Database means" refers to a system for safely storing received data and making it accessible when necessary.

[0291] "Natural language processing technology" is a technology for analyzing and understanding human language, and primarily uses machine learning and language models.

[0292] A "question and answer pair" refers to a set of questions entered by a user and their corresponding appropriate answers.

[0293] An "emotion analysis engine" refers to technology that analyzes a user's emotional state from collected data and assigns various emotional labels.

[0294] An "emotion label" is a tag attached to data to indicate the user's emotional state, such as "joy," "sadness," or "anxiety."

[0295] "Fine-tuning" is the process of fine-tuning an existing model to suit a specific dataset or purpose.

[0296] "Natural language model" refers to a machine learning model trained to generate and parse human language.

[0297] "Feedback" refers to the evaluation or opinion a user gives on a provided answer.

[0298] "Anonymization" is a technique that removes or transforms personally identifiable information to protect the privacy of data.

[0299] "Display means" refers to a device or software for visually presenting the generated answer to the user.

[0300] An embodiment of this invention is described below. This system is mainly composed of three elements: a server, a terminal, and a user. This system allows users to provide online conversation data, analyzes that data, and combines it with an emotion engine that recognizes the user's emotions, thereby providing a function that takes the user's emotional state into account when generating answers.

[0301] Device behavior

[0302] When a user has an online conversation or web conference, the device collects the conversation in real time using the device's microphone and camera. The collected data is saved in audio or video format and immediately converted into text format. The data is then encrypted and sent to the server in a secure manner. The encryption techniques used include AES-256 encryption and SSL / TLS protocol.

[0303] Server Operation

[0304] The server stores the received data in a database. Access restrictions are set to this database to ensure privacy. The data received by the server is analyzed using natural language processing (NLP) technology to extract important keywords and question-and-answer pairs. This analysis is performed using software such as Apache (registered trademark) Kafka and TENSORFLOW (registered trademark). The extracted text data is further analyzed using a sentiment analysis engine (e.g., IBM Watson (registered trademark) or Amazon Comprehend), and emotional labels are assigned to the data. Emotional labels such as "happiness," "sadness," and "anxiety" are taken into consideration.

[0305] Training the model

[0306] The server uses the extracted question and answer pairs and emotion labels to fine-tune a natural language model. This process uses Hugging Face's transformers library and OpenAI's GPT-3. The fine-tuned model learns the user's unique answer patterns and emotional state, allowing it to generate more appropriate answers.

[0307] Providing answers to users

[0308] Users input questions through the system interface (web app or mobile app). The device sends the user's question to the server, which analyzes it and generates the best answer using a fine-tuned natural language model and sentiment analysis engine. The generated answer is sent to the device and displayed to the user. For example, if an employee inputs "What should I do to prepare for the next meeting?", the server will provide a response such as "To prepare for the next meeting, first check the agenda and gather relevant materials."

[0309] Feedback and model improvement

[0310] Users can enter feedback on the answers provided, and the device collects and sends the feedback to the server, which analyzes the feedback and retunes the model based on the results to improve performance.

[0311] Data anonymization

[0312] The data collected is anonymized to protect your privacy - in this process, any personally identifiable information is removed and it is stored securely.

[0313] Examples of concrete examples and prompts

[0314] Example: After a meeting with their boss, an employee asks the system, "How should I approach the challenges in my new department?" The system recognizes that the employee is feeling anxious and provides advice that takes their feelings into consideration.

[0315] Example prompt: "I've just been assigned to a new department and I'm not sure how to solve a problem. Could you please give me some advice?"

[0316] As a result, this system provides appropriate answers that take into account the user's emotional state, protecting the privacy of data while providing an environment in which users can seek advice efficiently.

[0317] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0318] Step 1:

[0319] Collecting online conversation data

[0320] Device operation: When a user has an online conversation or web conference, the device's microphone and camera capture what is being said in real time and save it as audio or video data.

[0321] Input: User audio and video data.

[0322] Data processing: The device converts audio and video data into text format.

[0323] Output: Conversation data in text format.

[0324] What it does: Your device's microphone records your "discussion about a new project" and uses speech recognition software to convert it into text.

[0325] Step 2:

[0326] Data transmission and storage

[0327] What the device does: Encrypts the collected text data and sends it to the server in a secure manner.

[0328] Input: Conversation data in text format.

[0329] Data processing: Data is encrypted with AES-256 and transmitted using the SSL / TLS protocol.

[0330] Output: The encrypted data is sent to the server.

[0331] Specific operation: Text data collected on the device is encrypted with AES-256 and sent via SSL to the server, which then stores the received data in a database.

[0332] Step 3:

[0333] Data analysis

[0334] Server operation: The server analyzes the received text data using natural language processing technology and extracts important keywords and question-and-answer pairs.

[0335] Input: Conversation data in text format.

[0336] Data processing: Analyze using natural language processing techniques (e.g., Apache Kafka, TensorFlow) to extract keywords and question-answer pairs.

[0337] Output: Important keywords and question-answer pairs are extracted.

[0338] What it does: The server extracts keywords such as "plans to enter new markets" from the text data and identifies related questions and answers.

[0339] Step 4:

[0340] Emotion analysis using an emotion engine

[0341] Server operation: The server uses an emotion analysis engine to analyze the user's emotional state based on the data and assigns an emotion label.

[0342] Input: Keywords and question-answer pairs.

[0343] Data processing: Analyze the emotional state using a sentiment analysis engine (e.g., IBM Watson, Amazon Comprehend) and assign sentiment labels to the data.

[0344] Output: Data with emotion labels.

[0345] Specific operation: The server analyzes the emotion "anxiety" from statements such as "The project is progressing slowly" and assigns a label to it.

[0346] Step 5:

[0347] Training the model

[0348] Server operation: Fine-tune a natural language model using extracted question-answer pairs and sentiment labels.

[0349] Input: Question-answer pairs and sentiment labels.

[0350] Data processing: Fine-tune the model using Hugging Face's transformers library and OpenAI's GPT-3.

[0351] Output: A fine-tuned natural language model.

[0352] Specific operation: The server learns answers that adapt to the employee's question patterns based on guidelines and emotion labels.

[0353] Step 6:

[0354] Providing a consultation system

[0355] User action: The user enters a question through the system interface (web app or mobile app). The device sends the user's question to the server.

[0356] Input: The user's question.

[0357] Data processing: The user's question is formulated and sent to the server.

[0358] Output: The user's question sent to the server.

[0359] What happens: An employee types, "How do I prepare for a meeting?" and the device sends the question to the server.

[0360] Step 7:

[0361] Generate answers

[0362] Server operation: The server analyzes the user's question and generates the best answer using a fine-tuned natural language model and sentiment analysis engine based on the sentiment labels.

[0363] Input: User question and sentiment label.

[0364] Data processing: Generate answers using fine-tuned natural language models.

[0365] Output: The generated answer.

[0366] Specific operation: The server receives a question about "preparing for the meeting," generates a response such as "First, check the agenda and gather the materials," and sends it to the device.

[0367] Step 8:

[0368] Show Answers

[0369] What the device does: The device displays the answer sent by the server to the user.

[0370] Input: The answer sent by the server.

[0371] Data processing: Display the answer to the user visually or audibly.

[0372] Output: The answer displayed to the user.

[0373] What it does: The device prompts the employee, "To prepare for your next meeting, first review the agenda and gather any relevant materials."

[0374] Step 9:

[0375] Feedback and model improvement

[0376] User action: The user enters feedback on the provided answer. The device sends the feedback to the server.

[0377] Input: User feedback.

[0378] Data processing: Analyze feedback and use it to improve the model.

[0379] Output: An improved natural language model.

[0380] Specific behavior: The user enters feedback such as "This answer was helpful," and the device sends the feedback to the server. The server analyzes the feedback and recalibrates the model.

[0381] Step 10:

[0382] Data anonymization

[0383] What the device does: Anonymizes collected data to protect privacy.

[0384] Input: Conversation data, including personal information.

[0385] Data processing: removing or transforming personally identifiable information.

[0386] Output: Anonymized data.

[0387] What it does: It removes and securely stores your name and unique identifiers.

[0388] (Application example 2)

[0389] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0390] When it comes to communication between workers and robots in factories, there is a lack of a way to understand the emotions of workers and respond appropriately. This can have a negative impact on productivity and safety, especially in emergencies or stressful situations. To solve this problem, a system is needed that can analyze the emotional state of workers and respond appropriately according to that state.

[0391] The identification process by the identification 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 online conversation data from users, means for analyzing the collected data to extract question and answer pairs, means for fine-tuning a natural language model using the extracted question and answer pairs, means for analyzing the emotional state of the user using an emotion engine based on the collected data, means for generating an adapted answer based on the emotional state, and means for displaying the generated answer to the user. This makes it possible to understand the emotional state of employees and respond appropriately.

[0392] "User" refers to the factory employees who use the system.

[0393] "Online conversation data" refers to one-to-one voice or text interactions between employees and robots.

[0394] "Means of collection" refers to hardware and software that records or records online conversations between employees and robots.

[0395] "Means for analyzing and extracting question and answer pairs" refers to technology that uses natural language processing technology to analyze collected online conversation data and identify important questions and their answers.

[0396] "Fine-tuning" refers to the process of optimizing the performance of an existing natural language model using extracted question and answer pairs.

[0397] "Emotion engine" refers to software used to analyze users' emotions from online conversation data and identify their emotional states.

[0398] "Means for analyzing emotional state" refers to the process for determining a user's emotions from collected data using an emotion engine.

[0399] "Means for generating adapted answers" refers to techniques for generating appropriate and optimal answers based on the user's emotional state.

[0400] "Means for displaying" refers to a display or audio output device for visually or audibly communicating the generated response to the employee.

[0401] A system for implementing this invention allows a user to provide online conversation data, analyzes the data, and combines an emotion engine that recognizes the user's emotions with a natural language model to take the user's emotional state into account when generating responses. This system is particularly applicable to factory robots and is designed to facilitate smooth communication between factory workers and robots.

[0402] Hardware and Software Configuration

[0403] Hardware: Robots, microphones, and servers located in the factory

[0404] Software: Natural language processing engines (e.g., Transformer's NLP), sentiment analysis engines (e.g., Sentiment Analysis Pipeline), generative AI models (e.g., GPT-3)

[0405] Specific explanation of the system's operation

[0406] Collection and transmission of online conversation data

[0407] The robot collects online conversations with employees in real time via a microphone, and the collected voice data is encrypted and securely sent to a server.

[0408] Data analysis and text conversion

[0409] The server analyzes the received voice data and converts it into text using natural language processing (NLP) techniques, extracting important keywords and question / answer pairs from the converted text data.

[0410] Emotion analysis using an emotion engine

[0411] Next, the server analyzes the user's emotional state using an emotion engine and assigns emotion labels to the conversation data, thereby understanding the user's emotional state.

[0412] Training the model

[0413] The extracted question-answer pairs and sentiment labels are used to fine-tune an existing natural language model (e.g., GPT-3), learning user-specific answer patterns along with their emotional states in the process.

[0414] Responding to questions

[0415] When an employee types a question through the system's interface, the robot sends it to the server, which uses a fine-tuned natural language model and emotion engine to generate the best answer and communicate it to the user via a display or voice output device.

[0416] Specific examples

[0417] For example, if an employee in the quality control department asks the robot, "I don't know how to inspect the product," the robot's emotion analysis engine will recognize that the employee is feeling impatient. The generative AI model will respond by offering guidance that is adapted to the employee's impatient state: "Please stay calm. Let's inspect the product by following these steps."

[0418] Prompt Sentence Examples

[0419] "If an employee notices something is wrong but is feeling anxious, how do I support them?"

[0420] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0421] Step 1:

[0422] The terminal collects online conversation data. Specifically, the factory robot records the conversation with the employee in real time through a microphone. The collected data is in the form of audio, and the audio data is temporarily stored on the terminal for further processing.

[0423] Step 2:

[0424] The device encrypts the collected voice data and securely transmits it to a server using a secure communication protocol (e.g., HTTPS). The server then stores the received voice data in a database with access restrictions to ensure privacy.

[0425] Step 3:

[0426] The server analyzes the received voice data and converts it into text using natural language processing (NLP) techniques, such as using Transformers NLP models to convert speech to text, and stores the resulting text for use in the next step.

[0427] Step 4:

[0428] The server extracts important keywords and question / answer pairs from the converted text data. It uses natural language processing technology to identify which parts are questions and which parts are answers and extracts those pairs. This allows it to understand what information the user is looking for.

[0429] Step 5:

[0430] The server uses an emotion engine to analyze the user's emotional state based on the collected text data. An emotion analysis engine (e.g., Sentiment Analysis Pipeline) is used to identify emotions such as "anxiety" or "impatience" from the user's comments. Based on the results of this analysis, an emotion label is assigned to the text data.

[0431] Step 6:

[0432] The server uses the extracted question and answer pairs and sentiment labels to fine-tune existing natural language models, such as using GPT-3 to learn employee speech patterns and their corresponding emotional states. This process produces a model that understands the user's unique communication style.

[0433] Step 7:

[0434] The user inputs a question through the system interface. The factory robot recognizes this question and sends it to the server. The question is then analyzed in the next step.

[0435] Step 8:

[0436] The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. For example, if the user is in an "anxious" emotional state, the server will provide a calm tone and specific advice that is adapted to that state. The generated answer is then sent to the device.

[0437] Step 9:

[0438] The terminal displays the response sent from the server to the user, and the factory robot communicates the response to the employee via voice, allowing the user to take appropriate action.

[0439] Step 10:

[0440] The user provides feedback on the answers provided by the device, which collects this feedback and sends it to the server, which analyzes it and uses the results to retune the model and improve the system's performance.

[0441] Step 11:

[0442] The device anonymizes the data collected to protect your privacy, ensuring that personally identifiable information is removed and stored securely.

[0443] Example prompt sentence:

[0444] "If an employee notices something is wrong but is feeling anxious, how do I support them?"

[0445] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0446] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0447] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0448] [Second embodiment]

[0449] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0450] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0451] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0452] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0453] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0454] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0455] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0456] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0457] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0458] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0459] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0460] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0461] A system for implementing the present invention is configured by combining various software and hardware components for users to provide online conversation data and for analyzing that data.

[0462] Explanation of program processing

[0463] 1. Collecting online conversation data

[0464] What the device does: When a user has an online conversation or web conference, the device collects the conversation in real time and stores the collected data in text or audio format.

[0465] Example: An employee has a web conference with their manager about a particular project, and the content of the conference is collected.

[0466] 2. Data transmission and storage

[0467] What the device does: Collected data is encrypted and securely sent to the server.

[0468] Server behavior: The server stores the received data in a specific database with restricted access to ensure privacy.

[0469] 3. Data Analysis

[0470] Server operation: The server analyzes the received data and converts it into text using natural language processing (NLP) techniques. It then extracts important keywords and question / answer pairs from the converted text data.

[0471] Example: The server extracts questions and answers about "plans to enter new markets" from a conversation between a boss and an employee.

[0472] 4. Training the model

[0473] What the server does: It uses the extracted question and answer pairs to fine-tune an existing natural language model. Specifically, the fine-tuned model learns the user's unique answer patterns.

[0474] Example: The server uses the extracted data to train a natural language model to learn the speaking style of your boss.

[0475] 5. Provision of consultation system

[0476] User action: The user enters a question through the system's interface.

[0477] What the terminal does: Sends the entered question to the server.

[0478] What the server does: Analyzes the received question, uses a fine-tuned natural language model to generate the best answer, and sends this answer to the device.

[0479] Terminal behavior: Display the generated answer to the user.

[0480] For example: An employee asks, "How can I prepare for my next meeting?" and is provided with the best answer to that question.

[0481] 6. Feedback and model improvement

[0482] User Action: The user enters feedback on the answer provided.

[0483] Device Action: Sends feedback to the server.

[0484] Server action: The server analyzes the feedback and uses the results to retune the model and improve performance.

[0485] 7. Data anonymization

[0486] Device behavior: The data collected is anonymized to protect your privacy. This process removes any personally identifiable information, allowing you to use the system with confidence.

[0487] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized, so personal information will be protected.

[0488] With the above program and processing flow, the present invention is a system that allows users to consult with others and receive accurate answers regardless of time or place. This system is particularly useful for users who find it difficult to consult with superiors or seniors directly, and provides an environment where consultations can be conducted efficiently and safely.

[0489] The processing flow will be explained below.

[0490] Step 1:

[0491] Users can hold online conversations or web conferences, and the device can collect and store the conversations in real time in text or audio format.

