System

Abstract

A system is provided.SOLUTION: A system comprising: means for collecting meeting minutes data; means for training an artificial intelligence model using the meeting minutes data; and means for generating a response to a question from a user using the trained artificial intelligence model.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] Office workers often have to sift through a huge amount of meeting minutes when they want to find out the details of past meetings, rules, or the background of a project. This makes it difficult to obtain information efficiently, resulting in a decline in work productivity. Furthermore, the time it takes to search for people who might know the information makes it difficult to make quick decisions. [Means for solving the problem]

[0005] The present invention resides in a system that includes a means for collecting minutes data, a means for training an artificial intelligence model using the minutes data, and a means for generating responses to questions from users using the trained artificial intelligence model. This allows users to quickly and accurately obtain detailed information about past minutes and meetings, improving work efficiency and enabling quick decision-making.

[0006] "Minutes data" is text-format information that records the contents of meetings and discussions.

[0007] A "collecting means" is a method or device for obtaining and collecting minutes data from a particular directory or database.

[0008] An "artificial intelligence model" is a machine learning algorithm or system that can learn patterns from large amounts of data and then generate appropriate responses to new data.

[0009] A "training means" is a method or device for inputting meeting minutes data into an artificial intelligence model and assisting the model in the process of learning patterns and rules from the data.

[0010] A "trained artificial intelligence model" is an artificial intelligence model that has already been trained for a specific task and is capable of generating appropriate responses.

[0011] A "user" is a person or system user who wishes to retrieve information from the minutes data.

[0012] A "means for generating a response to a question" is a method or device that uses a trained artificial intelligence model to generate an appropriate response based on a query provided by a user. [Brief explanation of the drawings]

[0013] [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

[0014] 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.

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

[0016] 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).

[0017] 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.

[0018] 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.

[0019] 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.

[0020] 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."

[0021] [First embodiment]

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

[0023] 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.

[0024] 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).

[0025] 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.

[0026] 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.

[0027] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form 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.

[0028] 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.

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

[0030] 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.

[0031] 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.

[0032] 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.

[0033] 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."

[0034] This system collects data from company meeting minutes and generates responses to user questions using an artificial intelligence model. This system is composed of three main components: a server, a terminal, and a user.

[0035] 1. Collection of meeting minutes data

[0036] The server has the ability to scan specific directories and databases where the company's meeting minutes data is stored and collect all meeting minutes files, thereby consolidating scattered meeting minutes data in one place.

[0037] Specifically, the server reads all text files in a specified directory and stores their contents in a list or array, which is later used to train an artificial intelligence model.

[0038] 2. Training the AI ​​model

[0039] The server uses the collected meeting minutes data to train an AI model. Specifically, it combines the text-based meeting minutes data into one large text file and uses it to train a generative model such as GPT-2.

[0040] It creates a training text dataset, and after training, saves the trained model and tokenizer, which are then used to generate answers to user questions.

[0041] 3. Initializing the chat bot using the trained model

[0042] The server loads the trained AI model and tokenizer and initializes a new chat bot instance, which is responsible for generating responses to user queries.

[0043] 4. User Q&A

[0044] The user inputs a question to the chat bot using a terminal. This input is sent to the server via the terminal. The chat bot on the server uses a trained artificial intelligence model to generate an appropriate response to the question. The generated response is then sent back to the terminal and displayed to the user.

[0045] As a concrete example, consider a case where a user asks, "Please tell me about the progress of Project X last year." This question is sent from the device to a chat bot, which generates a response based on the collected meeting minutes data. The bot then provides the user with an answer such as, "Project X started in April of last year and achieved major milestones in June. The main challenges were A and B, and C and D were proposed as solutions for each."

[0046] In this way, the present invention helps office workers to efficiently search through minutes data and quickly obtain the information they need.

[0047] The processing flow will be explained below.

[0048] Step 1:

[0049] The server scans specific directories and databases where the company's meeting minutes data is stored and collects all meeting minutes files. At that time, it reads all text files in the specified directory and stores their contents in a list or array. This aggregates scattered meeting minutes data in one place.

[0050] Step 2:

[0051] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model. Specifically, it combines the individual meeting minutes texts and integrates them into one large text file. This single text file is then used in the subsequent learning process.

[0052] Step 3:

[0053] The server trains an artificial intelligence model using a generative model (e.g., GPT-2). It loads the generative model and tokenizer, creates a training dataset, and trains the model based on the configured parameters. After training is complete, it saves the trained model and tokenizer.

[0054] Step 4:

[0055] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which is responsible for generating responses to user queries.

[0056] Step 5:

[0057] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[0058] Step 6:

[0059] The chat bot on the server processes the received query and generates an appropriate response using an artificial intelligence model. The query is encoded with a tokenizer, and the generated tokens are input into the model to generate a predicted response.

[0060] Step 7:

[0061] The server uses a tokenizer to decode the generated response and convert it into a natural language response, such as "Project X was launched in April of last year and achieved a major milestone in June."

[0062] Step 8:

[0063] The terminal displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for.

[0064] Example 1

[0065] 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."

[0066] Conventional meeting minutes data management systems make it difficult to efficiently find the necessary information from the vast amount of data. Manual search and information extraction are time-consuming, hindering productivity. Furthermore, the lack of advanced technology to generate appropriate responses makes it difficult for users to quickly obtain useful information.

[0067] 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.

[0068] In this invention, the server includes a means for scanning specific company directories and databases to collect meeting minutes data, a means for integrating the collected meeting minutes data into a single large text file and training a generative AI model using the text, and a means for loading the trained model and tokenizer and initializing a chat bot instance, thereby enabling the server to generate prompt and appropriate responses to user questions.

[0069] A "server" is a computer system that provides and processes information via a communication network.

[0070] A "directory" is a hierarchical unit that lists files and folders in a computer.

[0071] A "database" is a system that systematically stores and manages large amounts of data, allowing it to be searched and retrieved quickly.

[0072] "Minutes data" is document data that records the contents of meetings and discussions.

[0073] A "text file" is a file format for storing text data, and is generally saved as plain text.

[0074] A "generative AI model" is an artificial intelligence model that generates new text or information based on input data.

[0075] A "trained model" is an artificial intelligence model that has been fully trained using a specific dataset.

[0076] A "tokenizer" is a tool that divides text data into small units (tokens) such as words and phrases.

[0077] A "chat bot instance" is an instance of a chat bot that has been initialized to simulate a conversation with a user.

[0078] "User" means an individual or organization that uses the system.

[0079] This invention is a system that collects internal meeting minutes data and generates responses to user questions using a generative AI model. The system mainly consists of three main components: a server, a terminal, and a user.

[0080] Hardware and Software Configuration

[0081] server

[0082] The server is responsible for data collection, AI model training, and chat bot initialization and operation. Specifically, it uses programming languages ​​and libraries such as Python, TensorFlow, and PyTorch. The server also operates chat bot instances using web frameworks such as Flask and Django. MySQL or PostgreSQL are commonly used as databases.

[0083] Terminal

[0084] The device is a computer, tablet, smartphone, etc. that provides an interface for users to enter questions. The device is connected to the server via a web browser. The user interface is developed using HTML, CSS, JavaScript, etc.

[0085] User

[0086] A user is an individual or organization that uses the system to enter questions and receive responses. Users access the system through a terminal and obtain the information they need.

[0087] Data processing and calculation

[0088] The server scans designated directories and databases to collect meeting minutes data stored within the company. The collected meeting minutes data is consolidated into one large text file. This consolidated text file is then used to train a generative AI model, such as GPT-2. The trained model is then used to generate responses to user questions.

[0089] The terminal transmits the user's input to the server and displays the responses received from the server.

[0090] For example, if a user asks, "Please tell me about the progress of Project X last year," this input is sent from the terminal to the server. The chat bot on the server generates a response based on the collected meeting minutes data and provides the user with an answer such as, "Project X started in April of last year and achieved a major milestone in June. The main challenges were A and B, and C and D were proposed as solutions for each."

[0091] Prompt Sentence Examples

[0092] Below are some example prompts to input to a generative AI model:

[0093] Search for "Progress of Project X last year" in your company's meeting minutes data.

[0094] In this way, the present invention helps office workers to efficiently search through minutes data and quickly obtain the information they need.

[0095] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0096] Step 1:

[0097] The server scans specific directories and databases within the company to collect meeting minutes data. During this process, the server searches for all text and PDF files in the specified directory and retrieves their contents. The collected files are saved in temporary storage on the server. For example, files such as "meeting_20230310.txt" and "strategy_session.pdf" are retrieved from the " / corporate / minutes / " directory. The input for this step is the directory path, and the output is a list of the collected meeting minutes files.

[0098] Step 2:

[0099] The server combines the collected minutes files into one large text file. It reads each file and concatenates their contents into a single text file called "combined_minutes.txt." For example, it combines the text content extracted from "meeting_20230310.txt" and "strategy_session.pdf." The input for this step is a list of minutes files, and the output is a combined text file.