[0492] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[0493] Step 2:

[0494] The device encrypts the collected data and sends it securely to a server, which then stores it in a database with access restrictions to ensure privacy.

[0495] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[0496] Step 3:

[0497] The server analyzes the received data and converts it into text using natural language processing (NLP) techniques, then extracts important keywords and question / answer pairs from the converted text data.

[0498] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[0499] Step 4:

[0500] The server uses the extracted question and answer pairs to fine-tune an existing natural language model (e.g., GPT-3), learning user-specific answer patterns in the process.

[0501] Example: The server feeds the extracted dataset into a natural language model to learn the speaking style of your boss.

[0502] Step 5:

[0503] The user inputs a question through the system interface, and the terminal sends the user's question to the server.

[0504] Example: An employee types, "How can I prepare for my next meeting?"

[0505] Step 6:

[0506] The server analyzes the user's question, uses a fine-tuned natural language model to generate the best answer, and sends the generated answer to the device.

[0507] Example: The server invokes a natural language model to answer "how to prepare for my next meeting."

[0508] Step 7:

[0509] The terminal displays the answer sent from the server to the user. Example: An employee's terminal displays the answer "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[0510] Step 8:

[0511] The user enters feedback on the provided answer, and the device collects the feedback and sends it to the server.

[0512] Example: An employee gives feedback saying, "This answer was helpful."

[0513] Step 9:

[0514] The server analyzes the feedback and retunes the model based on the results to improve performance.

[0515] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[0516] Step 10:

[0517] The device anonymizes the data collected to protect privacy, a process that removes any personally identifiable information.

[0518] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[0519] Example 1

[0520] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0521] Conventional online conversation systems make it difficult for users to seek advice or obtain information in real time, and information sharing is hindered, especially in environments where it is difficult to ask questions directly to superiors or senior colleagues. Furthermore, the quality of answers is inconsistent, preventing users from accurately obtaining useful information. Furthermore, from the perspective of privacy protection, secure data collection and analysis are required. Therefore, there is a need for the development of a system that can effectively collect and analyze online conversation data, provide appropriate answers to users, and ensure privacy protection.

[0522] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0523] In this invention, the server includes means for collecting online communication data from users, means for encrypting the collected data and securely transmitting it to the server, means for analyzing the received data and converting it into text using natural language processing technology, means for extracting important keywords and question-and-answer pairs from the converted text data, means for fine-tuning an existing generative AI model using the extracted question-and-answer pairs, means for accepting questions from users and generating optimal answers using the fine-tuned generative AI model, and means for displaying the generated answers to the users. This makes it possible to effectively collect and analyze online conversation data and provide appropriate answers to users.

[0524] "Online communication data" refers to data related to communication such as conversations and text messages that users have online.

[0525] "Encryption" is the process of transforming data using a specific algorithm to protect it from unauthorized access.

[0526] A "server" is a computer that provides various services within a network, and refers to hardware that runs software with specific functions.

[0527] "Natural language processing technology" refers to technology for processing and analyzing human language using computers, and includes tokenization, part-of-speech tagging, named entity recognition, etc.

[0528] "Convert to text" refers to the process of converting non-text data such as audio or images into data in character string format.

[0529] "Keywords" are words or phrases that contain particularly important information in the data, and are elements that serve as clues for understanding the data.

[0530] A "question and answer pair" is a dataset that combines a user's question with an appropriate answer.

[0531] A "generative AI model" is an artificial intelligence model that generates new information based on given input data.

[0532] "Fine-tuning" refers to the process of optimizing an existing artificial intelligence model to suit a specific task or user needs, and involves additional training.

[0533] "Feedback" refers to evaluations and opinions on the information and answers provided by users, which are useful for improving the system.

[0534] "Anonymization" is the removal of personally identifiable information from data, and is a privacy-protecting technique.

[0535] MODE FOR CARRYING OUT THE INVENTION

[0536] The system for carrying out the present invention is configured by combining various software and hardware for users to provide online conversation data and for analyzing the data.

[0537] Hardware and Software Configuration

[0538] Device configuration

[0539] A terminal is a device used by a user for online conversations and web conferences, such as a PC, smartphone, tablet, etc. The terminal has speech recognition software (e.g., Google Speech-to-Text API) and encryption software (e.g., AES encryption) installed.

[0540] Server Configuration

[0541] The server is a computer used to analyze and store collected data and fine-tune the generative AI model. The following software is installed on the server:

[0542] Natural language processing technology (e.g., SpaCy)

[0543] Database system (e.g. MySQL)

[0544] Generative AI models (e.g., GPT-3)

[0545] A concrete example of the processing flow

[0546] 1. Collecting online conversation data

[0547] The device collects conversations in real time when users are having online conversations or web conferences. The conversations are captured as voice data and converted into text data using voice recognition software. For example, when an employee discusses "sales strategies for a new product" with their boss in a web conference, the voice data is converted into text.

[0548] 2. Data transmission and storage

[0549] The collected text data is encrypted and sent to the server using AES encryption technology and HTTPS protocol, where it is decrypted and stored in a MySQL database.

[0550] 3. Data Analysis

[0551] The server analyzes the text data using natural language processing technology (e.g., SpaCy) and extracts important keywords and question-and-answer pairs. For example, questions and answers related to "sales strategies for new products" are extracted. The results of this analysis are also stored in a database.

[0552] 4. Training the model

[0553] The server uses the extracted question-answer pairs to fine-tune a generative AI model (e.g., GPT-3). This process optimizes the model over multiple epochs and learns the user's unique answer patterns. For example, the model is tuned to learn the speaking style of a boss.

[0554] 5. Provision of consultation system

[0555] Users input questions into the system through a dedicated interface. The input question is encoded and sent to the server. The server analyzes the question and uses a fine-tuned generative AI model to generate the best answer, which is then sent to the device. The device then displays the answer to the user. For example, if an employee asks, "How should I prepare for my next meeting?", the best answer will be displayed.

[0556] 6. Feedback and model improvement

[0557] Users input feedback on the answers they provide through the interface, which is sent to the server for analysis, and the server uses the feedback to retune the generative AI model and improve its performance.

[0558] 7. Data anonymization

[0559] The device anonymizes the collected data to protect privacy. This is a process that removes any personally identifiable information. For example, "consultation about interpersonal relationships at work" is anonymized to protect personal information.

[0560] Examples of prompt statements

[0561] Examples of prompts used in this system include:

[0562] "If you were to extract questions and answers about a sales strategy for a new product during a web conference with your boss and ask about preparations for the next meeting, what would your answer be?"

[0563] "Generate relevant advice from anonymized consultations about workplace relationships."

[0564] The above is an embodiment of the invention.

[0565] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0566] Step 1: Collect online conversation data

[0567] The device collects the content of conversations in real time when users are having online conversations or web conferences. The content of the conversation is captured as audio data and converted into text data using speech recognition software (e.g., Google Speech-to-Text API). Specifically, the collected audio data is input into the speech recognition software and output as text data. For example, when an employee discusses "sales strategies for a new product" with his or her boss in a web conference, the audio data is converted into text data.

[0568] Step 2: Send and store data

[0569] The terminal encrypts the collected text data (e.g., AES encryption). The encrypted data is sent to the server via the network. The server decrypts the received encrypted data and stores it in a database (e.g., MySQL). Specifically, the data is entered into a database system and stored as decrypted text data from its encrypted state. For example, the contents of a meeting may be encrypted and sent to the server, where it is decrypted and stored.

[0570] Step 3: Analyze the data

[0571] The server analyzes the stored data using natural language processing technology (e.g., SpaCy). Important keywords and question-and-answer pairs are extracted from the text data. Specifically, the text data is input into the natural language processing software, which then outputs the analyzed keywords and question-and-answer pairs. For example, questions and answers related to "sales strategies for new products" are extracted.

[0572] Step 4: Train the model

[0573] The server uses the extracted question-answer pairs to fine-tune a generative AI model (e.g., GPT-3). This fine-tuning allows the model to learn the user's unique answer patterns. Specifically, the question-answer pair data is input into the training interface of the generative AI model, and a fine-tuned model is output. For example, the model is adjusted to learn the speaking style of a boss.

[0574] Step 5: Providing a consultation system

[0575] A user inputs a question via a system interface (e.g., a web application). The device encodes the input question and sends it to the server. The server analyzes the received question and generates an optimal answer using a fine-tuned generative AI model. The answer is sent to the device, which then displays the generated answer to the user. Specifically, the system encodes the user input and sends it to the server, then decodes and displays the generated answer. For example, if an employee asks, "How should I prepare for the next meeting?", the optimal answer will be displayed.

[0576] Step 6: Feedback and model improvement

[0577] The user inputs feedback on the provided answer via the interface. The device encodes the feedback data and sends it to the server. The server analyzes the feedback and readjusts the generative AI model based on the results, improving its performance. Specifically, the feedback data is input into the analysis system, which outputs a readjusted model. For example, an evaluation of the appropriateness of the provided answer can be input to improve the quality of the model.

[0578] Step 7: Anonymize the data

[0579] The device anonymizes the collected data to protect privacy. This process removes any personally identifiable information. The server stores and analyzes the anonymized data to ensure the user's privacy. Specifically, the system filters personal information from the data and uses an anonymization tool to replace proper nouns with generic terms. For example, "consultation about interpersonal relationships at work" is anonymized, protecting personal information.

[0580] (Application example 1)

[0581] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0582] In today's information society, online communication is increasing. In this context, it is essential for companies and individuals to manage not only knowledge but also security risks. However, there is a lack of systems that can detect security risks in real time and respond immediately. The objective of this invention is to provide a system that collects and analyzes online conversation data, analyzes security risks in real time, and notifies users.

[0583] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0584] In this invention, the server includes means for collecting online conversation data from users, means for analyzing the collected data to extract question and answer pairs, means for fine-tuning a natural language model using the extracted question and answer pairs, means for generating answers to questions from users using the fine-tuned natural language model, means for collecting online conversation data in real time and analyzing security risks, and means for notifying users of risk alerts based on the analysis results. This makes it possible to efficiently manage online conversation data, detect security risks in real time, and respond immediately.

[0585] "Means for collecting online conversation data from users" refers to a combination of hardware and software for collecting the content of online meetings and conversations conducted by users in real time and saving it in text or audio format.

[0586] "Means for analyzing collected data to extract question and answer pairs" refers to software processes that apply natural language processing techniques to collected online conversation data to identify and extract key question and answer pairs.

[0587] "Means for fine-tuning natural language models using extracted question and answer pairs" refers to techniques for incrementally training existing natural language models using extracted data to improve their accuracy and responsiveness in specific contexts.

[0588] "Means for generating answers to questions from users using a fine-tuned natural language model" refers to algorithms and processes for generating optimal answers to questions entered by users using a fine-tuned natural language model.

[0589] "Means for displaying generated answers to a user" refers to an interface and software for displaying answers generated by a natural language model on a user's device.

[0590] "Means for collecting online conversation data in real time and analyzing security risks" refers to a combination of hardware and software for collecting online conversation data in real time and analyzing the data to detect security risks.

[0591] "Means for notifying users of risk alerts based on the analysis results" refers to alert systems and software that immediately notify users of security risks detected from the analyzed data.

[0592] The system for implementing this invention is configured by combining various software and hardware for allowing users to collect online conversation data in real time using smartphones or other devices and analyze that data.

[0593] System Overview

[0594] Acquisition System:

[0595] The device collects the conversation content in real time when the user is having an online conversation or web conference. The collected data is saved in text or audio format. Specifically, the voice data is converted to text using the Google Web Speech API. The collected text data is securely transmitted from the user's device to a server.

[0596] Data transmission and storage:

[0597] The device encrypts the collected data and sends it securely to a server, which then stores it in a specific database with access restrictions to ensure privacy.

[0598] Data Analysis:

[0599] The server parses the incoming data and converts it into text using natural language processing (NLP) techniques. From the converted text data, question and answer pairs are extracted. This extracted data is used to fine-tune an existing natural language model.

[0600] Security Risk Analysis:

[0601] The collected online conversation data is monitored in real time and security risks are analyzed. Risks are determined based on specific trigger words (e.g., "password," "confidential," etc.). This is done using an NLP model.

[0602] Risk Alert Notification:

[0603] Based on the analysis results, users are immediately notified of any security risks that are discovered. Once the data has been analyzed, the server sends an alert to the user along with a risk assessment. This notification is sent via an interface such as a push notification on a smartphone.

[0604] User Interface:

[0605] Users input questions through the interface, and the best answer to that question is generated by a fine-tuned natural language model. The generated answer is displayed on the user's device. In addition, users can input feedback on the provided answer, which is sent to the server and used to refine the model.

[0606] Examples and prompts

[0607] For example, if a user is using a project management system and during a conversation they say, "The budget for the next project has been decided to be 5 million yen," the system will collect and analyze that conversation in real time to determine whether it contains confidential information.

[0608] An example prompt is:

[0609] Online conversation data: "The budget for the next project has been decided to be 5 million yen."

[0610] Prompt: "Is this conversation a security risk?"

[0611] This system allows users to instantly identify potential security risks that arise during conversations and take appropriate measures.

[0612] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0613] Step 1:

[0614] The device collects conversation content in real time when a user is engaged in online conversations or web conferences. It receives voice or text data as input and converts it into text using speech recognition software (e.g., Google Web Speech API). The output is text-format conversation data.

[0615] Step 2:

[0616] The device encrypts the collected text data using a secure protocol (e.g., HTTPS) and securely sends it to the server. It receives text conversation data as input, encrypts it, and then sends it to the server. The output is encrypted text data.

[0617] Step 3:

[0618] The server stores the received data in a specific database. It receives encrypted text data as input, decrypts it, and stores it in the database. The output is the text data stored in the database.

[0619] Step 4:

[0620] The server analyzes the stored data and extracts question and answer pairs using natural language processing (NLP) techniques. It receives text data as input and applies NLP algorithms to extract important keywords and question and answer pairs. The output is the extracted question and answer pairs.

[0621] Step 5:

[0622] The server uses the extracted question and answer pairs to fine-tune the natural language model. It takes the question and answer pairs as input and retrains the generative AI model. The output is a fine-tuned natural language model.

[0623] Step 6:

[0624] The terminal receives the user's question through the system interface, and sends it to the server in text format. The output is the question in text format.

[0625] Step 7:

[0626] The server analyzes the received question and generates the best answer using a fine-tuned natural language model. It takes the user's question as input and generates an answer using a generative AI model. The output is the generated answer in text format.

[0627] Step 8:

[0628] The terminal displays the generated answer to the user. It takes as input the generated textual answer and displays it in a user interface. The output is the answer that is displayed to the user.

[0629] Step 9:

[0630] The server monitors the collected online conversation data in real time and analyzes security risks. It receives the conversation data as input and evaluates the security risk using an NLP model. The output is the security risk assessment result.

[0631] Step 10:

[0632] The server notifies the user of a risk alert based on the analysis results. It receives the security risk assessment results as input, generates a message for the alert notification, and sends it to the terminal. The output is an alert message that is notified to the user.

[0633] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0634] The system for implementing this invention has the ability to take emotional state into account when generating answers by allowing users to provide online conversation data, analyzing that data, and combining it with an emotion engine that recognizes the user's emotions.

[0635] Explanation of program processing

[0636] 1. Collecting online conversation data

[0637] Device operation: When a user engages in online conversations or web conferences, the device collects the conversation content in real time and stores it in text or audio format.

[0638] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[0639] 2. Data transmission and storage

[0640] What the device does: Collected data is encrypted and sent securely to a server, which stores the data in a database with access restrictions to ensure privacy.

[0641] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[0642] 3. Data Analysis

[0643] Server operation: The server analyzes the received data and converts it into text using natural language processing (NLP) techniques. It then extracts important keywords and question / answer pairs from the converted text data.

[0644] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[0645] 4. Emotion analysis using an emotion engine

[0646] Server operation: The server analyzes the user's emotional state based on the collected data using an emotion engine. Based on the analysis results, it assigns emotion labels to the conversation data.

[0647] Example: The server assigns emotion labels such as "anxiety" or "excitement" based on the user's speech.

[0648] 5. Training the model

[0649] What the server does: It uses the extracted question-answer pairs and sentiment labels to fine-tune an existing natural language model, learning user-specific answer patterns along with their emotional states in the process.

[0650] Example: The server learns the speaking style and corresponding emotional state of a boss and trains a natural language model (e.g., GPT-3).

[0651] 6. Provision of consultation system

[0652] User action: The user inputs a question through the system interface. The terminal sends the user's question to the server.