[0100] Step 3:

[0101] The server uses the combined text files to train a generative AI model (e.g., GPT-2). It trains the model using Python scripts and libraries (TensorFlow, PyTorch). During training, it uses the text data "combined_minutes.txt" as input and generates a trained model and tokenizer "trained_gpt2_model.pt" and "tokenizer.json" as outputs. Specifically, it uses a GPU to process each batch of data and optimize the model parameters.

[0102] Step 4:

[0103] The server loads the trained model and tokenizer and initializes a new chat bot instance. These are loaded on the server using a web framework such as Flask or Django. For example, start a Flask application, load "trained_gpt2_model.pt" and "tokenizer.json", and create a chat bot instance. The input of this step is the trained model and tokenizer, and the output is an initialized chat bot instance.

[0104] Step 5:

[0105] A user uses a terminal to input a question to the chat bot. For example, using a web browser, the user inputs a question such as "How did project X progress last year?" This input is sent to the server via the terminal. The input of this step is the user's question, and the output is a request to send to the server.

[0106] Step 6:

[0107] The chat bot on the server receives the user's question and generates an appropriate response using the trained model. As a specific example, the model generates text based on "Progress of Project X last year" and outputs the following response: "Project X started in April of last year and achieved major milestones in June. The main challenges were A and B, and C and D were proposed as solutions for each." The input of this step is the user's question, and the output is the generated response.

[0108] Step 7:

[0109] The server sends the generated response back to the device. The server sends the response text as an HTTP response. For example, it packs the generated response in JSON format and sends it back to the browser. The input of this step is the generated response, and the output is the response to the device.

[0110] Step 8:

[0111] The terminal receives the response from the server and displays it to the user. A concrete example is a web browser, which displays the received text on the screen for the user to view. The input to this step is the response from the server, and the output is what is displayed to the user.

[0112] (Application example 1)

[0113] 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."

[0114] With conventional methods for managing meeting minutes data, it was difficult to efficiently integrate data scattered across multiple locations, search it, and quickly obtain the necessary information. Furthermore, there was no environment in place for collecting work reports and meeting minutes data within the factory in real time and allowing on-site staff to access it immediately. This resulted in problems such as reduced work efficiency and delays in troubleshooting.

[0115] 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.

[0116] In this invention, the server includes means for collecting minutes data, means for training an artificial intelligence model using the minutes data, means for generating responses to questions from users using the trained artificial intelligence model, means for the user to input the questions using a terminal device and receive the responses, and means for automatically collecting the minutes data using machines in a factory and uploading it to the server. This enables efficient collection and integration of scattered minutes data, allowing users to quickly obtain the information they need in real time.

[0117] "Minutes data" is a text document that records the contents of meetings, work reports, and the like.

[0118] A "collection means" is a system component that scans a specific directory or database and aggregates the minutes data in one place.

[0119] The "artificial intelligence model" is a machine learning technology that learns from collected meeting minutes data and generates appropriate responses to questions from users.

[0120] "Training methods" are the processes and techniques used to train artificial intelligence models using collected meeting minutes data.

[0121] The "means for generating a response" is a system component that uses a trained artificial intelligence model to generate an appropriate answer to a question entered by a user.

[0122] A "terminal device" is a hardware device (e.g., smartphone, tablet, PC) through which a user accesses the system, enters questions, and receives responses.

[0123] The "machines in the factory" refer to devices and robots installed in the factory, which are devices that automatically collect meeting minutes data and upload it to the server.

[0124] A "server" is a computer system that centrally manages the collection of minutes data, learning of artificial intelligence models, and response generation.

[0125] The present invention is a system that collects meeting minutes data and generates responses to user questions using an artificial intelligence model. This system is mainly composed of three elements: a server, a terminal, and a user.

[0126] 1. Collection of meeting minutes data

[0127] The server has the function of scanning specific directories and databases and aggregating meeting minutes data automatically collected from machines in the factory into one place. For example, every time a meeting or work report is held in the factory, the data is uploaded to the server. This process allows scattered meeting minutes data to be collected in a centralized manner.

[0128] 2. Training the AI ​​model

[0129] The server uses the collected meeting minutes data to train an artificial intelligence model. Specifically, all meeting minutes data is integrated into one large text file, and a generative AI model such as GPT-2 is trained based on this. A training text dataset is created, and after learning, the trained model and tokenizer are saved. This trained model is used to generate responses to user questions.

[0130] 3. Initializing the chat bot using the trained model

[0131] The server loads the trained AI model and tokenizer and initializes a new chat bot instance, which is responsible for generating responses to user queries.

[0132] 4. User Q&A

[0133] A user uses a device (such as a smartphone or tablet) to input a question to the chat bot. The input question is sent to the server via the device. For example, a user might ask, "Please tell me the details of yesterday's troubleshooting." This question is sent from the device to the chat bot, and the chat bot generates a response based on the collected meeting minutes data. The user is provided with a response such as, "As for the details of yesterday's troubleshooting, a problem occurred with device A, and a software update was implemented as a solution."

[0134] To realize this system, the server uses the following hardware and software: The hardware includes the server device, in-factory collection equipment, and user terminals. The software includes a generative AI model such as GPT-2, a directory scanning program, a tokenizer, and a chat bot instance.

[0135] As a concrete example, consider the case where a factory engineer asks, "Please tell me what my boss said in the meeting yesterday." Examples of prompt sentences are as follows:

[0136] Prompt: "Tell me what your boss said in the meeting yesterday."

[0137] Based on this prompt, the trained model references detailed meeting minutes data to generate what the supervisor said in the meeting. For example, it can provide a response such as, "The supervisor presented a new schedule for equipment maintenance and instructed that it be carried out within one week." This system enables users to quickly and accurately obtain the information they need.

[0138] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0139] Step 1: Collecting meeting minutes data

[0140] The server stores meeting minutes data automatically collected from machines in the factory in a specific directory. The machines in the factory generate work reports and meeting minutes data, which are then uploaded to the server in real time. The input is the meeting minutes data, and the output is data integrated into a directory or database. Specifically, the server scans the directory to detect new files and centrally manages their contents.

[0141] Step 2: Integrating and learning from meeting minutes data

[0142] The server combines the collected meeting minutes data into one large text file and uses it to train an AI model. Specifically, it uses the combined text data to train the GPT-2 model. The input is the combined text data, and the output is a trained AI model and tokenizer. The server tokenizes the text data, feeds it to the model, and saves the learning results.

[0143] Step 3: Initialize the chat bot

[0144] The server loads the trained artificial intelligence model and tokenizer and initializes a new chat bot instance. The input is the trained model and tokenizer, and the output is the initialized chat bot instance. Specifically, the server loads the saved model and tokenizer into memory and prepares the dialogue system.

[0145] Step 4: Enter user questions

[0146] A user uses a device (such as a smartphone or tablet) to input a question to the chat bot. This input is sent to the server through the device. The input is the user's question, and the output is a question request to the server. Specifically, the user types a question through the UI, which is then sent to the server via an HTTP request or similar.

[0147] Step 5: Question processing and response generation

[0148] The server generates an appropriate response based on the user's question using a trained artificial intelligence model. The input is the user's question, and the output is the generated response. The server tokenizes the question, inputs it into the model, and decodes and formats the generated text.

[0149] Step 6: View the response

[0150] The terminal receives the response sent from the server and displays it to the user. The input is the response from the server, and the output is the text displayed to the user. Specifically, the terminal parses the response from the server and displays it on the screen.

[0151] This series of steps allows users to quickly and accurately obtain the information they need in real time, for example, by providing an appropriate response to a prompt such as "Please tell me what my boss said in the meeting yesterday."

[0152] 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.

[0153] This invention combines a system that collects data from company meeting minutes and generates responses to user questions using an artificial intelligence model with an emotion engine that recognizes user emotions. This system is mainly composed of three main components: a server, a terminal, and a user.

[0154] 1. Collection of meeting minutes data

[0155] The server scans specific directories and databases where the company's meeting minutes data is stored and collects all meeting minutes files. At that time, it reads all text files in the specified directory and stores their contents in a list or array. This aggregates scattered meeting minutes data in one place.

[0156] 2. Training the AI ​​model

[0157] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model, combining the individual meeting minutes texts and merging them into one large text file, which is then used in the subsequent training process.

[0158] Next, train an artificial intelligence model using a generative model (e.g., GPT-2). Load the generative model and tokenizer, create a training dataset, and train the model based on the set parameters. After training is complete, save the trained model and tokenizer.

[0159] 3. Initializing the chat bot using the trained model

[0160] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which is responsible for generating responses to user queries.

[0161] 4. User Q&A

[0162] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[0163] 5. Use of Emotion Engine

[0164] The server passes the query received from the user to the emotion engine, which analyzes the user's emotion. The emotion engine analyzes the query and recognizes the emotion the user is feeling (e.g., joy, sadness, anger, surprise, etc.).

[0165] 6. Emotion-based response generation

[0166] The chat bot on the server uses the trained artificial intelligence model to generate responses to user queries, taking into account the emotions identified by the emotion engine. For example, if the user is angry, a more careful and polite response will be generated.

[0167] 7. Displaying the Response

[0168] The server uses a tokenizer to decode the generated response and convert it into a natural language response, such as "Project X was launched in April of last year and achieved a major milestone in June."

[0169] 8. Providing a Response

[0170] The device displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for. Furthermore, the response provided takes into consideration the user's feelings, improving the user experience.

[0171] In this way, the present invention helps office-based business people to efficiently search through minutes data and quickly obtain the information they need, while providing responses that take into consideration the user's feelings.

[0172] The processing flow will be explained below.

[0173] Step 1:

[0174] The server scans a specific directory or database where the company's meeting minutes data is stored and collects all meeting minutes files, which involves reading all text files in the directory and storing their contents in a list or array.

[0175] Step 2:

[0176] The server consolidates the collected meeting minutes data into a single text file, sequentially reads each meeting minute, and combines them into a single large text file, which is then used to train the AI ​​model.

[0177] Step 3:

[0178] The server loads a generative model (e.g., GPT-2) and a tokenizer, creates a training dataset, trains the model using training text files, and saves the trained model and tokenizer, ready for subsequent query-answer processing.

[0179] Step 4:

[0180] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which responds to user questions.

[0181] Step 5:

[0182] The user uses the terminal to input a question to the chat bot, for example, "Please tell me how project X progressed last year." This input is then sent from the terminal to the server.

[0183] Step 6:

[0184] Before parsing the received query, the server passes it to an emotion engine to analyze the user's emotions. The emotion engine identifies the user's emotions (e.g., joy, anger, sadness, surprise, etc.) based on the text input.

[0185] Step 7:

[0186] The chat bot on the server uses the trained AI model to generate responses to queries, taking into account the emotions identified by the emotion engine. For example, if the user expresses anger, the response will be more careful and polite.

[0187] Step 8:

[0188] The server uses a tokenizer to decode the generated response and convert it into a natural language response, for example, generating text like "Project X started in April of last year and achieved a major milestone in June."

[0189] Step 9:

[0190] The device displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for. Emotionally sensitive responses improve the user experience.

[0191] Example 2

[0192] 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."

[0193] Conventional information search systems that use internal company meeting minutes data have the problem that they respond to user questions in a formulaic manner and are unable to provide flexible responses that take the user's emotions and situation into consideration, resulting in a poor user experience.

[0194] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes means for collecting minutes data, means for training an artificial intelligence model using the minutes data, means for initializing a chat interface using the trained artificial intelligence model and a tokenizer, means for analyzing emotions in response to a user query using an emotion engine, means for generating a response taking emotions into consideration, and means for displaying the generated response on a user terminal. This makes it possible to provide a response that takes the user's emotions into consideration.

[0195] "Minutes data" refers to a text file or document that records the contents of a conference or meeting.

[0196] An "artificial intelligence model" is an algorithm or system that uses data to learn and generate responses to user queries.

[0197] A "tokenizer" is a tool for converting text data into a format that can be processed by a model.

[0198] A "chat interface" is an interactive user interface that allows users to enter questions and receive responses.

[0199] An "emotion engine" is a system or software that analyzes emotions from a user's text and labels it based on those emotions.

[0200] A "query" is a question or request that a user enters into a system.

[0201] A "user terminal" is a device that a user directly operates and interacts with the system. For example, a PC or smartphone is an example.

[0202] This invention is a system that collects meeting minutes data and generates responses to user questions using an artificial intelligence model. The system also incorporates an emotion engine that recognizes user emotions. The system mainly consists of three main components: a server, a terminal, and a user.

[0203] Use of collected meeting transcript data

[0204] The server scans specific directories and databases within the company to collect all meeting minutes files. It reads all text files in the specified directory and stores their contents in a list or array. For example, it scans the directory / data / meeting_minutes and collects text files. This aggregates scattered meeting minutes data into one place.

[0205] The learning process of artificial intelligence models

[0206] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model. It combines the individual meeting minutes texts and merges them into one large text file. It processes this merged text file using a tokenizer to create the training dataset for the model. The generative model used is GPT-2, and it loads this model and tokenizer and trains the model based on the set parameters. After training is complete, it saves the trained model and tokenizer.

[0207] Chat BOT initialization

[0208] The server initializes the chat bot using the saved trained model and tokenizer. It loads the model and tokenizer into memory and creates a new chat bot instance, which is responsible for generating responses to user queries.

[0209] User questions and answers

[0210] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[0211] Using the Emotion Engine

[0212] The server passes the query received from the user to the emotion engine, which analyzes the user's emotion. The emotion engine analyzes the query and recognizes the emotion the user is feeling (e.g., joy, sadness, anger, surprise, etc.).

[0213] Emotion-based response generation

[0214] The chat bot on the server uses the trained artificial intelligence model to generate responses to user queries, taking into account the emotions identified by the emotion engine. For example, if the user is angry, a more careful and polite response will be generated.

[0215] Viewing and Providing Responses

[0216] The server uses a tokenizer to decode the generated response and convert it into a natural language response. For example, the server might generate an answer such as, "Project X was launched in April of last year and achieved a major milestone in June." The device then displays this generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for.

[0217] In this way, the present invention helps office-based business people to efficiently search through minutes data and quickly obtain the information they need, while providing responses that take into consideration the user's feelings.

[0218] Specific examples

[0219] Example question: "How did project X go last year?"

[0220] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0221] Step 1:

[0222] The server scans a specific directory or database where meeting minutes data is stored and collects all meeting minutes files. As input, it receives a directory path (e.g., / data / meeting_minutes) and stores all text files in the directory in a list or array. Specifically, it uses the Python os module to detect text files in the directory, reads them, and adds them to a list. As output, it obtains a list of the paths to the text files.

[0223] Step 2:

[0224] The server combines the collected minutes data into a single large text file to create a training dataset for the AI ​​model. As input, it receives the path list of the minutes files obtained in step 1 and combines them into a single text file. Specifically, it opens each text file, concatenates its contents into a single string, and writes it to a new file. As output, it saves the combined training dataset as a single text file.

[0225] Step 3:

[0226] The server uses the integrated text file to train an artificial intelligence model (generative model). As input, it receives the integrated text file and training parameters (e.g., number of epochs, batch size, learning rate). Specifically, it uses Hugging Face's Transformers library to load a tokenizer and generative model (e.g., GPT-2), tokenizes the text data with the tokenizer, and trains the model. As output, it saves the trained model and tokenizer.

[0227] Step 4:

[0228] The server initializes the chat bot using the saved trained model and tokenizer. As input, it receives the trained model and tokenizer obtained in step 3 and loads them into memory. Specifically, it loads the model and tokenizer using Hugging Face's Transformers library and creates a new chat bot instance. As output, it obtains the initialized chat bot instance.

[0229] Step 5:

[0230] The user uses the terminal to input a question to the chat bot and send it. As input, the bot receives the query that the user input into the terminal (e.g., "Please tell me the progress of project X last year."). As a specific operation, the bot sends the user's input to the server via the send button or input completion event. As output, the user's query is sent to the server.

[0231] Step 6:

[0232] The server passes the received query to the emotion engine and analyzes the user's emotion. The server receives the user's query as input and passes it to the emotion engine. Specifically, it uses an emotion analysis library (e.g., TextBlob) to analyze the emotion of the query and obtains an emotion label (e.g., joy, sadness, anger, surprise, etc.). The output is the emotion label and the analysis result.

[0233] Step 7:

[0234] The chat bot on the server generates a response to the user's query while taking into account the emotions identified by the emotion engine. As input, it receives the user's query and emotion label. Specifically, it uses the generative AI model to create a prompt sentence based on the emotion label and generate a response. As output, it obtains the generated response text.

[0235] Step 8:

[0236] The server decodes the generated response using a tokenizer and converts it into a natural language response. As input, it receives the generated response text. Specifically, it uses the generative AI model's tokenizer to decode the response text and converts it into a natural language response. As output, it obtains a natural language response.

[0237] Step 9:

[0238] The terminal displays the generated response to the user. As input, it receives a natural language response and displays it to the user. As a specific operation, it adds the natural language response to an area that displays responses through a user interface. As output, it allows the user to confirm the response.

[0239] (Application example 2)

[0240] 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."

[0241] Physical stores require a means to respond appropriately and quickly to customer questions. Furthermore, responses that do not take into account the user's emotions present a risk of lowering customer satisfaction. Furthermore, responses that take emotions into account require the assistance of a human operator, which is costly and labor-intensive. To address these issues, the present invention aims to provide a system that automates customer service in physical stores, understands emotions, and responds appropriately.

[0242] 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 minutes data, means for training an artificial intelligence model using the minutes data, means for generating responses to questions from users using the trained artificial intelligence model, means for analyzing the user's emotions, and means for adjusting the response based on the emotion analysis results. This enables quick and emotional responses to customer questions in physical stores.

[0243] "Minutes data" is document data that records the contents of meetings and discussions.

[0244] An "artificial intelligence model" is a program that uses machine learning algorithms to learn patterns from data and make inferences and predictions.

[0245] "User" is a concept that refers to a person or organization that uses a system.

[0246] An "emotion engine" is software that analyzes and recognizes emotions from input data such as text and voice.

[0247] A "server" is a part of a computer system that provides services to other computers over a network.

[0248] "Emotion analysis result" is user emotion information output as a result of analysis by the emotion engine.