[0653] Example: An employee types, "How can I prepare for my next meeting?"

[0654] 7. Answer Generation

[0655] Server operation: The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. The answer is generated based on the user's emotional state. The generated answer is then sent to the device.

[0656] Example: The server provides an "anxious" employee with a response such as, "To prepare for the next meeting, first check the agenda and gather any relevant materials. If you have any concerns, please check with your leader."

[0657] 8. View Answers

[0658] What the device does: The device displays the answer sent by the server to the user.

[0659] Example: The employee's device displays the following response: "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[0660] 9. Feedback and Model Improvement

[0661] User action: The user enters feedback on the provided answer. The device collects the feedback and sends it to the server.

[0662] Server action: The server analyzes the feedback and retunes the model based on the results to improve performance.

[0663] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[0664] 10. Data Anonymization

[0665] What the device does: The data collected is anonymized to protect your privacy. This process removes any personally identifiable information.

[0666] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[0667] With the above program and processing flow, the present invention is a system that allows users to consult with others and receive appropriate answers regardless of time or place. This system is particularly useful for users who find it difficult to consult directly with superiors or seniors, and provides an environment where consultations can be conducted efficiently and safely while taking into account their emotional state.

[0668] The processing flow will be explained below.

[0669] Step 1:

[0670] Users can hold online conversations or web conferences, and the device can collect and save the conversations in real time as text or audio.

[0671] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[0672] Step 2:

[0673] The device encrypts the collected data and sends it securely to a server, where it is stored in a specific database with access restrictions to ensure privacy.

[0674] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[0675] Step 3:

[0676] The server analyzes the received data and converts it into text using natural language processing (NLP) techniques, then extracts important keywords and question / answer pairs from the converted text data.

[0677] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[0678] Step 4:

[0679] The server analyzes the user's emotional state based on the collected data using an emotion engine, and assigns emotion labels to the conversation data based on the analysis results.

[0680] Example: The server assigns emotion labels such as "anxiety" or "excitement" based on the user's speech.

[0681] Step 5:

[0682] The server uses the extracted question and answer pairs and sentiment labels to fine-tune an existing natural language model, learning the user's unique answer patterns along with their emotional state in the process.

[0683] Example: The server learns the speaking style and corresponding emotional state of a boss and trains a natural language model.

[0684] Step 6:

[0685] The user inputs a question through the system interface, and the terminal sends the user's question to the server.

[0686] Example: An employee types, "How can I prepare for my next meeting?"

[0687] Step 7:

[0688] The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. The answer is generated based on the user's emotional state. The generated answer is then sent to the device.

[0689] Example: The server provides an "anxious" employee with a response such as, "To prepare for the next meeting, first check the agenda and gather any relevant materials. If you have any concerns, please check with your leader."

[0690] Step 8:

[0691] The terminal displays the answer sent from the server to the user.

[0692] Example: The employee's device displays the following response: "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[0693] Step 9:

[0694] The user enters feedback on the provided answer, and the device collects the feedback and sends it to the server.

[0695] Example: An employee gives feedback saying, "This answer was helpful."

[0696] Step 10:

[0697] The server analyzes the feedback and retunes the model based on the results to improve performance.

[0698] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[0699] Step 11:

[0700] The device anonymizes the data collected to protect privacy, a process that removes any personally identifiable information.

[0701] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[0702] Example 2

[0703] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0704] Current consultation systems that use online conversation data generate answers without considering the user's emotional state, which can result in the system failing to provide the most appropriate answer for the user. Furthermore, data privacy protection is insufficient, posing a risk of personal information leaks. The objective of this invention is to solve these problems and provide a system that generates appropriate answers that consider the user's emotional state and protects data privacy.

[0705] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0706] In this invention, the server includes terminal means for collecting online conversation data from users, terminal means for encrypting the collected data and transmitting it to the server in a secure manner, database means, means for the server to analyze the received data using natural language processing technology and extract question and answer pairs, means for the server to analyze the user's emotional state from the data using a sentiment analysis engine and assign sentiment labels, server means for fine-tuning a natural language model using the extracted question and answer pairs and sentiment labels, terminal means for accepting questions from users, means for the server to generate answers to the user's questions using the fine-tuned natural language model and sentiment analysis engine, and terminal means for displaying the generated answers to the users. This makes it possible to provide appropriate answers that take the user's emotional state into consideration and protect the privacy of the data.

[0707] "Terminal means" refers to electronic devices that users use to collect online conversation data and input questions.

[0708] "Server" refers to the computer system that receives, analyzes, and stores collected data.

[0709] "Encryption" is a technology that converts data using a specific algorithm to protect it from unauthorized access or tampering.

[0710] "Server means" refers to the hardware or software used by the server to perform a particular function.

[0711] "Database means" refers to a system for safely storing received data and making it accessible when necessary.

[0712] "Natural language processing technology" is a technology for analyzing and understanding human language, and primarily uses machine learning and language models.

[0713] A "question and answer pair" refers to a set of questions entered by a user and their corresponding appropriate answers.

[0714] An "emotion analysis engine" refers to technology that analyzes a user's emotional state from collected data and assigns various emotional labels.

[0715] An "emotion label" is a tag attached to data to indicate the user's emotional state, such as "joy," "sadness," or "anxiety."

[0716] "Fine-tuning" is the process of fine-tuning an existing model to suit a specific dataset or purpose.

[0717] "Natural language model" refers to a machine learning model trained to generate and parse human language.

[0718] "Feedback" refers to the evaluation or opinion a user gives on a provided answer.

[0719] "Anonymization" is a technique that removes or transforms personally identifiable information to protect the privacy of data.

[0720] "Display means" refers to a device or software for visually presenting the generated answer to the user.

[0721] An embodiment of this invention is described below. This system is mainly composed of three elements: a server, a terminal, and a user. This system allows users to provide online conversation data, analyzes that data, and combines it with an emotion engine that recognizes the user's emotions, thereby providing a function that takes the user's emotional state into account when generating answers.

[0722] Device behavior

[0723] When a user has an online conversation or web conference, the device collects the conversation in real time using the device's microphone and camera. The collected data is saved in audio or video format and immediately converted into text format. The data is then encrypted and sent to the server in a secure manner. The encryption techniques used include AES-256 encryption and SSL / TLS protocol.

[0724] Server Operation

[0725] The server stores the received data in a database. Access restrictions are set to this database to ensure privacy. The data received by the server is analyzed using natural language processing (NLP) techniques to extract important keywords and question-and-answer pairs. Software such as Apache Kafka and TensorFlow is used for this analysis. The extracted text data is further analyzed using a sentiment analysis engine (e.g., IBM Watson or Amazon Comprehend), and emotional labels are assigned to the data. Emotional labels such as "happiness," "sadness," and "anxiety" are taken into account.

[0726] Training the model

[0727] The server uses the extracted question and answer pairs and emotion labels to fine-tune a natural language model using Hugging Face's transformers library and OpenAI's GPT-3. The fine-tuned model learns the user's unique answer patterns and emotional state, allowing it to generate more appropriate answers.

[0728] Providing answers to users

[0729] Users input questions through the system interface (web app or mobile app). The device sends the user's question to the server, which analyzes it and generates the best answer using a fine-tuned natural language model and sentiment analysis engine. The generated answer is sent to the device and displayed to the user. For example, if an employee inputs "What should I do to prepare for the next meeting?", the server will provide a response such as "To prepare for the next meeting, first check the agenda and gather relevant materials."

[0730] Feedback and model improvement

[0731] Users can enter feedback on the answers provided, and the device collects and sends the feedback to the server, which analyzes the feedback and retunes the model based on the results to improve performance.

[0732] Data anonymization

[0733] The data collected is anonymized to protect your privacy - in this process, any personally identifiable information is removed and it is stored securely.

[0734] Examples of concrete examples and prompts

[0735] Example: After a meeting with their boss, an employee asks the system, "How should I approach the challenges in my new department?" The system recognizes that the employee is feeling anxious and provides advice that takes their feelings into consideration.

[0736] Example prompt: "I've just been assigned to a new department and I'm not sure how to solve a problem. Could you please give me some advice?"

[0737] As a result, this system provides appropriate answers that take into account the user's emotional state, protecting the privacy of data while providing an environment in which users can seek advice efficiently.

[0738] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0739] Step 1:

[0740] Collecting online conversation data

[0741] Device operation: When a user has an online conversation or web conference, the device's microphone and camera capture what is being said in real time and save it as audio or video data.

[0742] Input: User audio and video data.

[0743] Data processing: The device converts audio and video data into text format.

[0744] Output: Conversation data in text format.

[0745] What it does: Your device's microphone records your "discussion about a new project" and uses speech recognition software to convert it into text.

[0746] Step 2:

[0747] Data transmission and storage

[0748] What the device does: Encrypts the collected text data and sends it to the server in a secure manner.

[0749] Input: Conversation data in text format.

[0750] Data processing: Data is encrypted with AES-256 and transmitted using the SSL / TLS protocol.

[0751] Output: The encrypted data is sent to the server.

[0752] Specific operation: Text data collected on the device is encrypted with AES-256 and sent via SSL to the server, which then stores the received data in a database.

[0753] Step 3:

[0754] Data analysis

[0755] Server operation: The server analyzes the received text data using natural language processing technology and extracts important keywords and question-and-answer pairs.

[0756] Input: Conversation data in text format.

[0757] Data processing: Analyze using natural language processing techniques (e.g., Apache Kafka, TensorFlow) to extract keywords and question-answer pairs.

[0758] Output: Important keywords and question-answer pairs are extracted.

[0759] What it does: The server extracts keywords such as "plans to enter new markets" from the text data and identifies related questions and answers.

[0760] Step 4:

[0761] Emotion analysis using an emotion engine

[0762] Server operation: The server uses an emotion analysis engine to analyze the user's emotional state based on the data and assigns an emotion label.

[0763] Input: Keywords and question-answer pairs.

[0764] Data processing: Analyze the emotional state using a sentiment analysis engine (e.g., IBM Watson, Amazon Comprehend) and assign sentiment labels to the data.

[0765] Output: Data with emotion labels.

[0766] Specific operation: The server analyzes the emotion "anxiety" from statements such as "The project is progressing slowly" and assigns a label to it.

[0767] Step 5:

[0768] Training the model

[0769] Server operation: Fine-tune a natural language model using extracted question-answer pairs and sentiment labels.

[0770] Input: Question-answer pairs and sentiment labels.

[0771] Data processing: Fine-tune the model using Hugging Face's transformers library and OpenAI's GPT-3.

[0772] Output: A fine-tuned natural language model.

[0773] Specific operation: The server learns answers that adapt to the employee's question patterns based on guidelines and emotion labels.

[0774] Step 6:

[0775] Providing a consultation system

[0776] User action: The user enters a question through the system interface (web app or mobile app). The device sends the user's question to the server.

[0777] Input: The user's question.

[0778] Data processing: The user's question is formulated and sent to the server.

[0779] Output: The user's question sent to the server.

[0780] What happens: An employee types, "How do I prepare for a meeting?" and the device sends the question to the server.

[0781] Step 7:

[0782] Generate answers

[0783] Server operation: The server analyzes the user's question and generates the best answer using a fine-tuned natural language model and sentiment analysis engine based on the sentiment labels.

[0784] Input: User question and sentiment label.

[0785] Data processing: Generate answers using fine-tuned natural language models.

[0786] Output: The generated answer.

[0787] Specific operation: The server receives a question about "preparing for the meeting," generates a response such as "First, check the agenda and gather the materials," and sends it to the device.

[0788] Step 8:

[0789] Show Answers

[0790] What the device does: The device displays the answer sent by the server to the user.

[0791] Input: The answer sent by the server.

[0792] Data processing: Display the answer to the user visually or audibly.

[0793] Output: The answer displayed to the user.

[0794] What it does: The device prompts the employee, "To prepare for your next meeting, first review the agenda and gather any relevant materials."

[0795] Step 9:

[0796] Feedback and model improvement

[0797] User action: The user enters feedback on the provided answer. The device sends the feedback to the server.

[0798] Input: User feedback.

[0799] Data processing: Analyze feedback and use it to improve the model.

[0800] Output: An improved natural language model.

[0801] Specific behavior: The user enters feedback such as "This answer was helpful," and the device sends the feedback to the server. The server analyzes the feedback and recalibrates the model.

[0802] Step 10:

[0803] Data anonymization

[0804] What the device does: Anonymizes collected data to protect privacy.

[0805] Input: Conversation data, including personal information.

[0806] Data processing: removing or transforming personally identifiable information.

[0807] Output: Anonymized data.

[0808] What it does: It removes and securely stores your name and unique identifiers.

[0809] (Application example 2)

[0810] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0811] When it comes to communication between workers and robots in factories, there is a lack of a way to understand the emotions of workers and respond appropriately. This can have a negative impact on productivity and safety, especially in emergencies or stressful situations. To solve this problem, a system is needed that can analyze the emotional state of workers and respond appropriately according to that state.

[0812] The identification process by the identification 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 online conversation data from users, means for analyzing the collected data to extract question and answer pairs, means for fine-tuning a natural language model using the extracted question and answer pairs, means for analyzing the emotional state of the user using an emotion engine based on the collected data, means for generating an adapted answer based on the emotional state, and means for displaying the generated answer to the user. This makes it possible to understand the emotional state of employees and respond appropriately.

[0813] "User" refers to the factory employees who use the system.

[0814] "Online conversation data" refers to one-to-one voice or text interactions between employees and robots.

[0815] "Means of collection" refers to hardware and software that records or records online conversations between employees and robots.

[0816] "Means for analyzing and extracting question and answer pairs" refers to technology that uses natural language processing technology to analyze collected online conversation data and identify important questions and their answers.

[0817] "Fine-tuning" refers to the process of optimizing the performance of an existing natural language model using extracted question and answer pairs.

[0818] "Emotion engine" refers to software used to analyze users' emotions from online conversation data and identify their emotional states.

[0819] "Means for analyzing emotional state" refers to the process for determining a user's emotions from collected data using an emotion engine.

[0820] "Means for generating adapted answers" refers to techniques for generating appropriate and optimal answers based on the user's emotional state.

[0821] "Means for displaying" refers to a display or audio output device for visually or audibly communicating the generated response to the employee.

[0822] A system for implementing this invention allows a user to provide online conversation data, analyzes the data, and combines an emotion engine that recognizes the user's emotions with a natural language model to take the user's emotional state into account when generating responses. This system is particularly applicable to factory robots and is designed to facilitate smooth communication between factory workers and robots.

[0823] Hardware and Software Configuration

[0824] Hardware: Robots, microphones, and servers located in the factory

[0825] Software: Natural language processing engines (e.g., Transformer's NLP), sentiment analysis engines (e.g., Sentiment Analysis Pipeline), generative AI models (e.g., GPT-3)

[0826] Specific explanation of the system's operation

[0827] Collection and transmission of online conversation data

[0828] The robot collects online conversations with employees in real time via a microphone, and the collected voice data is encrypted and securely sent to a server.

[0829] Data analysis and text conversion

[0830] The server analyzes the received voice data and converts it into text using natural language processing (NLP) techniques, extracting important keywords and question / answer pairs from the converted text data.

[0831] Emotion analysis using an emotion engine

[0832] Next, the server analyzes the user's emotional state using an emotion engine and assigns emotion labels to the conversation data, thereby understanding the user's emotional state.

[0833] Training the model

[0834] The extracted question-answer pairs and sentiment labels are used to fine-tune an existing natural language model (e.g., GPT-3), learning user-specific answer patterns along with their emotional states in the process.

[0835] Responding to questions

[0836] When an employee types a question through the system's interface, the robot sends it to the server, which uses a fine-tuned natural language model and emotion engine to generate the best answer and communicate it to the user via a display or voice output device.

[0837] Specific examples

[0838] For example, if an employee in the quality control department asks the robot, "I don't know how to inspect the product," the robot's emotion analysis engine will recognize that the employee is feeling impatient. The generative AI model will respond by offering guidance that is adapted to the employee's impatient state: "Please stay calm. Let's inspect the product by following these steps."

[0839] Prompt Sentence Examples

[0840] "If an employee notices something is wrong but is feeling anxious, how do I support them?"

[0841] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0842] Step 1:

[0843] The terminal collects online conversation data. Specifically, the factory robot records the conversation with the employee in real time through a microphone. The collected data is in the form of audio, and the audio data is temporarily stored on the terminal for further processing.

[0844] Step 2:

[0845] The device encrypts the collected voice data and securely transmits it to a server using a secure communication protocol (e.g., HTTPS). The server then stores the received voice data in a database with access restrictions to ensure privacy.