[0249] The "means for generating a response" refers to the process of using an artificial intelligence model to generate an appropriate answer to a question from a user.

[0250] A "collection method" is a process or method for gathering specific data from a designated location.

[0251] This invention is a system for providing fast and emotionally sensitive responses to customer questions in a brick-and-mortar store. The system includes three main components: a server, a terminal, and a user.

[0252] server

[0253] The server has the following roles:

[0254] 1. Collection of meeting minutes data: The server collects meeting minutes data from the specified directory or database, thereby consolidating scattered meeting minutes data in one place.

[0255] 2. Training the AI ​​model: The server uses the collected meeting minutes data to train a generative model (e.g., GPT-2). The meeting minutes data is combined into one large text file, and this file is used to train the AI ​​model.

[0256] 3. Sentiment Analysis: Pass the user query to the sentiment engine to analyze the sentiment. For example, use the TextBlob library to analyze the sentiment of the text.

[0257] 4. Response generation and adjustment: Based on the accumulated data and sentiment analysis results, a generative model is used to generate responses to user questions and adjust them to take sentiment into account.

[0258] Terminal

[0259] The terminal functions as follows:

[0260] 1. Providing a user interface: An interface is provided for users to input questions. This is typically implemented as a smartphone app.

[0261] 2. Question submission: The system has the function of submitting questions entered by the user to the server, which analyzes the submitted data and generates an appropriate response.

[0262] 3. Display Response: Display the response received from the server to the user.

[0263] User

[0264] The user is a customer who uses this system.

[0265] 1. Entering a question: The user enters a question using a terminal. For example, the user enters a specific question such as "Can I return this product?"

[0266] 2. Receiving a response: The user can check the response displayed on the terminal and get the necessary information immediately.

[0267] Usage example

[0268] Specific use cases include the following scenarios:

[0269] Example 1:

[0270] User Input: "Can I return this item?"

[0271] Emotion analysis result: Neutral

[0272] Generated response: "We're glad to help. This item can be returned within 30 days of purchase."

[0273] Prompt Sentence Examples

[0274] Below are some example prompts to input to a generative AI model:

[0275] A customer asked: "Can I return this item?"

[0276] Sentiment analysis was performed using TextBlob, and the user's sentiment score was 0.0.

[0277] The response generated using the GPT-2 model is shown below:

[0278] "We're glad we could help. This item can be returned within 30 days of purchase."

[0279] In this way, the present invention realizes a system that improves the efficiency of customer service in physical stores and provides quick responses that take emotions into consideration.

[0280] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0281] Step 1:

[0282] The server collects meeting minutes data from a specified directory or database. It receives the path of the target directory as input. The server reads all text files in the directory and stores their contents in a list. This aggregates the scattered meeting minutes data in one place, and outputs one large text data.

[0283] Step 2:

[0284] The server trains an AI model using the collected meeting minutes data. As input, it receives the integrated meeting minutes text data. The server loads a generative model such as GPT-2 and a tokenizer to create a training dataset. It trains the model using this dataset and saves the trained model and tokenizer. As output, it obtains the trained AI model and tokenizer.

[0285] Step 3:

[0286] The server initializes the chat bot using the trained artificial intelligence model and tokenizer. As input, it loads the saved trained model and tokenizer. The server reads them into memory and creates a new chat bot instance. As output, it obtains the initialized chat bot instance.

[0287] Step 4:

[0288] The user inputs a question using a terminal and sends it to the server. As input, a specific question text (e.g., "Can I return this product?") is obtained. The terminal sends this input to the server. As output, the question text sent to the server is obtained.

[0289] Step 5:

[0290] The server passes the question text received from the user to the emotion engine to analyze the sentiment. The server receives the question text as input. The server calculates the sentiment score of the text using an emotion analysis library such as TextBlob. The server receives the sentiment score as output.

[0291] Step 6:

[0292] The server uses a trained artificial intelligence model to generate a response to the question, taking into account the sentiment analysis results. The input is the question text and a sentiment score. The server inputs the question text into the generative model and tailors the response based on the sentiment score. The output is a tailored natural language response.

[0293] Step 7:

[0294] The server sends the generated response to the terminal. As input, it receives a generated natural language response. The server sends this response to the terminal. As output, it receives a natural language response sent to the terminal.

[0295] Step 8:

[0296] The terminal displays the response received from the server to the user. As input, it receives a natural language response sent from the server. The terminal displays this response to the user. As output, it receives a response displayed on the user's terminal screen.

[0297] 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.

[0298] 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.

[0299] 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.

[0300] [Second embodiment]

[0301] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0302] 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.

[0303] 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).

[0304] 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.

[0305] 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.

[0306] 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).

[0307] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0308] 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.

[0309] 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.

[0310] 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.

[0311] In the smart glasses 214, 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.

[0312] 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."

[0313] This system collects data from company meeting minutes and generates responses to user questions using an artificial intelligence model. This system is composed of three main components: a server, a terminal, and a user.

[0314] 1. Collection of meeting minutes data

[0315] The server has the ability to scan specific directories and databases where the company's meeting minutes data is stored and collect all meeting minutes files, thereby consolidating scattered meeting minutes data in one place.

[0316] Specifically, the server reads all text files in a specified directory and stores their contents in a list or array, which is later used to train an artificial intelligence model.

[0317] 2. Training the AI ​​model

[0318] The server uses the collected meeting minutes data to train an AI model. Specifically, it combines the text-based meeting minutes data into one large text file and uses it to train a generative model such as GPT-2.

[0319] It creates a training text dataset, and after training, saves the trained model and tokenizer, which are then used to generate answers to user questions.

[0320] 3. Initializing the chat bot using the trained model

[0321] The server loads the trained AI model and tokenizer and initializes a new chat bot instance, which is responsible for generating responses to user queries.

[0322] 4. User Q&A

[0323] The user inputs a question to the chat bot using a terminal. This input is sent to the server via the terminal. The chat bot on the server uses a trained artificial intelligence model to generate an appropriate response to the question. The generated response is then sent back to the terminal and displayed to the user.

[0324] As a concrete example, consider a case where a user asks, "Please tell me about the progress of Project X last year." This question is sent from the device to a chat bot, which generates a response based on the collected meeting minutes data. The bot then provides the user with an answer such as, "Project X started in April of last year and achieved major milestones in June. The main challenges were A and B, and C and D were proposed as solutions for each."

[0325] In this way, the present invention helps office workers to efficiently search through minutes data and quickly obtain the information they need.

[0326] The processing flow will be explained below.

[0327] Step 1:

[0328] The server scans specific directories and databases where the company's meeting minutes data is stored and collects all meeting minutes files. At that time, it reads all text files in the specified directory and stores their contents in a list or array. This aggregates scattered meeting minutes data in one place.

[0329] Step 2:

[0330] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model. Specifically, it combines the individual meeting minutes texts and integrates them into one large text file. This single text file is then used in the subsequent learning process.

[0331] Step 3:

[0332] The server trains an artificial intelligence model using a generative model (e.g., GPT-2). It loads the generative model and tokenizer, creates a training dataset, and trains the model based on the configured parameters. After training is complete, it saves the trained model and tokenizer.

[0333] Step 4:

[0334] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which is responsible for generating responses to user queries.

[0335] Step 5:

[0336] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[0337] Step 6:

[0338] The chat bot on the server processes the received query and generates an appropriate response using an artificial intelligence model. The query is encoded with a tokenizer, and the generated tokens are input into the model to generate a predicted response.

[0339] Step 7:

[0340] The server uses a tokenizer to decode the generated response and convert it into a natural language response, such as "Project X was launched in April of last year and achieved a major milestone in June."

[0341] Step 8:

[0342] The terminal displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for.

[0343] Example 1

[0344] 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."

[0345] Conventional meeting minutes data management systems make it difficult to efficiently find the necessary information from the vast amount of data. Manual search and information extraction are time-consuming, hindering productivity. Furthermore, the lack of advanced technology to generate appropriate responses makes it difficult for users to quickly obtain useful information.

[0346] 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.

[0347] In this invention, the server includes a means for scanning specific company directories and databases to collect meeting minutes data, a means for integrating the collected meeting minutes data into a single large text file and training a generative AI model using the text, and a means for loading the trained model and tokenizer and initializing a chat bot instance, thereby enabling the server to generate prompt and appropriate responses to user questions.

[0348] A "server" is a computer system that provides and processes information via a communication network.

[0349] A "directory" is a hierarchical unit that lists files and folders in a computer.

[0350] A "database" is a system that systematically stores and manages large amounts of data, allowing it to be searched and retrieved quickly.

[0351] "Minutes data" is document data that records the contents of meetings and discussions.

[0352] A "text file" is a file format for storing text data, and is generally saved as plain text.

[0353] A "generative AI model" is an artificial intelligence model that generates new text or information based on input data.

[0354] A "trained model" is an artificial intelligence model that has been fully trained using a specific dataset.

[0355] A "tokenizer" is a tool that divides text data into small units (tokens) such as words and phrases.

[0356] A "chat bot instance" is an instance of a chat bot that has been initialized to simulate a conversation with a user.

[0357] "User" means an individual or organization that uses the system.

[0358] This invention is a system that collects internal meeting minutes data and generates responses to user questions using a generative AI model. The system mainly consists of three main components: a server, a terminal, and a user.

[0359] Hardware and Software Configuration

[0360] server

[0361] The server is responsible for data collection, AI model training, and chat bot initialization and operation. Specifically, it uses programming languages ​​and libraries such as Python, TensorFlow, and PyTorch. The server also operates chat bot instances using web frameworks such as Flask and Django. MySQL or PostgreSQL are commonly used as databases.

[0362] Terminal

[0363] The device is a computer, tablet, smartphone, etc. that provides an interface for users to enter questions. The device is connected to the server via a web browser. The user interface is developed using HTML, CSS, JavaScript, etc.

[0364] User

[0365] A user is an individual or organization that uses the system to enter questions and receive responses. Users access the system through a terminal and obtain the information they need.

[0366] Data processing and calculation

[0367] The server scans designated directories and databases to collect meeting minutes data stored within the company. The collected meeting minutes data is consolidated into one large text file. This consolidated text file is then used to train a generative AI model, such as GPT-2. The trained model is then used to generate responses to user questions.

[0368] The terminal transmits the user's input to the server and displays the responses received from the server.

[0369] For example, if a user asks, "Please tell me about the progress of Project X last year," this input is sent from the terminal to the server. The chat bot on the server generates a response based on the collected meeting minutes data and provides the user with an answer such as, "Project X started in April of last year and achieved a major milestone in June. The main challenges were A and B, and C and D were proposed as solutions for each."

[0370] Prompt Sentence Examples

[0371] Below are some example prompts to input to a generative AI model:

[0372] Search for "Progress of Project X last year" in your company's meeting minutes data.

[0373] In this way, the present invention helps office workers to efficiently search through minutes data and quickly obtain the information they need.

[0374] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0375] Step 1:

[0376] The server scans specific directories and databases within the company to collect meeting minutes data. During this process, the server searches for all text and PDF files in the specified directory and retrieves their contents. The collected files are saved in temporary storage on the server. For example, files such as "meeting_20230310.txt" and "strategy_session.pdf" are retrieved from the " / corporate / minutes / " directory. The input for this step is the directory path, and the output is a list of the collected meeting minutes files.

[0377] Step 2:

[0378] The server combines the collected minutes files into one large text file. It reads each file and concatenates their contents into a single text file called "combined_minutes.txt." For example, it combines the text content extracted from "meeting_20230310.txt" and "strategy_session.pdf." The input for this step is a list of minutes files, and the output is a combined text file.

[0379] Step 3:

[0380] The server uses the combined text files to train a generative AI model (e.g., GPT-2). It trains the model using Python scripts and libraries (TensorFlow, PyTorch). During training, it uses the text data "combined_minutes.txt" as input and generates a trained model and tokenizer "trained_gpt2_model.pt" and "tokenizer.json" as outputs. Specifically, it uses a GPU to process each batch of data and optimize the model parameters.

[0381] Step 4:

[0382] The server loads the trained model and tokenizer and initializes a new chat bot instance. These are loaded on the server using a web framework such as Flask or Django. For example, start a Flask application, load "trained_gpt2_model.pt" and "tokenizer.json", and create a chat bot instance. The input of this step is the trained model and tokenizer, and the output is an initialized chat bot instance.

[0383] Step 5:

[0384] A user uses a terminal to input a question to the chat bot. For example, using a web browser, the user inputs a question such as "How did project X progress last year?" This input is sent to the server via the terminal. The input of this step is the user's question, and the output is a request to send to the server.

[0385] Step 6:

[0386] The chat bot on the server receives the user's question and generates an appropriate response using the trained model. As a specific example, the model generates text based on "Progress of Project X last year" and outputs the following response: "Project X started in April of last year and achieved major milestones in June. The main challenges were A and B, and C and D were proposed as solutions for each." The input of this step is the user's question, and the output is the generated response.

[0387] Step 7:

[0388] The server sends the generated response back to the device. The server sends the response text as an HTTP response. For example, it packs the generated response in JSON format and sends it back to the browser. The input of this step is the generated response, and the output is the response to the device.

[0389] Step 8:

[0390] The terminal receives the response from the server and displays it to the user. A concrete example is a web browser, which displays the received text on the screen for the user to view. The input to this step is the response from the server, and the output is what is displayed to the user.

[0391] (Application example 1)

[0392] 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."

[0393] With conventional methods for managing meeting minutes data, it was difficult to efficiently integrate data scattered across multiple locations, search it, and quickly obtain the necessary information. Furthermore, there was no environment in place for collecting work reports and meeting minutes data within the factory in real time and allowing on-site staff to access it immediately. This resulted in problems such as reduced work efficiency and delays in troubleshooting.

[0394] 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.

[0395] In this invention, the server includes means for collecting minutes data, means for training an artificial intelligence model using the minutes data, means for generating responses to questions from users using the trained artificial intelligence model, means for the user to input the questions using a terminal device and receive the responses, and means for automatically collecting the minutes data using machines in a factory and uploading it to the server. This enables efficient collection and integration of scattered minutes data, allowing users to quickly obtain the information they need in real time.

[0396] "Minutes data" is a text document that records the contents of meetings, work reports, and the like.

[0397] A "collection means" is a system component that scans a specific directory or database and aggregates the minutes data in one place.

[0398] The "artificial intelligence model" is a machine learning technology that learns from collected meeting minutes data and generates appropriate responses to questions from users.

[0399] "Training methods" are the processes and techniques used to train artificial intelligence models using collected meeting minutes data.

[0400] The "means for generating a response" is a system component that uses a trained artificial intelligence model to generate an appropriate answer to a question entered by a user.

[0401] A "terminal device" is a hardware device (e.g., smartphone, tablet, PC) through which a user accesses the system, enters questions, and receives responses.

[0402] The "machines in the factory" refer to devices and robots installed in the factory, which are devices that automatically collect meeting minutes data and upload it to the server.

[0403] A "server" is a computer system that centrally manages the collection of minutes data, learning of artificial intelligence models, and response generation.

[0404] The present invention is a system that collects meeting minutes data and generates responses to user questions using an artificial intelligence model. This system is mainly composed of three elements: a server, a terminal, and a user.

[0405] 1. Collection of meeting minutes data

[0406] The server has the function of scanning specific directories and databases and aggregating meeting minutes data automatically collected from machines in the factory into one place. For example, every time a meeting or work report is held in the factory, the data is uploaded to the server. This process allows scattered meeting minutes data to be collected in a centralized manner.

[0407] 2. Training the AI ​​model

[0408] The server uses the collected meeting minutes data to train an artificial intelligence model. Specifically, all meeting minutes data is integrated into one large text file, and a generative AI model such as GPT-2 is trained based on this. A training text dataset is created, and after learning, the trained model and tokenizer are saved. This trained model is used to generate responses to user questions.

[0409] 3. Initializing the chat bot using the trained model

[0410] The server loads the trained AI model and tokenizer and initializes a new chat bot instance, which is responsible for generating responses to user queries.

[0411] 4. User Q&A

[0412] A user uses a device (such as a smartphone or tablet) to input a question to the chat bot. The input question is sent to the server via the device. For example, a user might ask, "Please tell me the details of yesterday's troubleshooting." This question is sent from the device to the chat bot, and the chat bot generates a response based on the collected meeting minutes data. The user is provided with a response such as, "As for the details of yesterday's troubleshooting, a problem occurred with device A, and a software update was implemented as a solution."

[0413] To realize this system, the server uses the following hardware and software: The hardware includes the server device, in-factory collection equipment, and user terminals. The software includes a generative AI model such as GPT-2, a directory scanning program, a tokenizer, and a chat bot instance.

[0414] As a concrete example, consider the case where a factory engineer asks, "Please tell me what my boss said in the meeting yesterday." Examples of prompt sentences are as follows:

[0415] Prompt: "Tell me what your boss said in the meeting yesterday."

[0416] Based on this prompt, the trained model references detailed meeting minutes data to generate what the supervisor said in the meeting. For example, it can provide a response such as, "The supervisor presented a new schedule for equipment maintenance and instructed that it be carried out within one week." This system enables users to quickly and accurately obtain the information they need.

[0417] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0418] Step 1: Collecting meeting minutes data

[0419] The server stores meeting minutes data automatically collected from machines in the factory in a specific directory. The machines in the factory generate work reports and meeting minutes data, which are then uploaded to the server in real time. The input is the meeting minutes data, and the output is data integrated into a directory or database. Specifically, the server scans the directory to detect new files and centrally manages their contents.

[0420] Step 2: Integrating and learning from meeting minutes data

[0421] The server combines the collected meeting minutes data into one large text file and uses it to train an AI model. Specifically, it uses the combined text data to train the GPT-2 model. The input is the combined text data, and the output is a trained AI model and tokenizer. The server tokenizes the text data, feeds it to the model, and saves the learning results.

[0422] Step 3: Initialize the chat bot

[0423] The server loads the trained artificial intelligence model and tokenizer and initializes a new chat bot instance. The input is the trained model and tokenizer, and the output is the initialized chat bot instance. Specifically, the server loads the saved model and tokenizer into memory and prepares the dialogue system.