[0846] Step 3:

[0847] The server analyzes the received voice data and converts it into text using natural language processing (NLP) techniques, such as using Transformers NLP models to convert speech to text, and stores the resulting text for use in the next step.

[0848] Step 4:

[0849] The server extracts important keywords and question / answer pairs from the converted text data. It uses natural language processing technology to identify which parts are questions and which parts are answers and extracts those pairs. This allows it to understand what information the user is looking for.

[0850] Step 5:

[0851] The server uses an emotion engine to analyze the user's emotional state based on the collected text data. An emotion analysis engine (e.g., Sentiment Analysis Pipeline) is used to identify emotions such as "anxiety" or "impatience" from the user's comments. Based on the results of this analysis, an emotion label is assigned to the text data.

[0852] Step 6:

[0853] The server uses the extracted question and answer pairs and sentiment labels to fine-tune existing natural language models, such as using GPT-3 to learn employee speech patterns and their corresponding emotional states. This process produces a model that understands the user's unique communication style.

[0854] Step 7:

[0855] The user inputs a question through the system interface. The factory robot recognizes this question and sends it to the server. The question is then analyzed in the next step.

[0856] Step 8:

[0857] The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. For example, if the user is in an "anxious" emotional state, the server will provide a calm tone and specific advice that is adapted to that state. The generated answer is then sent to the device.

[0858] Step 9:

[0859] The terminal displays the response sent from the server to the user, and the factory robot communicates the response to the employee via voice, allowing the user to take appropriate action.

[0860] Step 10:

[0861] The user provides feedback on the answers provided by the device, which collects this feedback and sends it to the server, which analyzes it and uses the results to retune the model and improve the system's performance.

[0862] Step 11:

[0863] The device anonymizes the data collected to protect your privacy, ensuring that personally identifiable information is removed and stored securely.

[0864] Example prompt sentence:

[0865] "If an employee notices something is wrong but is feeling anxious, how do I support them?"

[0866] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0867] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0868] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0869] [Third embodiment]

[0870] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0871] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0872] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0873] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0874] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0875] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0876] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0877] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0878] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0879] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0880] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0881] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0882] A system for implementing the present invention is configured by combining various software and hardware components for users to provide online conversation data and for analyzing that data.

[0883] Explanation of program processing

[0884] 1. Collecting online conversation data

[0885] What the device does: When a user has an online conversation or web conference, the device collects the conversation in real time and stores the collected data in text or audio format.

[0886] Example: An employee has a web conference with their manager about a particular project, and the content of the conference is collected.

[0887] 2. Data transmission and storage

[0888] What the device does: Collected data is encrypted and securely sent to the server.

[0889] Server behavior: The server stores the received data in a specific database with restricted access to ensure privacy.

[0890] 3. Data Analysis

[0891] Server operation: The server analyzes the received data and converts it into text using natural language processing (NLP) techniques. It then extracts important keywords and question / answer pairs from the converted text data.

[0892] Example: The server extracts questions and answers about "plans to enter new markets" from a conversation between a boss and an employee.

[0893] 4. Training the model

[0894] What the server does: It uses the extracted question and answer pairs to fine-tune an existing natural language model. Specifically, the fine-tuned model learns the user's unique answer patterns.

[0895] Example: The server uses the extracted data to train a natural language model to learn the speaking style of your boss.

[0896] 5. Provision of consultation system

[0897] User action: The user enters a question through the system's interface.

[0898] What the terminal does: Sends the entered question to the server.

[0899] What the server does: Analyzes the received question, uses a fine-tuned natural language model to generate the best answer, and sends this answer to the device.

[0900] Terminal behavior: Display the generated answer to the user.

[0901] For example: An employee asks, "How can I prepare for my next meeting?" and is provided with the best answer to that question.

[0902] 6. Feedback and model improvement

[0903] User Action: The user enters feedback on the answer provided.

[0904] Device Action: Sends feedback to the server.

[0905] Server action: The server analyzes the feedback and uses the results to retune the model and improve performance.

[0906] 7. Data anonymization

[0907] Device behavior: The data collected is anonymized to protect your privacy. This process removes any personally identifiable information, allowing you to use the system with confidence.

[0908] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized, so personal information will be protected.

[0909] With the above program and processing flow, the present invention is a system that allows users to consult with others and receive accurate answers regardless of time or place. This system is particularly useful for users who find it difficult to consult with superiors or seniors directly, and provides an environment where consultations can be conducted efficiently and safely.

[0910] The processing flow will be explained below.

[0911] Step 1:

[0912] Users can hold online conversations or web conferences, and the device can collect and store the conversations in real time in text or audio format.

[0913] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[0914] Step 2:

[0915] The device encrypts the collected data and sends it securely to a server, which then stores it in a database with access restrictions to ensure privacy.

[0916] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[0917] Step 3:

[0918] The server analyzes the received data and converts it into text using natural language processing (NLP) techniques, then extracts important keywords and question / answer pairs from the converted text data.

[0919] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[0920] Step 4:

[0921] The server uses the extracted question and answer pairs to fine-tune an existing natural language model (e.g., GPT-3), learning user-specific answer patterns in the process.

[0922] Example: The server feeds the extracted dataset into a natural language model to learn the speaking style of your boss.

[0923] Step 5:

[0924] The user inputs a question through the system interface, and the terminal sends the user's question to the server.

[0925] Example: An employee types, "How can I prepare for my next meeting?"

[0926] Step 6:

[0927] The server analyzes the user's question, uses a fine-tuned natural language model to generate the best answer, and sends the generated answer to the device.

[0928] Example: The server invokes a natural language model to answer "how to prepare for my next meeting."

[0929] Step 7:

[0930] The terminal displays the answer sent from the server to the user. Example: An employee's terminal displays the answer "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[0931] Step 8:

[0932] The user enters feedback on the provided answer, and the device collects the feedback and sends it to the server.

[0933] Example: An employee gives feedback saying, "This answer was helpful."

[0934] Step 9:

[0935] The server analyzes the feedback and retunes the model based on the results to improve performance.

[0936] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[0937] Step 10:

[0938] The device anonymizes the data collected to protect privacy, a process that removes any personally identifiable information.

[0939] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[0940] Example 1

[0941] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0942] Conventional online conversation systems make it difficult for users to seek advice or obtain information in real time, and information sharing is hindered, especially in environments where it is difficult to ask questions directly to superiors or senior colleagues. Furthermore, the quality of answers is inconsistent, preventing users from accurately obtaining useful information. Furthermore, from the perspective of privacy protection, secure data collection and analysis are required. Therefore, there is a need for the development of a system that can effectively collect and analyze online conversation data, provide appropriate answers to users, and ensure privacy protection.

[0943] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0944] In this invention, the server includes means for collecting online communication data from users, means for encrypting the collected data and securely transmitting it to the server, means for analyzing the received data and converting it into text using natural language processing technology, means for extracting important keywords and question-and-answer pairs from the converted text data, means for fine-tuning an existing generative AI model using the extracted question-and-answer pairs, means for accepting questions from users and generating optimal answers using the fine-tuned generative AI model, and means for displaying the generated answers to the users. This makes it possible to effectively collect and analyze online conversation data and provide appropriate answers to users.

[0945] "Online communication data" refers to data related to communication such as conversations and text messages that users have online.

[0946] "Encryption" is the process of transforming data using a specific algorithm to protect it from unauthorized access.

[0947] A "server" is a computer that provides various services within a network, and refers to hardware that runs software with specific functions.

[0948] "Natural language processing technology" refers to technology for processing and analyzing human language using computers, and includes tokenization, part-of-speech tagging, named entity recognition, etc.

[0949] "Convert to text" refers to the process of converting non-text data such as audio or images into data in character string format.

[0950] "Keywords" are words or phrases that contain particularly important information in the data, and are elements that serve as clues for understanding the data.

[0951] A "question and answer pair" is a dataset that combines a user's question with an appropriate answer.

[0952] A "generative AI model" is an artificial intelligence model that generates new information based on given input data.

[0953] "Fine-tuning" refers to the process of optimizing an existing artificial intelligence model to suit a specific task or user needs, and involves additional training.

[0954] "Feedback" refers to evaluations and opinions on the information and answers provided by users, which are useful for improving the system.

[0955] "Anonymization" is the removal of personally identifiable information from data, and is a privacy-protecting technique.

[0956] MODE FOR CARRYING OUT THE INVENTION

[0957] The system for carrying out the present invention is configured by combining various software and hardware for users to provide online conversation data and for analyzing the data.

[0958] Hardware and Software Configuration

[0959] Device configuration

[0960] A terminal is a device used by a user for online conversations and web conferences, such as a PC, smartphone, tablet, etc. The terminal has speech recognition software (e.g., Google Speech-to-Text API) and encryption software (e.g., AES encryption) installed.

[0961] Server Configuration

[0962] The server is a computer used to analyze and store collected data and fine-tune the generative AI model. The following software is installed on the server:

[0963] Natural language processing technology (e.g., SpaCy)

[0964] Database system (e.g. MySQL)

[0965] Generative AI models (e.g., GPT-3)

[0966] A concrete example of the processing flow

[0967] 1. Collecting online conversation data

[0968] The device collects conversations in real time when users are having online conversations or web conferences. The conversations are captured as voice data and converted into text data using voice recognition software. For example, when an employee discusses "sales strategies for a new product" with their boss in a web conference, the voice data is converted into text.

[0969] 2. Data transmission and storage

[0970] The collected text data is encrypted and sent to the server using AES encryption technology and HTTPS protocol, where it is decrypted and stored in a MySQL database.

[0971] 3. Data Analysis

[0972] The server analyzes the text data using natural language processing technology (e.g., SpaCy) and extracts important keywords and question-and-answer pairs. For example, questions and answers related to "sales strategies for new products" are extracted. The results of this analysis are also stored in a database.

[0973] 4. Training the model

[0974] The server uses the extracted question-answer pairs to fine-tune a generative AI model (e.g., GPT-3). This process optimizes the model over multiple epochs and learns the user's unique answer patterns. For example, the model is tuned to learn the speaking style of a boss.

[0975] 5. Provision of consultation system

[0976] Users input questions into the system through a dedicated interface. The input question is encoded and sent to the server. The server analyzes the question and uses a fine-tuned generative AI model to generate the best answer, which is then sent to the device. The device then displays the answer to the user. For example, if an employee asks, "How should I prepare for my next meeting?", the best answer will be displayed.

[0977] 6. Feedback and model improvement

[0978] Users input feedback on the answers they provide through the interface, which is sent to the server for analysis, and the server uses the feedback to retune the generative AI model and improve its performance.

[0979] 7. Data anonymization

[0980] The device anonymizes the collected data to protect privacy. This is a process that removes any personally identifiable information. For example, "consultation about interpersonal relationships at work" is anonymized to protect personal information.

[0981] Examples of prompt statements

[0982] Examples of prompts used in this system include:

[0983] "If you were to extract questions and answers about a sales strategy for a new product during a web conference with your boss and ask about preparations for the next meeting, what would your answer be?"

[0984] "Generate relevant advice from anonymized consultations about workplace relationships."

[0985] The above is an embodiment of the invention.

[0986] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0987] Step 1: Collect online conversation data

[0988] The device collects the content of conversations in real time when users are having online conversations or web conferences. The content of the conversation is captured as audio data and converted into text data using speech recognition software (e.g., Google Speech-to-Text API). Specifically, the collected audio data is input into the speech recognition software and output as text data. For example, when an employee discusses "sales strategies for a new product" with his or her boss in a web conference, the audio data is converted into text data.

[0989] Step 2: Send and store data

[0990] The terminal encrypts the collected text data (e.g., AES encryption). The encrypted data is sent to the server via the network. The server decrypts the received encrypted data and stores it in a database (e.g., MySQL). Specifically, the data is entered into a database system and stored as decrypted text data from its encrypted state. For example, the contents of a meeting may be encrypted and sent to the server, where it is decrypted and stored.

[0991] Step 3: Analyze the data

[0992] The server analyzes the stored data using natural language processing technology (e.g., SpaCy). Important keywords and question-and-answer pairs are extracted from the text data. Specifically, the text data is input into the natural language processing software, which then outputs the analyzed keywords and question-and-answer pairs. For example, questions and answers related to "sales strategies for new products" are extracted.

[0993] Step 4: Train the model

[0994] The server uses the extracted question-answer pairs to fine-tune a generative AI model (e.g., GPT-3). This fine-tuning allows the model to learn the user's unique answer patterns. Specifically, the question-answer pair data is input into the training interface of the generative AI model, and a fine-tuned model is output. For example, the model is adjusted to learn the speaking style of a boss.

[0995] Step 5: Providing a consultation system

[0996] A user inputs a question via a system interface (e.g., a web application). The device encodes the input question and sends it to the server. The server analyzes the received question and generates an optimal answer using a fine-tuned generative AI model. The answer is sent to the device, which then displays the generated answer to the user. Specifically, the system encodes the user input and sends it to the server, then decodes and displays the generated answer. For example, if an employee asks, "How should I prepare for the next meeting?", the optimal answer will be displayed.

[0997] Step 6: Feedback and model improvement

[0998] The user inputs feedback on the provided answer via the interface. The device encodes the feedback data and sends it to the server. The server analyzes the feedback and readjusts the generative AI model based on the results, improving its performance. Specifically, the feedback data is input into the analysis system, which outputs a readjusted model. For example, an evaluation of the appropriateness of the provided answer can be input to improve the quality of the model.

[0999] Step 7: Anonymize the data

[1000] The device anonymizes the collected data to protect privacy. This process removes any personally identifiable information. The server stores and analyzes the anonymized data to ensure the user's privacy. Specifically, the system filters personal information from the data and uses an anonymization tool to replace proper nouns with generic terms. For example, "consultation about interpersonal relationships at work" is anonymized, protecting personal information.

[1001] (Application example 1)

[1002] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1003] In today's information society, online communication is increasing. In this context, it is essential for companies and individuals to manage not only knowledge but also security risks. However, there is a lack of systems that can detect security risks in real time and respond immediately. The objective of this invention is to provide a system that collects and analyzes online conversation data, analyzes security risks in real time, and notifies users.

[1004] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1005] In this invention, the server includes means for collecting online conversation data from users, means for analyzing the collected data to extract question and answer pairs, means for fine-tuning a natural language model using the extracted question and answer pairs, means for generating answers to questions from users using the fine-tuned natural language model, means for collecting online conversation data in real time and analyzing security risks, and means for notifying users of risk alerts based on the analysis results. This makes it possible to efficiently manage online conversation data, detect security risks in real time, and respond immediately.

[1006] "Means for collecting online conversation data from users" refers to a combination of hardware and software for collecting the content of online meetings and conversations conducted by users in real time and saving it in text or audio format.

[1007] "Means for analyzing collected data to extract question and answer pairs" refers to software processes that apply natural language processing techniques to collected online conversation data to identify and extract key question and answer pairs.

[1008] "Means for fine-tuning natural language models using extracted question and answer pairs" refers to techniques for incrementally training existing natural language models using extracted data to improve their accuracy and responsiveness in specific contexts.

[1009] "Means for generating answers to questions from users using a fine-tuned natural language model" refers to algorithms and processes for generating optimal answers to questions entered by users using a fine-tuned natural language model.

[1010] "Means for displaying generated answers to a user" refers to an interface and software for displaying answers generated by a natural language model on a user's device.

[1011] "Means for collecting online conversation data in real time and analyzing security risks" refers to a combination of hardware and software for collecting online conversation data in real time and analyzing the data to detect security risks.

[1012] "Means for notifying users of risk alerts based on the analysis results" refers to alert systems and software that immediately notify users of security risks detected from the analyzed data.

[1013] The system for implementing this invention is configured by combining various software and hardware for allowing users to collect online conversation data in real time using smartphones or other devices and analyze that data.

[1014] System Overview

[1015] Acquisition System:

[1016] The device collects the conversation content in real time when the user is having an online conversation or web conference. The collected data is saved in text or audio format. Specifically, the voice data is converted to text using the Google Web Speech API. The collected text data is securely transmitted from the user's device to a server.

[1017] Data transmission and storage:

[1018] The device encrypts the collected data and sends it securely to a server, which then stores it in a specific database with access restrictions to ensure privacy.

[1019] Data Analysis:

[1020] The server parses the incoming data and converts it into text using natural language processing (NLP) techniques. From the converted text data, question and answer pairs are extracted. This extracted data is used to fine-tune an existing natural language model.