[0424] Step 4: Enter user questions

[0425] A user uses a device (such as a smartphone or tablet) to input a question to the chat bot. This input is sent to the server through the device. The input is the user's question, and the output is a question request to the server. Specifically, the user types a question through the UI, which is then sent to the server via an HTTP request or similar.

[0426] Step 5: Question processing and response generation

[0427] The server generates an appropriate response based on the user's question using a trained artificial intelligence model. The input is the user's question, and the output is the generated response. The server tokenizes the question, inputs it into the model, and decodes and formats the generated text.

[0428] Step 6: View the response

[0429] The terminal receives the response sent from the server and displays it to the user. The input is the response from the server, and the output is the text displayed to the user. Specifically, the terminal parses the response from the server and displays it on the screen.

[0430] This series of steps allows users to quickly and accurately obtain the information they need in real time, for example, by providing an appropriate response to a prompt such as "Please tell me what my boss said in the meeting yesterday."

[0431] 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.

[0432] This invention combines a system that collects data from company meeting minutes and generates responses to user questions using an artificial intelligence model with an emotion engine that recognizes user emotions. This system is mainly composed of three main components: a server, a terminal, and a user.

[0433] 1. Collection of meeting minutes data

[0434] The server scans specific directories and databases where the company's meeting minutes data is stored and collects all meeting minutes files. At that time, it reads all text files in the specified directory and stores their contents in a list or array. This aggregates scattered meeting minutes data in one place.

[0435] 2. Training the AI ​​model

[0436] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model, combining the individual meeting minutes texts and merging them into one large text file, which is then used in the subsequent training process.

[0437] Next, train an artificial intelligence model using a generative model (e.g., GPT-2). Load the generative model and tokenizer, create a training dataset, and train the model based on the set parameters. After training is complete, save the trained model and tokenizer.

[0438] 3. Initializing the chat bot using the trained model

[0439] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which is responsible for generating responses to user queries.

[0440] 4. User Q&A

[0441] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[0442] 5. Use of Emotion Engine

[0443] The server passes the query received from the user to the emotion engine, which analyzes the user's emotion. The emotion engine analyzes the query and recognizes the emotion the user is feeling (e.g., joy, sadness, anger, surprise, etc.).

[0444] 6. Emotion-based response generation

[0445] The chat bot on the server uses the trained artificial intelligence model to generate responses to user queries, taking into account the emotions identified by the emotion engine. For example, if the user is angry, a more careful and polite response will be generated.

[0446] 7. Displaying the Response

[0447] The server uses a tokenizer to decode the generated response and convert it into a natural language response, such as "Project X was launched in April of last year and achieved a major milestone in June."

[0448] 8. Providing a Response

[0449] The device displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for. Furthermore, the response provided takes into consideration the user's feelings, improving the user experience.

[0450] In this way, the present invention helps office-based business people to efficiently search through minutes data and quickly obtain the information they need, while providing responses that take into consideration the user's feelings.

[0451] The processing flow will be explained below.

[0452] Step 1:

[0453] The server scans a specific directory or database where the company's meeting minutes data is stored and collects all meeting minutes files, which involves reading all text files in the directory and storing their contents in a list or array.

[0454] Step 2:

[0455] The server consolidates the collected meeting minutes data into a single text file, sequentially reads each meeting minute, and combines them into a single large text file, which is then used to train the AI ​​model.

[0456] Step 3:

[0457] The server loads a generative model (e.g., GPT-2) and a tokenizer, creates a training dataset, trains the model using training text files, and saves the trained model and tokenizer, ready for subsequent query-answer processing.

[0458] Step 4:

[0459] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which responds to user questions.

[0460] Step 5:

[0461] The user uses the terminal to input a question to the chat bot, for example, "Please tell me how project X progressed last year." This input is then sent from the terminal to the server.

[0462] Step 6:

[0463] Before parsing the received query, the server passes it to an emotion engine to analyze the user's emotions. The emotion engine identifies the user's emotions (e.g., joy, anger, sadness, surprise, etc.) based on the text input.

[0464] Step 7:

[0465] The chat bot on the server uses the trained AI model to generate responses to queries, taking into account the emotions identified by the emotion engine. For example, if the user expresses anger, the response will be more careful and polite.

[0466] Step 8:

[0467] The server uses a tokenizer to decode the generated response and convert it into a natural language response, for example, generating text like "Project X started in April of last year and achieved a major milestone in June."

[0468] Step 9:

[0469] The device displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for. Emotionally sensitive responses improve the user experience.

[0470] Example 2

[0471] 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."

[0472] Conventional information search systems that use internal company meeting minutes data have the problem that they respond to user questions in a formulaic manner and are unable to provide flexible responses that take the user's emotions and situation into consideration, resulting in a poor user experience.

[0473] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes means for collecting minutes data, means for training an artificial intelligence model using the minutes data, means for initializing a chat interface using the trained artificial intelligence model and a tokenizer, means for analyzing emotions in response to a user query using an emotion engine, means for generating a response taking emotions into consideration, and means for displaying the generated response on a user terminal. This makes it possible to provide a response that takes the user's emotions into consideration.

[0474] "Minutes data" refers to a text file or document that records the contents of a conference or meeting.

[0475] An "artificial intelligence model" is an algorithm or system that uses data to learn and generate responses to user queries.

[0476] A "tokenizer" is a tool for converting text data into a format that can be processed by a model.

[0477] A "chat interface" is an interactive user interface that allows users to enter questions and receive responses.

[0478] An "emotion engine" is a system or software that analyzes emotions from a user's text and labels it based on those emotions.

[0479] A "query" is a question or request that a user enters into a system.

[0480] A "user terminal" is a device that a user directly operates and interacts with the system. For example, a PC or smartphone is an example.

[0481] This invention is a system that collects meeting minutes data and generates responses to user questions using an artificial intelligence model. The system also incorporates an emotion engine that recognizes user emotions. The system mainly consists of three main components: a server, a terminal, and a user.

[0482] Use of collected meeting transcript data

[0483] The server scans specific directories and databases within the company to collect all meeting minutes files. It reads all text files in the specified directory and stores their contents in a list or array. For example, it scans the directory / data / meeting_minutes and collects text files. This aggregates scattered meeting minutes data into one place.

[0484] The learning process of artificial intelligence models

[0485] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model. It combines the individual meeting minutes texts and merges them into one large text file. It processes this merged text file using a tokenizer to create the training dataset for the model. The generative model used is GPT-2, and it loads this model and tokenizer and trains the model based on the set parameters. After training is complete, it saves the trained model and tokenizer.

[0486] Chat BOT initialization

[0487] The server initializes the chat bot using the saved trained model and tokenizer. It loads the model and tokenizer into memory and creates a new chat bot instance, which is responsible for generating responses to user queries.

[0488] User questions and answers

[0489] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[0490] Using the Emotion Engine

[0491] The server passes the query received from the user to the emotion engine, which analyzes the user's emotion. The emotion engine analyzes the query and recognizes the emotion the user is feeling (e.g., joy, sadness, anger, surprise, etc.).

[0492] Emotion-based response generation

[0493] The chat bot on the server uses the trained artificial intelligence model to generate responses to user queries, taking into account the emotions identified by the emotion engine. For example, if the user is angry, a more careful and polite response will be generated.

[0494] Viewing and Providing Responses

[0495] The server uses a tokenizer to decode the generated response and convert it into a natural language response. For example, the server might generate an answer such as, "Project X was launched in April of last year and achieved a major milestone in June." The device then displays this generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for.

[0496] In this way, the present invention helps office-based business people to efficiently search through minutes data and quickly obtain the information they need, while providing responses that take into consideration the user's feelings.

[0497] Specific examples

[0498] Example question: "How did project X go last year?"

[0499] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0500] Step 1:

[0501] The server scans a specific directory or database where meeting minutes data is stored and collects all meeting minutes files. As input, it receives a directory path (e.g., / data / meeting_minutes) and stores all text files in the directory in a list or array. Specifically, it uses the Python os module to detect text files in the directory, reads them, and adds them to a list. As output, it obtains a list of the paths to the text files.

[0502] Step 2:

[0503] The server combines the collected minutes data into a single large text file to create a training dataset for the AI ​​model. As input, it receives the path list of the minutes files obtained in step 1 and combines them into a single text file. Specifically, it opens each text file, concatenates its contents into a single string, and writes it to a new file. As output, it saves the combined training dataset as a single text file.

[0504] Step 3:

[0505] The server uses the integrated text file to train an artificial intelligence model (generative model). As input, it receives the integrated text file and training parameters (e.g., number of epochs, batch size, learning rate). Specifically, it uses Hugging Face's Transformers library to load a tokenizer and generative model (e.g., GPT-2), tokenizes the text data with the tokenizer, and trains the model. As output, it saves the trained model and tokenizer.

[0506] Step 4:

[0507] The server initializes the chat bot using the saved trained model and tokenizer. As input, it receives the trained model and tokenizer obtained in step 3 and loads them into memory. Specifically, it loads the model and tokenizer using Hugging Face's Transformers library and creates a new chat bot instance. As output, it obtains the initialized chat bot instance.

[0508] Step 5:

[0509] The user uses the terminal to input a question to the chat bot and send it. As input, the bot receives the query that the user input into the terminal (e.g., "Please tell me the progress of project X last year."). As a specific operation, the bot sends the user's input to the server via the send button or input completion event. As output, the user's query is sent to the server.

[0510] Step 6:

[0511] The server passes the received query to the emotion engine and analyzes the user's emotion. The server receives the user's query as input and passes it to the emotion engine. Specifically, it uses an emotion analysis library (e.g., TextBlob) to analyze the emotion of the query and obtains an emotion label (e.g., joy, sadness, anger, surprise, etc.). The output is the emotion label and the analysis result.

[0512] Step 7:

[0513] The chat bot on the server generates a response to the user's query while taking into account the emotions identified by the emotion engine. As input, it receives the user's query and emotion label. Specifically, it uses the generative AI model to create a prompt sentence based on the emotion label and generate a response. As output, it obtains the generated response text.

[0514] Step 8:

[0515] The server decodes the generated response using a tokenizer and converts it into a natural language response. As input, it receives the generated response text. Specifically, it uses the generative AI model's tokenizer to decode the response text and converts it into a natural language response. As output, it obtains a natural language response.

[0516] Step 9:

[0517] The terminal displays the generated response to the user. As input, it receives a natural language response and displays it to the user. As a specific operation, it adds the natural language response to an area that displays responses through a user interface. As output, it allows the user to confirm the response.

[0518] (Application example 2)

[0519] 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."

[0520] Physical stores require a means to respond appropriately and quickly to customer questions. Furthermore, responses that do not take into account the user's emotions present a risk of lowering customer satisfaction. Furthermore, responses that take emotions into account require the assistance of a human operator, which is costly and labor-intensive. To address these issues, the present invention aims to provide a system that automates customer service in physical stores, understands emotions, and responds appropriately.

[0521] 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 minutes data, means for training an artificial intelligence model using the minutes data, means for generating responses to questions from users using the trained artificial intelligence model, means for analyzing the user's emotions, and means for adjusting the response based on the emotion analysis results. This enables quick and emotional responses to customer questions in physical stores.

[0522] "Minutes data" is document data that records the contents of meetings and discussions.

[0523] An "artificial intelligence model" is a program that uses machine learning algorithms to learn patterns from data and make inferences and predictions.

[0524] "User" is a concept that refers to a person or organization that uses a system.

[0525] An "emotion engine" is software that analyzes and recognizes emotions from input data such as text and voice.

[0526] A "server" is a part of a computer system that provides services to other computers over a network.

[0527] "Emotion analysis result" is user emotion information output as a result of analysis by the emotion engine.

[0528] The "means for generating a response" refers to the process of using an artificial intelligence model to generate an appropriate answer to a question from a user.

[0529] A "collection method" is a process or method for gathering specific data from a designated location.

[0530] This invention is a system for providing fast and emotionally sensitive responses to customer questions in a brick-and-mortar store. The system includes three main components: a server, a terminal, and a user.

[0531] server

[0532] The server has the following roles:

[0533] 1. Collection of meeting minutes data: The server collects meeting minutes data from the specified directory or database, thereby consolidating scattered meeting minutes data in one place.

[0534] 2. Training the AI ​​model: The server uses the collected meeting minutes data to train a generative model (e.g., GPT-2). The meeting minutes data is combined into one large text file, and this file is used to train the AI ​​model.

[0535] 3. Sentiment Analysis: Pass the user query to the sentiment engine to analyze the sentiment. For example, use the TextBlob library to analyze the sentiment of the text.

[0536] 4. Response generation and adjustment: Based on the accumulated data and sentiment analysis results, a generative model is used to generate responses to user questions and adjust them to take sentiment into account.

[0537] Terminal

[0538] The terminal functions as follows:

[0539] 1. Providing a user interface: An interface is provided for users to input questions. This is typically implemented as a smartphone app.

[0540] 2. Question submission: The system has the function of submitting questions entered by the user to the server, which analyzes the submitted data and generates an appropriate response.

[0541] 3. Display Response: Display the response received from the server to the user.

[0542] User

[0543] The user is a customer who uses this system.

[0544] 1. Entering a question: The user enters a question using a terminal. For example, the user enters a specific question such as "Can I return this product?"

[0545] 2. Receiving a response: The user can check the response displayed on the terminal and get the necessary information immediately.

[0546] Usage example

[0547] Specific use cases include the following scenarios:

[0548] Example 1:

[0549] User Input: "Can I return this item?"

[0550] Emotion analysis result: Neutral

[0551] Generated response: "We're glad to help. This item can be returned within 30 days of purchase."

[0552] Prompt Sentence Examples

[0553] Below are some example prompts to input to a generative AI model:

[0554] A customer asked: "Can I return this item?"

[0555] Sentiment analysis was performed using TextBlob, and the user's sentiment score was 0.0.

[0556] The response generated using the GPT-2 model is shown below:

[0557] "We're glad we could help. This item can be returned within 30 days of purchase."

[0558] In this way, the present invention realizes a system that improves the efficiency of customer service in physical stores and provides quick responses that take emotions into consideration.

[0559] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0560] Step 1:

[0561] The server collects meeting minutes data from a specified directory or database. It receives the path of the target directory as input. The server reads all text files in the directory and stores their contents in a list. This aggregates the scattered meeting minutes data in one place, and outputs one large text data.

[0562] Step 2:

[0563] The server trains an AI model using the collected meeting minutes data. As input, it receives the integrated meeting minutes text data. The server loads a generative model such as GPT-2 and a tokenizer to create a training dataset. It trains the model using this dataset and saves the trained model and tokenizer. As output, it obtains the trained AI model and tokenizer.

[0564] Step 3:

[0565] The server initializes the chat bot using the trained artificial intelligence model and tokenizer. As input, it loads the saved trained model and tokenizer. The server reads them into memory and creates a new chat bot instance. As output, it obtains the initialized chat bot instance.

[0566] Step 4:

[0567] The user inputs a question using a terminal and sends it to the server. As input, a specific question text (e.g., "Can I return this product?") is obtained. The terminal sends this input to the server. As output, the question text sent to the server is obtained.

[0568] Step 5:

[0569] The server passes the question text received from the user to the emotion engine to analyze the sentiment. The server receives the question text as input. The server calculates the sentiment score of the text using an emotion analysis library such as TextBlob. The server receives the sentiment score as output.

[0570] Step 6:

[0571] The server uses a trained artificial intelligence model to generate a response to the question, taking into account the sentiment analysis results. The input is the question text and a sentiment score. The server inputs the question text into the generative model and tailors the response based on the sentiment score. The output is a tailored natural language response.

[0572] Step 7:

[0573] The server sends the generated response to the terminal. As input, it receives a generated natural language response. The server sends this response to the terminal. As output, it receives a natural language response sent to the terminal.

[0574] Step 8:

[0575] The terminal displays the response received from the server to the user. As input, it receives a natural language response sent from the server. The terminal displays this response to the user. As output, it receives a response displayed on the user's terminal screen.

[0576] 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.

[0577] 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.

[0578] 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.

[0579] [Third embodiment]

[0580] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0581] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0582] 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).

[0583] 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.

[0584] 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.

[0585] 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).

[0586] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0587] 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.

[0588] 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.

[0589] 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.

[0590] 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.

[0591] 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."

[0592] This system collects data from company meeting minutes and generates responses to user questions using an artificial intelligence model. This system is composed of three main components: a server, a terminal, and a user.

[0593] 1. Collection of meeting minutes data

[0594] The server has the ability to scan specific directories and databases where the company's meeting minutes data is stored and collect all meeting minutes files, thereby consolidating scattered meeting minutes data in one place.

[0595] Specifically, the server reads all text files in a specified directory and stores their contents in a list or array, which is later used to train an artificial intelligence model.

[0596] 2. Training the AI ​​model

[0597] The server uses the collected meeting minutes data to train an AI model. Specifically, it combines the text-based meeting minutes data into one large text file and uses it to train a generative model such as GPT-2.

[0598] It creates a training text dataset, and after training, saves the trained model and tokenizer, which are then used to generate answers to user questions.

[0599] 3. Initializing the chat bot using the trained model

[0600] The server loads the trained AI model and tokenizer and initializes a new chat bot instance, which is responsible for generating responses to user queries.

[0601] 4. User Q&A

[0602] The user inputs a question to the chat bot using a terminal. This input is sent to the server via the terminal. The chat bot on the server uses a trained artificial intelligence model to generate an appropriate response to the question. The generated response is then sent back to the terminal and displayed to the user.

[0603] As a concrete example, consider a case where a user asks, "Please tell me about the progress of Project X last year." This question is sent from the device to a chat bot, which generates a response based on the collected meeting minutes data. The bot then provides the user with an answer such as, "Project X started in April of last year and achieved major milestones in June. The main challenges were A and B, and C and D were proposed as solutions for each."

[0604] In this way, the present invention helps office workers to efficiently search through minutes data and quickly obtain the information they need.

[0605] The processing flow will be explained below.

[0606] Step 1:

[0607] The server scans specific directories and databases where the company's meeting minutes data is stored and collects all meeting minutes files. At that time, it reads all text files in the specified directory and stores their contents in a list or array. This aggregates scattered meeting minutes data in one place.