[1021] Security Risk Analysis:

[1022] The collected online conversation data is monitored in real time and security risks are analyzed. Risks are determined based on specific trigger words (e.g., "password," "confidential," etc.). This is done using an NLP model.

[1023] Risk Alert Notification:

[1024] Based on the analysis results, users are immediately notified of any security risks that are discovered. Once the data has been analyzed, the server sends an alert to the user along with a risk assessment. This notification is sent via an interface such as a push notification on a smartphone.

[1025] User Interface:

[1026] Users input questions through the interface, and the best answer to that question is generated by a fine-tuned natural language model. The generated answer is displayed on the user's device. In addition, users can input feedback on the provided answer, which is sent to the server and used to refine the model.

[1027] Examples and prompts

[1028] For example, if a user is using a project management system and during a conversation they say, "The budget for the next project has been decided to be 5 million yen," the system will collect and analyze that conversation in real time to determine whether it contains confidential information.

[1029] An example prompt is:

[1030] Online conversation data: "The budget for the next project has been decided to be 5 million yen."

[1031] Prompt: "Is this conversation a security risk?"

[1032] This system allows users to instantly identify potential security risks that arise during conversations and take appropriate measures.

[1033] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1034] Step 1:

[1035] The device collects conversation content in real time when a user is engaged in online conversations or web conferences. It receives voice or text data as input and converts it into text using speech recognition software (e.g., Google Web Speech API). The output is text-format conversation data.

[1036] Step 2:

[1037] The device encrypts the collected text data using a secure protocol (e.g., HTTPS) and securely sends it to the server. It receives text conversation data as input, encrypts it, and then sends it to the server. The output is encrypted text data.

[1038] Step 3:

[1039] The server stores the received data in a specific database. It receives encrypted text data as input, decrypts it, and stores it in the database. The output is the text data stored in the database.

[1040] Step 4:

[1041] The server analyzes the stored data and extracts question and answer pairs using natural language processing (NLP) techniques. It receives text data as input and applies NLP algorithms to extract important keywords and question and answer pairs. The output is the extracted question and answer pairs.

[1042] Step 5:

[1043] The server uses the extracted question and answer pairs to fine-tune the natural language model. It takes the question and answer pairs as input and retrains the generative AI model. The output is a fine-tuned natural language model.

[1044] Step 6:

[1045] The terminal receives the user's question through the system interface, and sends it to the server in text format. The output is the question in text format.

[1046] Step 7:

[1047] The server analyzes the received question and generates the best answer using a fine-tuned natural language model. It takes the user's question as input and generates an answer using a generative AI model. The output is the generated answer in text format.

[1048] Step 8:

[1049] The terminal displays the generated answer to the user. It takes as input the generated textual answer and displays it in a user interface. The output is the answer that is displayed to the user.

[1050] Step 9:

[1051] The server monitors the collected online conversation data in real time and analyzes security risks. It receives the conversation data as input and evaluates the security risk using an NLP model. The output is the security risk assessment result.

[1052] Step 10:

[1053] The server notifies the user of a risk alert based on the analysis results. It receives the security risk assessment results as input, generates a message for the alert notification, and sends it to the terminal. The output is an alert message that is notified to the user.

[1054] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1055] The system for implementing this invention has the ability to take emotional state into account when generating answers by allowing users to provide online conversation data, analyzing that data, and combining it with an emotion engine that recognizes the user's emotions.

[1056] Explanation of program processing

[1057] 1. Collecting online conversation data

[1058] Device operation: When a user engages in online conversations or web conferences, the device collects the conversation content in real time and stores it in text or audio format.

[1059] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[1060] 2. Data transmission and storage

[1061] What the device does: Collected data is encrypted and sent securely to a server, which stores the data in a database with access restrictions to ensure privacy.

[1062] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[1063] 3. Data Analysis

[1064] Server operation: The server analyzes the received data and converts it into text using natural language processing (NLP) techniques. It then extracts important keywords and question / answer pairs from the converted text data.

[1065] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[1066] 4. Emotion analysis using an emotion engine

[1067] Server operation: The server analyzes the user's emotional state based on the collected data using an emotion engine. Based on the analysis results, it assigns emotion labels to the conversation data.

[1068] Example: The server assigns emotion labels such as "anxiety" or "excitement" based on the user's speech.

[1069] 5. Training the model

[1070] What the server does: It uses the extracted question-answer pairs and sentiment labels to fine-tune an existing natural language model, learning user-specific answer patterns along with their emotional states in the process.

[1071] Example: The server learns the speaking style and corresponding emotional state of a boss and trains a natural language model (e.g., GPT-3).

[1072] 6. Provision of consultation system

[1073] User action: The user inputs a question through the system interface. The terminal sends the user's question to the server.

[1074] Example: An employee types, "How can I prepare for my next meeting?"

[1075] 7. Answer Generation

[1076] Server operation: The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. The answer is generated based on the user's emotional state. The generated answer is then sent to the device.

[1077] Example: The server provides an "anxious" employee with a response such as, "To prepare for the next meeting, first check the agenda and gather any relevant materials. If you have any concerns, please check with your leader."

[1078] 8. View Answers

[1079] What the device does: The device displays the answer sent by the server to the user.

[1080] Example: The employee's device displays the following response: "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[1081] 9. Feedback and Model Improvement

[1082] User action: The user enters feedback on the provided answer. The device collects the feedback and sends it to the server.

[1083] Server action: The server analyzes the feedback and retunes the model based on the results to improve performance.

[1084] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[1085] 10. Data Anonymization

[1086] What the device does: The data collected is anonymized to protect your privacy. This process removes any personally identifiable information.

[1087] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[1088] With the above program and processing flow, the present invention is a system that allows users to consult with others and receive appropriate answers regardless of time or place. This system is particularly useful for users who find it difficult to consult directly with superiors or seniors, and provides an environment where consultations can be conducted efficiently and safely while taking into account their emotional state.

[1089] The processing flow will be explained below.

[1090] Step 1:

[1091] Users can hold online conversations or web conferences, and the device can collect and save the conversations in real time as text or audio.

[1092] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[1093] Step 2:

[1094] The device encrypts the collected data and sends it securely to a server, where it is stored in a specific database with access restrictions to ensure privacy.

[1095] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[1096] Step 3:

[1097] The server analyzes the received data and converts it into text using natural language processing (NLP) techniques, then extracts important keywords and question / answer pairs from the converted text data.

[1098] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[1099] Step 4:

[1100] The server analyzes the user's emotional state based on the collected data using an emotion engine, and assigns emotion labels to the conversation data based on the analysis results.

[1101] Example: The server assigns emotion labels such as "anxiety" or "excitement" based on the user's speech.

[1102] Step 5:

[1103] The server uses the extracted question and answer pairs and sentiment labels to fine-tune an existing natural language model, learning the user's unique answer patterns along with their emotional state in the process.

[1104] Example: The server learns the speaking style and corresponding emotional state of a boss and trains a natural language model.

[1105] Step 6:

[1106] The user inputs a question through the system interface, and the terminal sends the user's question to the server.

[1107] Example: An employee types, "How can I prepare for my next meeting?"

[1108] Step 7:

[1109] The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. The answer is generated based on the user's emotional state. The generated answer is then sent to the device.

[1110] Example: The server provides an "anxious" employee with a response such as, "To prepare for the next meeting, first check the agenda and gather any relevant materials. If you have any concerns, please check with your leader."

[1111] Step 8:

[1112] The terminal displays the answer sent from the server to the user.

[1113] Example: The employee's device displays the following response: "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[1114] Step 9:

[1115] The user enters feedback on the provided answer, and the device collects the feedback and sends it to the server.

[1116] Example: An employee gives feedback saying, "This answer was helpful."

[1117] Step 10:

[1118] The server analyzes the feedback and retunes the model based on the results to improve performance.

[1119] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[1120] Step 11:

[1121] The device anonymizes the data collected to protect privacy, a process that removes any personally identifiable information.

[1122] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[1123] Example 2

[1124] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1125] Current consultation systems that use online conversation data generate answers without considering the user's emotional state, which can result in the system failing to provide the most appropriate answer for the user. Furthermore, data privacy protection is insufficient, posing a risk of personal information leaks. The objective of this invention is to solve these problems and provide a system that generates appropriate answers that consider the user's emotional state and protects data privacy.

[1126] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1127] In this invention, the server includes terminal means for collecting online conversation data from users, terminal means for encrypting the collected data and transmitting it to the server in a secure manner, database means, means for the server to analyze the received data using natural language processing technology and extract question and answer pairs, means for the server to analyze the user's emotional state from the data using a sentiment analysis engine and assign sentiment labels, server means for fine-tuning a natural language model using the extracted question and answer pairs and sentiment labels, terminal means for accepting questions from users, means for the server to generate answers to the user's questions using the fine-tuned natural language model and sentiment analysis engine, and terminal means for displaying the generated answers to the users. This makes it possible to provide appropriate answers that take the user's emotional state into consideration and protect the privacy of the data.

[1128] "Terminal means" refers to electronic devices that users use to collect online conversation data and input questions.

[1129] "Server" refers to the computer system that receives, analyzes, and stores collected data.

[1130] "Encryption" is a technology that converts data using a specific algorithm to protect it from unauthorized access or tampering.

[1131] "Server means" refers to the hardware or software used by the server to perform a particular function.

[1132] "Database means" refers to a system for safely storing received data and making it accessible when necessary.

[1133] "Natural language processing technology" is a technology for analyzing and understanding human language, and primarily uses machine learning and language models.

[1134] A "question and answer pair" refers to a set of questions entered by a user and their corresponding appropriate answers.

[1135] An "emotion analysis engine" refers to technology that analyzes a user's emotional state from collected data and assigns various emotional labels.

[1136] An "emotion label" is a tag attached to data to indicate the user's emotional state, such as "joy," "sadness," or "anxiety."

[1137] "Fine-tuning" is the process of fine-tuning an existing model to suit a specific dataset or purpose.

[1138] "Natural language model" refers to a machine learning model trained to generate and parse human language.

[1139] "Feedback" refers to the evaluation or opinion a user gives on a provided answer.

[1140] "Anonymization" is a technique that removes or transforms personally identifiable information to protect the privacy of data.

[1141] "Display means" refers to a device or software for visually presenting the generated answer to the user.

[1142] An embodiment of this invention is described below. This system is mainly composed of three elements: a server, a terminal, and a user. This system allows users to provide online conversation data, analyzes that data, and combines it with an emotion engine that recognizes the user's emotions, thereby providing a function that takes the user's emotional state into account when generating answers.

[1143] Device behavior

[1144] When a user has an online conversation or web conference, the device collects the conversation in real time using the device's microphone and camera. The collected data is saved in audio or video format and immediately converted into text format. The data is then encrypted and sent to the server in a secure manner. The encryption techniques used include AES-256 encryption and SSL / TLS protocol.

[1145] Server Operation

[1146] The server stores the received data in a database. Access restrictions are set to this database to ensure privacy. The data received by the server is analyzed using natural language processing (NLP) techniques to extract important keywords and question-and-answer pairs. Software such as Apache Kafka and TensorFlow is used for this analysis. The extracted text data is further analyzed using a sentiment analysis engine (e.g., IBM Watson or Amazon Comprehend), and emotional labels are assigned to the data. Emotional labels such as "happiness," "sadness," and "anxiety" are taken into account.

[1147] Training the model

[1148] The server uses the extracted question and answer pairs and emotion labels to fine-tune a natural language model using Hugging Face's transformers library and OpenAI's GPT-3. The fine-tuned model learns the user's unique answer patterns and emotional state, allowing it to generate more appropriate answers.

[1149] Providing answers to users

[1150] Users input questions through the system interface (web app or mobile app). The device sends the user's question to the server, which analyzes it and generates the best answer using a fine-tuned natural language model and sentiment analysis engine. The generated answer is sent to the device and displayed to the user. For example, if an employee inputs "What should I do to prepare for the next meeting?", the server will provide a response such as "To prepare for the next meeting, first check the agenda and gather relevant materials."

[1151] Feedback and model improvement

[1152] Users can enter feedback on the answers provided, and the device collects and sends the feedback to the server, which analyzes the feedback and retunes the model based on the results to improve performance.

[1153] Data anonymization

[1154] The data collected is anonymized to protect your privacy - in this process, any personally identifiable information is removed and it is stored securely.

[1155] Examples of concrete examples and prompts

[1156] Example: After a meeting with their boss, an employee asks the system, "How should I approach the challenges in my new department?" The system recognizes that the employee is feeling anxious and provides advice that takes their feelings into consideration.

[1157] Example prompt: "I've just been assigned to a new department and I'm not sure how to solve a problem. Could you please give me some advice?"

[1158] As a result, this system provides appropriate answers that take into account the user's emotional state, protecting the privacy of data while providing an environment in which users can seek advice efficiently.

[1159] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1160] Step 1:

[1161] Collecting online conversation data

[1162] Device operation: When a user has an online conversation or web conference, the device's microphone and camera capture what is being said in real time and save it as audio or video data.

[1163] Input: User audio and video data.

[1164] Data processing: The device converts audio and video data into text format.

[1165] Output: Conversation data in text format.

[1166] What it does: Your device's microphone records your "discussion about a new project" and uses speech recognition software to convert it into text.

[1167] Step 2:

[1168] Data transmission and storage

[1169] What the device does: Encrypts the collected text data and sends it to the server in a secure manner.

[1170] Input: Conversation data in text format.

[1171] Data processing: Data is encrypted with AES-256 and transmitted using the SSL / TLS protocol.

[1172] Output: The encrypted data is sent to the server.

[1173] Specific operation: Text data collected on the device is encrypted with AES-256 and sent via SSL to the server, which then stores the received data in a database.

[1174] Step 3:

[1175] Data analysis

[1176] Server operation: The server analyzes the received text data using natural language processing technology and extracts important keywords and question-and-answer pairs.

[1177] Input: Conversation data in text format.

[1178] Data processing: Analyze using natural language processing techniques (e.g., Apache Kafka, TensorFlow) to extract keywords and question-answer pairs.

[1179] Output: Important keywords and question-answer pairs are extracted.

[1180] What it does: The server extracts keywords such as "plans to enter new markets" from the text data and identifies related questions and answers.

[1181] Step 4:

[1182] Emotion analysis using an emotion engine

[1183] Server operation: The server uses an emotion analysis engine to analyze the user's emotional state based on the data and assigns an emotion label.

[1184] Input: Keywords and question-answer pairs.

[1185] Data processing: Analyze the emotional state using a sentiment analysis engine (e.g., IBM Watson, Amazon Comprehend) and assign sentiment labels to the data.

[1186] Output: Data with emotion labels.

[1187] Specific operation: The server analyzes the emotion "anxiety" from statements such as "The project is progressing slowly" and assigns a label to it.

[1188] Step 5:

[1189] Training the model

[1190] Server operation: Fine-tune a natural language model using extracted question-answer pairs and sentiment labels.

[1191] Input: Question-answer pairs and sentiment labels.

[1192] Data processing: Fine-tune the model using Hugging Face's transformers library and OpenAI's GPT-3.

[1193] Output: A fine-tuned natural language model.

[1194] Specific operation: The server learns answers that adapt to the employee's question patterns based on guidelines and emotion labels.

[1195] Step 6:

[1196] Providing a consultation system

[1197] User action: The user enters a question through the system interface (web app or mobile app). The device sends the user's question to the server.

[1198] Input: The user's question.

[1199] Data processing: The user's question is formulated and sent to the server.

[1200] Output: The user's question sent to the server.

[1201] What happens: An employee types, "How do I prepare for a meeting?" and the device sends the question to the server.

[1202] Step 7:

[1203] Generate answers

[1204] Server operation: The server analyzes the user's question and generates the best answer using a fine-tuned natural language model and sentiment analysis engine based on the sentiment labels.

[1205] Input: User question and sentiment label.

[1206] Data processing: Generate answers using fine-tuned natural language models.

[1207] Output: The generated answer.

[1208] Specific operation: The server receives a question about "preparing for the meeting," generates a response such as "First, check the agenda and gather the materials," and sends it to the device.

[1209] Step 8:

[1210] Show Answers

[1211] What the device does: The device displays the answer sent by the server to the user.

[1212] Input: The answer sent by the server.

[1213] Data processing: Display the answer to the user visually or audibly.

[1214] Output: The answer displayed to the user.

[1215] What it does: The device prompts the employee, "To prepare for your next meeting, first review the agenda and gather any relevant materials."

[1216] Step 9:

[1217] Feedback and model improvement

[1218] User action: The user enters feedback on the provided answer. The device sends the feedback to the server.

[1219] Input: User feedback.

[1220] Data processing: Analyze feedback and use it to improve the model.

[1221] Output: An improved natural language model.

[1222] Specific behavior: The user enters feedback such as "This answer was helpful," and the device sends the feedback to the server. The server analyzes the feedback and recalibrates the model.

[1223] Step 10:

[1224] Data anonymization

[1225] What the device does: Anonymizes collected data to protect privacy.

[1226] Input: Conversation data, including personal information.

[1227] Data processing: removing or transforming personally identifiable information.

[1228] Output: Anonymized data.

[1229] What it does: It removes and securely stores your name and unique identifiers.

[1230] (Application example 2)

[1231] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1232] When it comes to communication between workers and robots in factories, there is a lack of a way to understand the emotions of workers and respond appropriately. This can have a negative impact on productivity and safety, especially in emergencies or stressful situations. To solve this problem, a system is needed that can analyze the emotional state of workers and respond appropriately according to that state.

[1233] The identification process by the identification 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 online conversation data from users, means for analyzing the collected data to extract question and answer pairs, means for fine-tuning a natural language model using the extracted question and answer pairs, means for analyzing the emotional state of the user using an emotion engine based on the collected data, means for generating an adapted answer based on the emotional state, and means for displaying the generated answer to the user. This makes it possible to understand the emotional state of employees and respond appropriately.

[1234] "User" refers to the factory employees who use the system.

[1235] "Online conversation data" refers to one-to-one voice or text interactions between employees and robots.

[1236] "Means of collection" refers to hardware and software that records or records online conversations between employees and robots.

[1237] "Means for analyzing and extracting question and answer pairs" refers to technology that uses natural language processing technology to analyze collected online conversation data and identify important questions and their answers.

[1238] "Fine-tuning" refers to the process of optimizing the performance of an existing natural language model using extracted question and answer pairs.

[1239] "Emotion engine" refers to software used to analyze users' emotions from online conversation data and identify their emotional states.

[1240] "Means for analyzing emotional state" refers to the process for determining a user's emotions from collected data using an emotion engine.

[1241] "Means for generating adapted answers" refers to techniques for generating appropriate and optimal answers based on the user's emotional state.

[1242] "Means for displaying" refers to a display or audio output device for visually or audibly communicating the generated response to the employee.

[1243] A system for implementing this invention allows a user to provide online conversation data, analyzes the data, and combines an emotion engine that recognizes the user's emotions with a natural language model to take the user's emotional state into account when generating responses. This system is particularly applicable to factory robots and is designed to facilitate smooth communication between factory workers and robots.

[1244] Hardware and Software Configuration

[1245] Hardware: Robots, microphones, and servers located in the factory

[1246] Software: Natural language processing engines (e.g., Transformer's NLP), sentiment analysis engines (e.g., Sentiment Analysis Pipeline), generative AI models (e.g., GPT-3)

[1247] Specific explanation of the system's operation

[1248] Collection and transmission of online conversation data

[1249] The robot collects online conversations with employees in real time via a microphone, and the collected voice data is encrypted and securely sent to a server.

[1250] Data analysis and text conversion

[1251] The server analyzes the received voice data and converts it into text using natural language processing (NLP) techniques, extracting important keywords and question / answer pairs from the converted text data.

[1252] Emotion analysis using an emotion engine

[1253] Next, the server analyzes the user's emotional state using an emotion engine and assigns emotion labels to the conversation data, thereby understanding the user's emotional state.

[1254] Training the model

[1255] The extracted question-answer pairs and sentiment labels are used to fine-tune an existing natural language model (e.g., GPT-3), learning user-specific answer patterns along with their emotional states in the process.

[1256] Responding to questions

[1257] When an employee types a question through the system's interface, the robot sends it to the server, which uses a fine-tuned natural language model and emotion engine to generate the best answer and communicate it to the user via a display or voice output device.

[1258] Specific examples

[1259] For example, if an employee in the quality control department asks the robot, "I don't know how to inspect the product," the robot's emotion analysis engine will recognize that the employee is feeling impatient. The generative AI model will respond by offering guidance that is adapted to the employee's impatient state: "Please stay calm. Let's inspect the product by following these steps."

[1260] Prompt Sentence Examples

[1261] "If an employee notices something is wrong but is feeling anxious, how do I support them?"

[1262] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1263] Step 1:

[1264] The terminal collects online conversation data. Specifically, the factory robot records the conversation with the employee in real time through a microphone. The collected data is in the form of audio, and the audio data is temporarily stored on the terminal for further processing.

[1265] Step 2:

[1266] The device encrypts the collected voice data and securely transmits it to a server using a secure communication protocol (e.g., HTTPS). The server then stores the received voice data in a database with access restrictions to ensure privacy.

[1267] Step 3:

[1268] The server analyzes the received voice data and converts it into text using natural language processing (NLP) techniques, such as using Transformers NLP models to convert speech to text, and stores the resulting text for use in the next step.

[1269] Step 4:

[1270] The server extracts important keywords and question / answer pairs from the converted text data. It uses natural language processing technology to identify which parts are questions and which parts are answers and extracts those pairs. This allows it to understand what information the user is looking for.

[1271] Step 5:

[1272] The server uses an emotion engine to analyze the user's emotional state based on the collected text data. An emotion analysis engine (e.g., Sentiment Analysis Pipeline) is used to identify emotions such as "anxiety" or "impatience" from the user's comments. Based on the results of this analysis, an emotion label is assigned to the text data.

[1273] Step 6:

[1274] The server uses the extracted question and answer pairs and sentiment labels to fine-tune existing natural language models, such as using GPT-3 to learn employee speech patterns and their corresponding emotional states. This process produces a model that understands the user's unique communication style.

[1275] Step 7:

[1276] The user inputs a question through the system interface. The factory robot recognizes this question and sends it to the server. The question is then analyzed in the next step.

[1277] Step 8:

[1278] The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. For example, if the user is in an "anxious" emotional state, the server will provide a calm tone and specific advice that is adapted to that state. The generated answer is then sent to the device.

[1279] Step 9:

[1280] The terminal displays the response sent from the server to the user, and the factory robot communicates the response to the employee via voice, allowing the user to take appropriate action.

[1281] Step 10:

[1282] The user provides feedback on the answers provided by the device, which collects this feedback and sends it to the server, which analyzes it and uses the results to retune the model and improve the system's performance.

[1283] Step 11:

[1284] The device anonymizes the data collected to protect your privacy, ensuring that personally identifiable information is removed and stored securely.

[1285] Example prompt sentence:

[1286] "If an employee notices something is wrong but is feeling anxious, how do I support them?"

[1287] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1288] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1289] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1290] [Fourth embodiment]

[1291] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1292] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1293] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1294] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1295] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1296] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1297] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1298] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1299] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1300] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1301] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1302] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1303] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1304] A system for implementing the present invention is configured by combining various software and hardware components for users to provide online conversation data and for analyzing that data.

[1305] Explanation of program processing

[1306] 1. Collecting online conversation data

[1307] What the device does: When a user has an online conversation or web conference, the device collects the conversation in real time and stores the collected data in text or audio format.

[1308] Example: An employee has a web conference with their manager about a particular project, and the content of the conference is collected.

[1309] 2. Data transmission and storage

[1310] What the device does: Collected data is encrypted and securely sent to the server.

[1311] Server behavior: The server stores the received data in a specific database with restricted access to ensure privacy.

[1312] 3. Data Analysis

[1313] Server operation: The server analyzes the received data and converts it into text using natural language processing (NLP) techniques. It then extracts important keywords and question / answer pairs from the converted text data.

[1314] Example: The server extracts questions and answers about "plans to enter new markets" from a conversation between a boss and an employee.

[1315] 4. Training the model

[1316] What the server does: It uses the extracted question and answer pairs to fine-tune an existing natural language model. Specifically, the fine-tuned model learns the user's unique answer patterns.

[1317] Example: The server uses the extracted data to train a natural language model to learn the speaking style of your boss.

[1318] 5. Provision of consultation system

[1319] User action: The user enters a question through the system's interface.

[1320] What the terminal does: Sends the entered question to the server.

[1321] What the server does: Analyzes the received question, uses a fine-tuned natural language model to generate the best answer, and sends this answer to the device.

[1322] Terminal behavior: Display the generated answer to the user.

[1323] For example: An employee asks, "How can I prepare for my next meeting?" and is provided with the best answer to that question.

[1324] 6. Feedback and model improvement

[1325] User Action: The user enters feedback on the answer provided.

[1326] Device Action: Sends feedback to the server.

[1327] Server action: The server analyzes the feedback and uses the results to retune the model and improve performance.

[1328] 7. Data anonymization

[1329] Device behavior: The data collected is anonymized to protect your privacy. This process removes any personally identifiable information, allowing you to use the system with confidence.

[1330] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized, so personal information will be protected.

[1331] With the above program and processing flow, the present invention is a system that allows users to consult with others and receive accurate answers regardless of time or place. This system is particularly useful for users who find it difficult to consult with superiors or seniors directly, and provides an environment where consultations can be conducted efficiently and safely.

[1332] The processing flow will be explained below.

[1333] Step 1:

[1334] Users can hold online conversations or web conferences, and the device can collect and store the conversations in real time in text or audio format.

[1335] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[1336] Step 2:

[1337] The device encrypts the collected data and sends it securely to a server, which then stores it in a database with access restrictions to ensure privacy.

[1338] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[1339] Step 3:

[1340] The server analyzes the received data and converts it into text using natural language processing (NLP) techniques, then extracts important keywords and question / answer pairs from the converted text data.

[1341] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[1342] Step 4:

[1343] The server uses the extracted question and answer pairs to fine-tune an existing natural language model (e.g., GPT-3), learning user-specific answer patterns in the process.

[1344] Example: The server feeds the extracted dataset into a natural language model to learn the speaking style of your boss.

[1345] Step 5:

[1346] The user inputs a question through the system interface, and the terminal sends the user's question to the server.

[1347] Example: An employee types, "How can I prepare for my next meeting?"

[1348] Step 6:

[1349] The server analyzes the user's question, uses a fine-tuned natural language model to generate the best answer, and sends the generated answer to the device.

[1350] Example: The server invokes a natural language model to answer "how to prepare for my next meeting."

[1351] Step 7:

[1352] The terminal displays the answer sent from the server to the user. Example: An employee's terminal displays the answer "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[1353] Step 8:

[1354] The user enters feedback on the provided answer, and the device collects the feedback and sends it to the server.

[1355] Example: An employee gives feedback saying, "This answer was helpful."

[1356] Step 9:

[1357] The server analyzes the feedback and retunes the model based on the results to improve performance.

[1358] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[1359] Step 10:

[1360] The device anonymizes the data collected to protect privacy, a process that removes any personally identifiable information.

[1361] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[1362] Example 1

[1363] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1364] Conventional online conversation systems make it difficult for users to seek advice or obtain information in real time, and information sharing is hindered, especially in environments where it is difficult to ask questions directly to superiors or senior colleagues. Furthermore, the quality of answers is inconsistent, preventing users from accurately obtaining useful information. Furthermore, from the perspective of privacy protection, secure data collection and analysis are required. Therefore, there is a need for the development of a system that can effectively collect and analyze online conversation data, provide appropriate answers to users, and ensure privacy protection.

[1365] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1366] In this invention, the server includes means for collecting online communication data from users, means for encrypting the collected data and securely transmitting it to the server, means for analyzing the received data and converting it into text using natural language processing technology, means for extracting important keywords and question-and-answer pairs from the converted text data, means for fine-tuning an existing generative AI model using the extracted question-and-answer pairs, means for accepting questions from users and generating optimal answers using the fine-tuned generative AI model, and means for displaying the generated answers to the users. This makes it possible to effectively collect and analyze online conversation data and provide appropriate answers to users.

[1367] "Online communication data" refers to data related to communication such as conversations and text messages that users have online.

[1368] "Encryption" is the process of transforming data using a specific algorithm to protect it from unauthorized access.

[1369] A "server" is a computer that provides various services within a network, and refers to hardware that runs software with specific functions.

[1370] "Natural language processing technology" refers to technology for processing and analyzing human language using computers, and includes tokenization, part-of-speech tagging, named entity recognition, etc.

[1371] "Convert to text" refers to the process of converting non-text data such as audio or images into data in character string format.

[1372] "Keywords" are words or phrases that contain particularly important information in the data, and are elements that serve as clues for understanding the data.

[1373] A "question and answer pair" is a dataset that combines a user's question with an appropriate answer.

[1374] A "generative AI model" is an artificial intelligence model that generates new information based on given input data.

[1375] "Fine-tuning" refers to the process of optimizing an existing artificial intelligence model to suit a specific task or user needs, and involves additional training.

[1376] "Feedback" refers to evaluations and opinions on the information and answers provided by users, which are useful for improving the system.

[1377] "Anonymization" is the removal of personally identifiable information from data, and is a privacy-protecting technique.

[1378] MODE FOR CARRYING OUT THE INVENTION

[1379] The system for carrying out the present invention is configured by combining various software and hardware for users to provide online conversation data and for analyzing the data.

[1380] Hardware and Software Configuration

[1381] Device configuration

[1382] A terminal is a device used by a user for online conversations and web conferences, such as a PC, smartphone, tablet, etc. The terminal has speech recognition software (e.g., Google Speech-to-Text API) and encryption software (e.g., AES encryption) installed.

[1383] Server Configuration

[1384] The server is a computer used to analyze and store collected data and fine-tune the generative AI model. The following software is installed on the server:

[1385] Natural language processing technology (e.g., SpaCy)

[1386] Database system (e.g. MySQL)

[1387] Generative AI models (e.g., GPT-3)

[1388] A concrete example of the processing flow

[1389] 1. Collecting online conversation data

[1390] The device collects conversations in real time when users are having online conversations or web conferences. The conversations are captured as voice data and converted into text data using voice recognition software. For example, when an employee discusses "sales strategies for a new product" with their boss in a web conference, the voice data is converted into text.

[1391] 2. Data transmission and storage

[1392] The collected text data is encrypted and sent to the server using AES encryption technology and HTTPS protocol, where it is decrypted and stored in a MySQL database.

[1393] 3. Data Analysis

[1394] The server analyzes the text data using natural language processing technology (e.g., SpaCy) and extracts important keywords and question-and-answer pairs. For example, questions and answers related to "sales strategies for new products" are extracted. The results of this analysis are also stored in a database.

[1395] 4. Training the model

[1396] The server uses the extracted question-answer pairs to fine-tune a generative AI model (e.g., GPT-3). This process optimizes the model over multiple epochs and learns the user's unique answer patterns. For example, the model is tuned to learn the speaking style of a boss.

[1397] 5. Provision of consultation system

[1398] Users input questions into the system through a dedicated interface. The input question is encoded and sent to the server. The server analyzes the question and uses a fine-tuned generative AI model to generate the best answer, which is then sent to the device. The device then displays the answer to the user. For example, if an employee asks, "How should I prepare for my next meeting?", the best answer will be displayed.

[1399] 6. Feedback and model improvement

[1400] Users input feedback on the answers they provide through the interface, which is sent to the server for analysis, and the server uses the feedback to retune the generative AI model and improve its performance.

[1401] 7. Data anonymization

[1402] The device anonymizes the collected data to protect privacy. This is a process that removes any personally identifiable information. For example, "consultation about interpersonal relationships at work" is anonymized to protect personal information.

[1403] Examples of prompt statements

[1404] Examples of prompts used in this system include:

[1405] "If you were to extract questions and answers about a sales strategy for a new product during a web conference with your boss and ask about preparations for the next meeting, what would your answer be?"

[1406] "Generate relevant advice from anonymized consultations about workplace relationships."

[1407] The above is an embodiment of the invention.

[1408] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1409] Step 1: Collect online conversation data

[1410] The device collects the content of conversations in real time when users are having online conversations or web conferences. The content of the conversation is captured as audio data and converted into text data using speech recognition software (e.g., Google Speech-to-Text API). Specifically, the collected audio data is input into the speech recognition software and output as text data. For example, when an employee discusses "sales strategies for a new product" with his or her boss in a web conference, the audio data is converted into text data.

[1411] Step 2: Send and store data

[1412] The terminal encrypts the collected text data (e.g., AES encryption). The encrypted data is sent to the server via the network. The server decrypts the received encrypted data and stores it in a database (e.g., MySQL). Specifically, the data is entered into a database system and stored as decrypted text data from its encrypted state. For example, the contents of a meeting may be encrypted and sent to the server, where it is decrypted and stored.

[1413] Step 3: Analyze the data

[1414] The server analyzes the stored data using natural language processing technology (e.g., SpaCy). Important keywords and question-and-answer pairs are extracted from the text data. Specifically, the text data is input into the natural language processing software, which then outputs the analyzed keywords and question-and-answer pairs. For example, questions and answers related to "sales strategies for new products" are extracted.

[1415] Step 4: Train the model

[1416] The server uses the extracted question-answer pairs to fine-tune a generative AI model (e.g., GPT-3). This fine-tuning allows the model to learn the user's unique answer patterns. Specifically, the question-answer pair data is input into the training interface of the generative AI model, and a fine-tuned model is output. For example, the model is adjusted to learn the speaking style of a boss.

[1417] Step 5: Providing a consultation system

[1418] A user inputs a question via a system interface (e.g., a web application). The device encodes the input question and sends it to the server. The server analyzes the received question and generates an optimal answer using a fine-tuned generative AI model. The answer is sent to the device, which then displays the generated answer to the user. Specifically, the system encodes the user input and sends it to the server, then decodes and displays the generated answer. For example, if an employee asks, "How should I prepare for the next meeting?", the optimal answer will be displayed.

[1419] Step 6: Feedback and model improvement

[1420] The user inputs feedback on the provided answer via the interface. The device encodes the feedback data and sends it to the server. The server analyzes the feedback and readjusts the generative AI model based on the results, improving its performance. Specifically, the feedback data is input into the analysis system, which outputs a readjusted model. For example, an evaluation of the appropriateness of the provided answer can be input to improve the quality of the model.

[1421] Step 7: Anonymize the data

[1422] The device anonymizes the collected data to protect privacy. This process removes any personally identifiable information. The server stores and analyzes the anonymized data to ensure the user's privacy. Specifically, the system filters personal information from the data and uses an anonymization tool to replace proper nouns with generic terms. For example, "consultation about interpersonal relationships at work" is anonymized, protecting personal information.

[1423] (Application example 1)

[1424] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1425] In today's information society, online communication is increasing. In this context, it is essential for companies and individuals to manage not only knowledge but also security risks. However, there is a lack of systems that can detect security risks in real time and respond immediately. The objective of this invention is to provide a system that collects and analyzes online conversation data, analyzes security risks in real time, and notifies users.

[1426] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1427] In this invention, the server includes means for collecting online conversation data from users, means for analyzing the collected data to extract question and answer pairs, means for fine-tuning a natural language model using the extracted question and answer pairs, means for generating answers to questions from users using the fine-tuned natural language model, means for collecting online conversation data in real time and analyzing security risks, and means for notifying users of risk alerts based on the analysis results. This makes it possible to efficiently manage online conversation data, detect security risks in real time, and respond immediately.

[1428] "Means for collecting online conversation data from users" refers to a combination of hardware and software for collecting the content of online meetings and conversations conducted by users in real time and saving it in text or audio format.

[1429] "Means for analyzing collected data to extract question and answer pairs" refers to software processes that apply natural language processing techniques to collected online conversation data to identify and extract key question and answer pairs.

[1430] "Means for fine-tuning natural language models using extracted question and answer pairs" refers to techniques for incrementally training existing natural language models using extracted data to improve their accuracy and responsiveness in specific contexts.

[1431] "Means for generating answers to questions from users using a fine-tuned natural language model" refers to algorithms and processes for generating optimal answers to questions entered by users using a fine-tuned natural language model.

[1432] "Means for displaying generated answers to a user" refers to an interface and software for displaying answers generated by a natural language model on a user's device.

[1433] "Means for collecting online conversation data in real time and analyzing security risks" refers to a combination of hardware and software for collecting online conversation data in real time and analyzing the data to detect security risks.

[1434] "Means for notifying users of risk alerts based on the analysis results" refers to alert systems and software that immediately notify users of security risks detected from the analyzed data.

[1435] The system for implementing this invention is configured by combining various software and hardware for allowing users to collect online conversation data in real time using smartphones or other devices and analyze that data.

[1436] System Overview

[1437] Acquisition System:

[1438] The device collects the conversation content in real time when the user is having an online conversation or web conference. The collected data is saved in text or audio format. Specifically, the voice data is converted to text using the Google Web Speech API. The collected text data is securely transmitted from the user's device to a server.

[1439] Data transmission and storage:

[1440] The device encrypts the collected data and sends it securely to a server, which then stores it in a specific database with access restrictions to ensure privacy.

[1441] Data Analysis:

[1442] The server parses the incoming data and converts it into text using natural language processing (NLP) techniques. From the converted text data, question and answer pairs are extracted. This extracted data is used to fine-tune an existing natural language model.

[1443] Security Risk Analysis:

[1444] The collected online conversation data is monitored in real time and security risks are analyzed. Risks are determined based on specific trigger words (e.g., "password," "confidential," etc.). This is done using an NLP model.

[1445] Risk Alert Notification:

[1446] Based on the analysis results, users are immediately notified of any security risks that are discovered. Once the data has been analyzed, the server sends an alert to the user along with a risk assessment. This notification is sent via an interface such as a push notification on a smartphone.

[1447] User Interface:

[1448] Users input questions through the interface, and the best answer to that question is generated by a fine-tuned natural language model. The generated answer is displayed on the user's device. In addition, users can input feedback on the provided answer, which is sent to the server and used to refine the model.

[1449] Examples and prompts

[1450] For example, if a user is using a project management system and during a conversation they say, "The budget for the next project has been decided to be 5 million yen," the system will collect and analyze that conversation in real time to determine whether it contains confidential information.

[1451] An example prompt is:

[1452] Online conversation data: "The budget for the next project has been decided to be 5 million yen."

[1453] Prompt: "Is this conversation a security risk?"

[1454] This system allows users to instantly identify potential security risks that arise during conversations and take appropriate measures.

[1455] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1456] Step 1:

[1457] The device collects conversation content in real time when a user is engaged in online conversations or web conferences. It receives voice or text data as input and converts it into text using speech recognition software (e.g., Google Web Speech API). The output is text-format conversation data.

[1458] Step 2:

[1459] The device encrypts the collected text data using a secure protocol (e.g., HTTPS) and securely sends it to the server. It receives text conversation data as input, encrypts it, and then sends it to the server. The output is encrypted text data.

[1460] Step 3:

[1461] The server stores the received data in a specific database. It receives encrypted text data as input, decrypts it, and stores it in the database. The output is the text data stored in the database.

[1462] Step 4:

[1463] The server analyzes the stored data and extracts question and answer pairs using natural language processing (NLP) techniques. It receives text data as input and applies NLP algorithms to extract important keywords and question and answer pairs. The output is the extracted question and answer pairs.

[1464] Step 5:

[1465] The server uses the extracted question and answer pairs to fine-tune the natural language model. It takes the question and answer pairs as input and retrains the generative AI model. The output is a fine-tuned natural language model.

[1466] Step 6:

[1467] The terminal receives the user's question through the system interface, and sends it to the server in text format. The output is the question in text format.

[1468] Step 7:

[1469] The server analyzes the received question and generates the best answer using a fine-tuned natural language model. It takes the user's question as input and generates an answer using a generative AI model. The output is the generated answer in text format.

[1470] Step 8:

[1471] The terminal displays the generated answer to the user. It takes as input the generated textual answer and displays it in a user interface. The output is the answer that is displayed to the user.

[1472] Step 9:

[1473] The server monitors the collected online conversation data in real time and analyzes security risks. It receives the conversation data as input and evaluates the security risk using an NLP model. The output is the security risk assessment result.

[1474] Step 10:

[1475] The server notifies the user of a risk alert based on the analysis results. It receives the security risk assessment results as input, generates a message for the alert notification, and sends it to the terminal. The output is an alert message that is notified to the user.

[1476] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1477] The system for implementing this invention has the ability to take emotional state into account when generating answers by allowing users to provide online conversation data, analyzing that data, and combining it with an emotion engine that recognizes the user's emotions.

[1478] Explanation of program processing

[1479] 1. Collecting online conversation data

[1480] Device operation: When a user engages in online conversations or web conferences, the device collects the conversation content in real time and stores it in text or audio format.

[1481] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[1482] 2. Data transmission and storage

[1483] What the device does: Collected data is encrypted and sent securely to a server, which stores the data in a database with access restrictions to ensure privacy.

[1484] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[1485] 3. Data Analysis

[1486] Server operation: The server analyzes the received data and converts it into text using natural language processing (NLP) techniques. It then extracts important keywords and question / answer pairs from the converted text data.

[1487] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[1488] 4. Emotion analysis using an emotion engine

[1489] Server operation: The server analyzes the user's emotional state based on the collected data using an emotion engine. Based on the analysis results, it assigns emotion labels to the conversation data.

[1490] Example: The server assigns emotion labels such as "anxiety" or "excitement" based on the user's speech.

[1491] 5. Training the model

[1492] What the server does: It uses the extracted question-answer pairs and sentiment labels to fine-tune an existing natural language model, learning user-specific answer patterns along with their emotional states in the process.

[1493] Example: The server learns the speaking style and corresponding emotional state of a boss and trains a natural language model (e.g., GPT-3).

[1494] 6. Provision of consultation system

[1495] User action: The user inputs a question through the system interface. The terminal sends the user's question to the server.

[1496] Example: An employee types, "How can I prepare for my next meeting?"

[1497] 7. Answer Generation

[1498] Server operation: The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. The answer is generated based on the user's emotional state. The generated answer is then sent to the device.

[1499] Example: The server provides an "anxious" employee with a response such as, "To prepare for the next meeting, first check the agenda and gather any relevant materials. If you have any concerns, please check with your leader."

[1500] 8. View Answers

[1501] What the device does: The device displays the answer sent by the server to the user.

[1502] Example: The employee's device displays the following response: "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[1503] 9. Feedback and Model Improvement

[1504] User action: The user enters feedback on the provided answer. The device collects the feedback and sends it to the server.

[1505] Server action: The server analyzes the feedback and retunes the model based on the results to improve performance.

[1506] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[1507] 10. Data Anonymization

[1508] What the device does: The data collected is anonymized to protect your privacy. This process removes any personally identifiable information.

[1509] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[1510] With the above program and processing flow, the present invention is a system that allows users to consult with others and receive appropriate answers regardless of time or place. This system is particularly useful for users who find it difficult to consult directly with superiors or seniors, and provides an environment where consultations can be conducted efficiently and safely while taking into account their emotional state.

[1511] The processing flow will be explained below.

[1512] Step 1:

[1513] Users can hold online conversations or web conferences, and the device can collect and save the conversations in real time as text or audio.

[1514] Example: An employee holds a web conference with their boss about a specific project, and the content of the conference is collected on the device.

[1515] Step 2:

[1516] The device encrypts the collected data and sends it securely to a server, where it is stored in a specific database with access restrictions to ensure privacy.

[1517] Example: The device sends the recording data of a web conference to the server, which then stores it in a database.

[1518] Step 3:

[1519] The server analyzes the received data and converts it into text using natural language processing (NLP) techniques, then extracts important keywords and question / answer pairs from the converted text data.

[1520] Example: A server extracts questions and answers about "plans to enter new markets" from text data.

[1521] Step 4:

[1522] The server analyzes the user's emotional state based on the collected data using an emotion engine, and assigns emotion labels to the conversation data based on the analysis results.

[1523] Example: The server assigns emotion labels such as "anxiety" or "excitement" based on the user's speech.

[1524] Step 5:

[1525] The server uses the extracted question and answer pairs and sentiment labels to fine-tune an existing natural language model, learning the user's unique answer patterns along with their emotional state in the process.

[1526] Example: The server learns the speaking style and corresponding emotional state of a boss and trains a natural language model.

[1527] Step 6:

[1528] The user inputs a question through the system interface, and the terminal sends the user's question to the server.

[1529] Example: An employee types, "How can I prepare for my next meeting?"

[1530] Step 7:

[1531] The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. The answer is generated based on the user's emotional state. The generated answer is then sent to the device.

[1532] Example: The server provides an "anxious" employee with a response such as, "To prepare for the next meeting, first check the agenda and gather any relevant materials. If you have any concerns, please check with your leader."

[1533] Step 8:

[1534] The terminal displays the answer sent from the server to the user.

[1535] Example: The employee's device displays the following response: "To prepare for the next meeting, please first review the agenda and gather any relevant materials."

[1536] Step 9:

[1537] The user enters feedback on the provided answer, and the device collects the feedback and sends it to the server.

[1538] Example: An employee gives feedback saying, "This answer was helpful."

[1539] Step 10:

[1540] The server analyzes the feedback and retunes the model based on the results to improve performance.

[1541] Example: The server uses the feedback to further improve the accuracy of the natural language model.

[1542] Step 11:

[1543] The device anonymizes the data collected to protect privacy, a process that removes any personally identifiable information.

[1544] Example: Even if a user asks for advice about interpersonal relationships at work, the content will be anonymized and stored securely.

[1545] Example 2

[1546] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1547] Current consultation systems that use online conversation data generate answers without considering the user's emotional state, which can result in the system failing to provide the most appropriate answer for the user. Furthermore, data privacy protection is insufficient, posing a risk of personal information leaks. The objective of this invention is to solve these problems and provide a system that generates appropriate answers that consider the user's emotional state and protects data privacy.

[1548] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1549] In this invention, the server includes terminal means for collecting online conversation data from users, terminal means for encrypting the collected data and transmitting it to the server in a secure manner, database means, means for the server to analyze the received data using natural language processing technology and extract question and answer pairs, means for the server to analyze the user's emotional state from the data using a sentiment analysis engine and assign sentiment labels, server means for fine-tuning a natural language model using the extracted question and answer pairs and sentiment labels, terminal means for accepting questions from users, means for the server to generate answers to the user's questions using the fine-tuned natural language model and sentiment analysis engine, and terminal means for displaying the generated answers to the users. This makes it possible to provide appropriate answers that take the user's emotional state into consideration and protect the privacy of the data.

[1550] "Terminal means" refers to electronic devices that users use to collect online conversation data and input questions.

[1551] "Server" refers to the computer system that receives, analyzes, and stores collected data.

[1552] "Encryption" is a technology that converts data using a specific algorithm to protect it from unauthorized access or tampering.

[1553] "Server means" refers to the hardware or software used by the server to perform a particular function.

[1554] "Database means" refers to a system for safely storing received data and making it accessible when necessary.

[1555] "Natural language processing technology" is a technology for analyzing and understanding human language, and primarily uses machine learning and language models.

[1556] A "question and answer pair" refers to a set of questions entered by a user and their corresponding appropriate answers.

[1557] An "emotion analysis engine" refers to technology that analyzes a user's emotional state from collected data and assigns various emotional labels.

[1558] An "emotion label" is a tag attached to data to indicate the user's emotional state, such as "joy," "sadness," or "anxiety."

[1559] "Fine-tuning" is the process of fine-tuning an existing model to suit a specific dataset or purpose.

[1560] "Natural language model" refers to a machine learning model trained to generate and parse human language.

[1561] "Feedback" refers to the evaluation or opinion a user gives on a provided answer.

[1562] "Anonymization" is a technique that removes or transforms personally identifiable information to protect the privacy of data.

[1563] "Display means" refers to a device or software for visually presenting the generated answer to the user.

[1564] An embodiment of this invention is described below. This system is mainly composed of three elements: a server, a terminal, and a user. This system allows users to provide online conversation data, analyzes that data, and combines it with an emotion engine that recognizes the user's emotions, thereby providing a function that takes the user's emotional state into account when generating answers.

[1565] Device behavior

[1566] When a user has an online conversation or web conference, the device collects the conversation in real time using the device's microphone and camera. The collected data is saved in audio or video format and immediately converted into text format. The data is then encrypted and sent to the server in a secure manner. The encryption techniques used include AES-256 encryption and SSL / TLS protocol.

[1567] Server Operation

[1568] The server stores the received data in a database. Access restrictions are set to this database to ensure privacy. The data received by the server is analyzed using natural language processing (NLP) techniques to extract important keywords and question-and-answer pairs. Software such as Apache Kafka and TensorFlow is used for this analysis. The extracted text data is further analyzed using a sentiment analysis engine (e.g., IBM Watson or Amazon Comprehend), and emotional labels are assigned to the data. Emotional labels such as "happiness," "sadness," and "anxiety" are taken into account.

[1569] Training the model

[1570] The server uses the extracted question and answer pairs and emotion labels to fine-tune a natural language model using Hugging Face's transformers library and OpenAI's GPT-3. The fine-tuned model learns the user's unique answer patterns and emotional state, allowing it to generate more appropriate answers.

[1571] Providing answers to users

[1572] Users input questions through the system interface (web app or mobile app). The device sends the user's question to the server, which analyzes it and generates the best answer using a fine-tuned natural language model and sentiment analysis engine. The generated answer is sent to the device and displayed to the user. For example, if an employee inputs "What should I do to prepare for the next meeting?", the server will provide a response such as "To prepare for the next meeting, first check the agenda and gather relevant materials."

[1573] Feedback and model improvement

[1574] Users can enter feedback on the answers provided, and the device collects and sends the feedback to the server, which analyzes the feedback and retunes the model based on the results to improve performance.

[1575] Data anonymization

[1576] The data collected is anonymized to protect your privacy - in this process, any personally identifiable information is removed and it is stored securely.

[1577] Examples of concrete examples and prompts

[1578] Example: After a meeting with their boss, an employee asks the system, "How should I approach the challenges in my new department?" The system recognizes that the employee is feeling anxious and provides advice that takes their feelings into consideration.

[1579] Example prompt: "I've just been assigned to a new department and I'm not sure how to solve a problem. Could you please give me some advice?"

[1580] As a result, this system provides appropriate answers that take into account the user's emotional state, protecting the privacy of data while providing an environment in which users can seek advice efficiently.

[1581] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1582] Step 1:

[1583] Collecting online conversation data

[1584] Device operation: When a user has an online conversation or web conference, the device's microphone and camera capture what is being said in real time and save it as audio or video data.

[1585] Input: User audio and video data.

[1586] Data processing: The device converts audio and video data into text format.

[1587] Output: Conversation data in text format.

[1588] What it does: Your device's microphone records your "discussion about a new project" and uses speech recognition software to convert it into text.

[1589] Step 2:

[1590] Data transmission and storage

[1591] What the device does: Encrypts the collected text data and sends it to the server in a secure manner.

[1592] Input: Conversation data in text format.

[1593] Data processing: Data is encrypted with AES-256 and transmitted using the SSL / TLS protocol.

[1594] Output: The encrypted data is sent to the server.

[1595] Specific operation: Text data collected on the device is encrypted with AES-256 and sent via SSL to the server, which then stores the received data in a database.

[1596] Step 3:

[1597] Data analysis

[1598] Server operation: The server analyzes the received text data using natural language processing technology and extracts important keywords and question-and-answer pairs.

[1599] Input: Conversation data in text format.

[1600] Data processing: Analyze using natural language processing techniques (e.g., Apache Kafka, TensorFlow) to extract keywords and question-answer pairs.

[1601] Output: Important keywords and question-answer pairs are extracted.

[1602] What it does: The server extracts keywords such as "plans to enter new markets" from the text data and identifies related questions and answers.

[1603] Step 4:

[1604] Emotion analysis using an emotion engine

[1605] Server operation: The server uses an emotion analysis engine to analyze the user's emotional state based on the data and assigns an emotion label.

[1606] Input: Keywords and question-answer pairs.

[1607] Data processing: Analyze the emotional state using a sentiment analysis engine (e.g., IBM Watson, Amazon Comprehend) and assign sentiment labels to the data.

[1608] Output: Data with emotion labels.

[1609] Specific operation: The server analyzes the emotion "anxiety" from statements such as "The project is progressing slowly" and assigns a label to it.

[1610] Step 5:

[1611] Training the model

[1612] Server operation: Fine-tune a natural language model using extracted question-answer pairs and sentiment labels.

[1613] Input: Question-answer pairs and sentiment labels.

[1614] Data processing: Fine-tune the model using Hugging Face's transformers library and OpenAI's GPT-3.

[1615] Output: A fine-tuned natural language model.

[1616] Specific operation: The server learns answers that adapt to the employee's question patterns based on guidelines and emotion labels.

[1617] Step 6:

[1618] Providing a consultation system

[1619] User action: The user enters a question through the system interface (web app or mobile app). The device sends the user's question to the server.

[1620] Input: The user's question.

[1621] Data processing: The user's question is formulated and sent to the server.

[1622] Output: The user's question sent to the server.

[1623] What happens: An employee types, "How do I prepare for a meeting?" and the device sends the question to the server.

[1624] Step 7:

[1625] Generate answers

[1626] Server operation: The server analyzes the user's question and generates the best answer using a fine-tuned natural language model and sentiment analysis engine based on the sentiment labels.

[1627] Input: User question and sentiment label.

[1628] Data processing: Generate answers using fine-tuned natural language models.

[1629] Output: The generated answer.

[1630] Specific operation: The server receives a question about "preparing for the meeting," generates a response such as "First, check the agenda and gather the materials," and sends it to the device.

[1631] Step 8:

[1632] Show Answers

[1633] What the device does: The device displays the answer sent by the server to the user.

[1634] Input: The answer sent by the server.

[1635] Data processing: Display the answer to the user visually or audibly.

[1636] Output: The answer displayed to the user.

[1637] What it does: The device prompts the employee, "To prepare for your next meeting, first review the agenda and gather any relevant materials."

[1638] Step 9:

[1639] Feedback and model improvement

[1640] User action: The user enters feedback on the provided answer. The device sends the feedback to the server.

[1641] Input: User feedback.

[1642] Data processing: Analyze feedback and use it to improve the model.

[1643] Output: An improved natural language model.

[1644] Specific behavior: The user enters feedback such as "This answer was helpful," and the device sends the feedback to the server. The server analyzes the feedback and recalibrates the model.

[1645] Step 10:

[1646] Data anonymization

[1647] What the device does: Anonymizes collected data to protect privacy.

[1648] Input: Conversation data, including personal information.

[1649] Data processing: removing or transforming personally identifiable information.

[1650] Output: Anonymized data.

[1651] What it does: It removes and securely stores your name and unique identifiers.

[1652] (Application example 2)

[1653] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1654] When it comes to communication between workers and robots in factories, there is a lack of a way to understand the emotions of workers and respond appropriately. This can have a negative impact on productivity and safety, especially in emergencies or stressful situations. To solve this problem, a system is needed that can analyze the emotional state of workers and respond appropriately according to that state.

[1655] The identification process by the identification 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 online conversation data from users, means for analyzing the collected data to extract question and answer pairs, means for fine-tuning a natural language model using the extracted question and answer pairs, means for analyzing the emotional state of the user using an emotion engine based on the collected data, means for generating an adapted answer based on the emotional state, and means for displaying the generated answer to the user. This makes it possible to understand the emotional state of employees and respond appropriately.

[1656] "User" refers to the factory employees who use the system.

[1657] "Online conversation data" refers to one-to-one voice or text interactions between employees and robots.

[1658] "Means of collection" refers to hardware and software that records or records online conversations between employees and robots.

[1659] "Means for analyzing and extracting question and answer pairs" refers to technology that uses natural language processing technology to analyze collected online conversation data and identify important questions and their answers.

[1660] "Fine-tuning" refers to the process of optimizing the performance of an existing natural language model using extracted question and answer pairs.

[1661] "Emotion engine" refers to software used to analyze users' emotions from online conversation data and identify their emotional states.

[1662] "Means for analyzing emotional state" refers to the process for determining a user's emotions from collected data using an emotion engine.

[1663] "Means for generating adapted answers" refers to techniques for generating appropriate and optimal answers based on the user's emotional state.

[1664] "Means for displaying" refers to a display or audio output device for visually or audibly communicating the generated response to the employee.

[1665] A system for implementing this invention allows a user to provide online conversation data, analyzes the data, and combines an emotion engine that recognizes the user's emotions with a natural language model to take the user's emotional state into account when generating responses. This system is particularly applicable to factory robots and is designed to facilitate smooth communication between factory workers and robots.

[1666] Hardware and Software Configuration

[1667] Hardware: Robots, microphones, and servers located in the factory

[1668] Software: Natural language processing engines (e.g., Transformer's NLP), sentiment analysis engines (e.g., Sentiment Analysis Pipeline), generative AI models (e.g., GPT-3)

[1669] Specific explanation of the system's operation

[1670] Collection and transmission of online conversation data

[1671] The robot collects online conversations with employees in real time via a microphone, and the collected voice data is encrypted and securely sent to a server.

[1672] Data analysis and text conversion

[1673] The server analyzes the received voice data and converts it into text using natural language processing (NLP) techniques, extracting important keywords and question / answer pairs from the converted text data.

[1674] Emotion analysis using an emotion engine

[1675] Next, the server analyzes the user's emotional state using an emotion engine and assigns emotion labels to the conversation data, thereby understanding the user's emotional state.

[1676] Training the model

[1677] The extracted question-answer pairs and sentiment labels are used to fine-tune an existing natural language model (e.g., GPT-3), learning user-specific answer patterns along with their emotional states in the process.

[1678] Responding to questions

[1679] When an employee types a question through the system's interface, the robot sends it to the server, which uses a fine-tuned natural language model and emotion engine to generate the best answer and communicate it to the user via a display or voice output device.

[1680] Specific examples

[1681] For example, if an employee in the quality control department asks the robot, "I don't know how to inspect the product," the robot's emotion analysis engine will recognize that the employee is feeling impatient. The generative AI model will respond by offering guidance that is adapted to the employee's impatient state: "Please stay calm. Let's inspect the product by following these steps."

[1682] Prompt Sentence Examples

[1683] "If an employee notices something is wrong but is feeling anxious, how do I support them?"

[1684] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1685] Step 1:

[1686] The terminal collects online conversation data. Specifically, the factory robot records the conversation with the employee in real time through a microphone. The collected data is in the form of audio, and the audio data is temporarily stored on the terminal for further processing.

[1687] Step 2:

[1688] The device encrypts the collected voice data and securely transmits it to a server using a secure communication protocol (e.g., HTTPS). The server then stores the received voice data in a database with access restrictions to ensure privacy.

[1689] Step 3:

[1690] The server analyzes the received voice data and converts it into text using natural language processing (NLP) techniques, such as using Transformers NLP models to convert speech to text, and stores the resulting text for use in the next step.

[1691] Step 4:

[1692] The server extracts important keywords and question / answer pairs from the converted text data. It uses natural language processing technology to identify which parts are questions and which parts are answers and extracts those pairs. This allows it to understand what information the user is looking for.

[1693] Step 5:

[1694] The server uses an emotion engine to analyze the user's emotional state based on the collected text data. An emotion analysis engine (e.g., Sentiment Analysis Pipeline) is used to identify emotions such as "anxiety" or "impatience" from the user's comments. Based on the results of this analysis, an emotion label is assigned to the text data.

[1695] Step 6:

[1696] The server uses the extracted question and answer pairs and sentiment labels to fine-tune existing natural language models, such as using GPT-3 to learn employee speech patterns and their corresponding emotional states. This process produces a model that understands the user's unique communication style.

[1697] Step 7:

[1698] The user inputs a question through the system interface. The factory robot recognizes this question and sends it to the server. The question is then analyzed in the next step.

[1699] Step 8:

[1700] The server analyzes the user's question and generates the optimal answer using a fine-tuned natural language model and emotion engine. For example, if the user is in an "anxious" emotional state, the server will provide a calm tone and specific advice that is adapted to that state. The generated answer is then sent to the device.

[1701] Step 9:

[1702] The terminal displays the response sent from the server to the user, and the factory robot communicates the response to the employee via voice, allowing the user to take appropriate action.

[1703] Step 10:

[1704] The user provides feedback on the answers provided by the device, which collects this feedback and sends it to the server, which analyzes it and uses the results to retune the model and improve the system's performance.

[1705] Step 11:

[1706] The device anonymizes the data collected to protect your privacy, ensuring that personally identifiable information is removed and stored securely.

[1707] Example prompt sentence:

[1708] "If an employee notices something is wrong but is feeling anxious, how do I support them?"

[1709] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1710] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1711] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1712] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1713] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1714] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1715] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1716] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1717] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1718] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1719] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1720] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1721] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1722] 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.

[1723] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1724] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1725] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1726] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1727] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1728] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1729] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1730] The following is further disclosed regarding the above embodiment.

[1731] (Claim 1)

[1732] means for collecting online conversation data from users;

[1733] a means for analyzing the collected data to extract question and answer pairs;

[1734] means for fine-tuning a natural language model using the extracted question and answer pairs;

[1735] A means for generating answers to questions from users using a fine-tuned natural language model;

[1736] means for displaying the generated answers to the user;

[1737] A system including:

[1738] (Claim 2)

[1739] 10. The system of claim 1, further comprising means for collecting user feedback and continuously improving the fine-tuned natural language model based on the feedback.

[1740] (Claim 3)

[1741] 10. The system of claim 1, further comprising means for anonymizing data to provide privacy protection for online conversation data from users.

[1742] "Example 1"

[1743] (Claim 1)

[1744] means for collecting online communication data from a user;

[1745] A means for encrypting and securely transmitting collected data to a server;

[1746] means for parsing the received data and converting it into text using natural language processing techniques;

[1747] A means for extracting important keywords and question and answer pairs from the converted text data;

[1748] a means for fine-tuning an existing generative AI model using the extracted question and answer pairs; and

[1749] A means of accepting user questions and generating optimal answers using a fine-tuned generative AI model; and

[1750] means for displaying the generated answers to the user;

[1751] A system including:

[1752] (Claim 2)

[1753] 10. The system of claim 1, further comprising means for collecting user feedback, analyzing the feedback, and continually improving the fine-tuned generative AI model based on the results.

[1754] (Claim 3)

[1755] 10. The system of claim 1, further comprising means for anonymizing data to provide privacy protection for online communication data from a user.

[1756] "Application Example 1"

[1757] (Claim 1)

[1758] means for collecting online conversation data from users;

[1759] a means for analyzing the collected data to extract question and answer pairs;

[1760] means for fine-tuning a natural language model using the extracted question and answer pairs;

[1761] A means for generating answers to questions from users using a fine-tuned natural language model;

[1762] means for displaying the generated answers to the user;

[1763] A means of collecting online conversation data in real time and analyzing security risks;

[1764] a means for notifying a user of a risk alert based on the analysis results;

[1765] A system including:

[1766] (Claim 2)

[1767] 10. The system of claim 1, further comprising means for collecting user feedback and continuously improving the fine-tuned natural language model based on the feedback.

[1768] (Claim 3)

[1769] 10. The system of claim 1, further comprising means for anonymizing data to provide privacy protection for online conversation data from users.

[1770] "Example 2: Combining Emotion Engines"

[1771] (Claim 1)

[1772] terminal means for collecting online conversation data from users;

[1773] terminal means for encrypting the collected data and transmitting it to a server in a secure manner;

[1774] a database means for storing the received data by the server;

[1775] a means for analyzing the data received by the server using natural language processing technology and extracting question and answer pairs;

[1776] A means for the server to analyze the user's emotional state from the data using an emotion analysis engine and assign an emotion label;

[1777] a server means for fine-tuning a natural language model using the extracted question and answer pairs and sentiment labels;

[1778] a terminal means for accepting questions from users;

[1779] a means for the server to generate answers to the user's questions using a fine-tuned natural language model and sentiment analysis engine;

[1780] a terminal means for displaying the generated answer to the user;

[1781] A system including:

[1782] (Claim 2)

[1783] 10. The system of claim 1, further comprising server means for collecting user feedback and continuously improving the fine-tuned natural language model based on the feedback.

[1784] (Claim 3)

[1785] 10. The system of claim 1, further comprising terminal means for anonymizing data to provide privacy protection for online conversation data from users.

[1786] "Application example 2 when combining emotion engines"

[1787] (Claim 1)

[1788] means for collecting online conversation data from users;

[1789] a means for analyzing the collected data to extract question and answer pairs;

[1790] means for fine-tuning a natural language model using the extracted question and answer pairs;

[1791] means for analyzing the user's emotional state using an emotion engine based on the collected data;

[1792] means for generating adapted responses based on emotional state;

[1793] means for displaying the generated answers to the user;

[1794] A system including:

[1795] (Claim 2)

[1796] 10. The system of claim 1, further comprising means for collecting user feedback and continuously improving the fine-tuned natural language model based on the feedback.

[1797] (Claim 3)

[1798] 10. The system of claim 1, further comprising means for anonymizing data to provide privacy protection for online conversation data from users. [Explanation of symbols]

[1799] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. means for collecting online conversation data from users; a means for analyzing the collected data to extract question and answer pairs; means for fine-tuning a natural language model using the extracted question and answer pairs; A means for generating answers to questions from users using a fine-tuned natural language model; means for displaying the generated answers to the user; A system including:

2. The system of claim 1 , further comprising means for collecting user feedback and continuously improving the fine-tuned natural language model based on the feedback.

3. The system of claim 1 further comprising means for anonymizing data to provide privacy protection for online conversation data from users.

Images