[0608] Step 2:

[0609] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model. Specifically, it combines the individual meeting minutes texts and integrates them into one large text file. This single text file is then used in the subsequent learning process.

[0610] Step 3:

[0611] The server trains an artificial intelligence model using a generative model (e.g., GPT-2). It loads the generative model and tokenizer, creates a training dataset, and trains the model based on the configured parameters. After training is complete, it saves the trained model and tokenizer.

[0612] Step 4:

[0613] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which is responsible for generating responses to user queries.

[0614] Step 5:

[0615] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[0616] Step 6:

[0617] The chat bot on the server processes the received query and generates an appropriate response using an artificial intelligence model. The query is encoded with a tokenizer, and the generated tokens are input into the model to generate a predicted response.

[0618] Step 7:

[0619] The server uses a tokenizer to decode the generated response and convert it into a natural language response, such as "Project X was launched in April of last year and achieved a major milestone in June."

[0620] Step 8:

[0621] The terminal displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for.

[0622] Example 1

[0623] 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."

[0624] Conventional meeting minutes data management systems make it difficult to efficiently find the necessary information from the vast amount of data. Manual search and information extraction are time-consuming, hindering productivity. Furthermore, the lack of advanced technology to generate appropriate responses makes it difficult for users to quickly obtain useful information.

[0625] 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.

[0626] In this invention, the server includes a means for scanning specific company directories and databases to collect meeting minutes data, a means for integrating the collected meeting minutes data into a single large text file and training a generative AI model using the text, and a means for loading the trained model and tokenizer and initializing a chat bot instance, thereby enabling the server to generate prompt and appropriate responses to user questions.

[0627] A "server" is a computer system that provides and processes information via a communication network.

[0628] A "directory" is a hierarchical unit that lists files and folders in a computer.

[0629] A "database" is a system that systematically stores and manages large amounts of data, allowing it to be searched and retrieved quickly.

[0630] "Minutes data" is document data that records the contents of meetings and discussions.

[0631] A "text file" is a file format for storing text data, and is generally saved as plain text.

[0632] A "generative AI model" is an artificial intelligence model that generates new text or information based on input data.

[0633] A "trained model" is an artificial intelligence model that has been fully trained using a specific dataset.

[0634] A "tokenizer" is a tool that divides text data into small units (tokens) such as words and phrases.

[0635] A "chat bot instance" is an instance of a chat bot that has been initialized to simulate a conversation with a user.

[0636] "User" means an individual or organization that uses the system.

[0637] This invention is a system that collects internal meeting minutes data and generates responses to user questions using a generative AI model. The system mainly consists of three main components: a server, a terminal, and a user.

[0638] Hardware and Software Configuration

[0639] server

[0640] The server is responsible for data collection, AI model training, and chat bot initialization and operation. Specifically, it uses programming languages ​​and libraries such as Python, TensorFlow, and PyTorch. The server also operates chat bot instances using web frameworks such as Flask and Django. MySQL or PostgreSQL are commonly used as databases.

[0641] Terminal

[0642] The device is a computer, tablet, smartphone, etc. that provides an interface for users to enter questions. The device is connected to the server via a web browser. The user interface is developed using HTML, CSS, JavaScript, etc.

[0643] User

[0644] A user is an individual or organization that uses the system to enter questions and receive responses. Users access the system through a terminal and obtain the information they need.

[0645] Data processing and calculation

[0646] The server scans designated directories and databases to collect meeting minutes data stored within the company. The collected meeting minutes data is consolidated into one large text file. This consolidated text file is then used to train a generative AI model, such as GPT-2. The trained model is then used to generate responses to user questions.

[0647] The terminal transmits the user's input to the server and displays the responses received from the server.

[0648] For example, if a user asks, "Please tell me about the progress of Project X last year," this input is sent from the terminal to the server. The chat bot on the server generates a response based on the collected meeting minutes data and provides the user with an answer such as, "Project X started in April of last year and achieved a major milestone in June. The main challenges were A and B, and C and D were proposed as solutions for each."

[0649] Prompt Sentence Examples

[0650] Below are some example prompts to input to a generative AI model:

[0651] Search for "Progress of Project X last year" in your company's meeting minutes data.

[0652] In this way, the present invention helps office workers to efficiently search through minutes data and quickly obtain the information they need.

[0653] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0654] Step 1:

[0655] The server scans specific directories and databases within the company to collect meeting minutes data. During this process, the server searches for all text and PDF files in the specified directory and retrieves their contents. The collected files are saved in temporary storage on the server. For example, files such as "meeting_20230310.txt" and "strategy_session.pdf" are retrieved from the " / corporate / minutes / " directory. The input for this step is the directory path, and the output is a list of the collected meeting minutes files.

[0656] Step 2:

[0657] The server combines the collected minutes files into one large text file. It reads each file and concatenates their contents into a single text file called "combined_minutes.txt." For example, it combines the text content extracted from "meeting_20230310.txt" and "strategy_session.pdf." The input for this step is a list of minutes files, and the output is a combined text file.

[0658] Step 3:

[0659] The server uses the combined text files to train a generative AI model (e.g., GPT-2). It trains the model using Python scripts and libraries (TensorFlow, PyTorch). During training, it uses the text data "combined_minutes.txt" as input and generates a trained model and tokenizer "trained_gpt2_model.pt" and "tokenizer.json" as outputs. Specifically, it uses a GPU to process each batch of data and optimize the model parameters.

[0660] Step 4:

[0661] The server loads the trained model and tokenizer and initializes a new chat bot instance. These are loaded on the server using a web framework such as Flask or Django. For example, start a Flask application, load "trained_gpt2_model.pt" and "tokenizer.json", and create a chat bot instance. The input of this step is the trained model and tokenizer, and the output is an initialized chat bot instance.

[0662] Step 5:

[0663] A user uses a terminal to input a question to the chat bot. For example, using a web browser, the user inputs a question such as "How did project X progress last year?" This input is sent to the server via the terminal. The input of this step is the user's question, and the output is a request to send to the server.

[0664] Step 6:

[0665] The chat bot on the server receives the user's question and generates an appropriate response using the trained model. As a specific example, the model generates text based on "Progress of Project X last year" and outputs the following response: "Project X started in April of last year and achieved major milestones in June. The main challenges were A and B, and C and D were proposed as solutions for each." The input of this step is the user's question, and the output is the generated response.

[0666] Step 7:

[0667] The server sends the generated response back to the device. The server sends the response text as an HTTP response. For example, it packs the generated response in JSON format and sends it back to the browser. The input of this step is the generated response, and the output is the response to the device.

[0668] Step 8:

[0669] The terminal receives the response from the server and displays it to the user. A concrete example is a web browser, which displays the received text on the screen for the user to view. The input to this step is the response from the server, and the output is what is displayed to the user.

[0670] (Application example 1)

[0671] 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."

[0672] With conventional methods for managing meeting minutes data, it was difficult to efficiently integrate data scattered across multiple locations, search it, and quickly obtain the necessary information. Furthermore, there was no environment in place for collecting work reports and meeting minutes data within the factory in real time and allowing on-site staff to access it immediately. This resulted in problems such as reduced work efficiency and delays in troubleshooting.

[0673] 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.

[0674] In this invention, the server includes means for collecting minutes data, means for training an artificial intelligence model using the minutes data, means for generating responses to questions from users using the trained artificial intelligence model, means for the user to input the questions using a terminal device and receive the responses, and means for automatically collecting the minutes data using machines in a factory and uploading it to the server. This enables efficient collection and integration of scattered minutes data, allowing users to quickly obtain the information they need in real time.

[0675] "Minutes data" is a text document that records the contents of meetings, work reports, and the like.

[0676] A "collection means" is a system component that scans a specific directory or database and aggregates the minutes data in one place.

[0677] The "artificial intelligence model" is a machine learning technology that learns from collected meeting minutes data and generates appropriate responses to questions from users.

[0678] "Training methods" are the processes and techniques used to train artificial intelligence models using collected meeting minutes data.

[0679] The "means for generating a response" is a system component that uses a trained artificial intelligence model to generate an appropriate answer to a question entered by a user.

[0680] A "terminal device" is a hardware device (e.g., smartphone, tablet, PC) through which a user accesses the system, enters questions, and receives responses.

[0681] The "machines in the factory" refer to devices and robots installed in the factory, which are devices that automatically collect meeting minutes data and upload it to the server.

[0682] A "server" is a computer system that centrally manages the collection of minutes data, learning of artificial intelligence models, and response generation.

[0683] The present invention is a system that collects meeting minutes data and generates responses to user questions using an artificial intelligence model. This system is mainly composed of three elements: a server, a terminal, and a user.

[0684] 1. Collection of meeting minutes data

[0685] The server has the function of scanning specific directories and databases and aggregating meeting minutes data automatically collected from machines in the factory into one place. For example, every time a meeting or work report is held in the factory, the data is uploaded to the server. This process allows scattered meeting minutes data to be collected in a centralized manner.

[0686] 2. Training the AI ​​model

[0687] The server uses the collected meeting minutes data to train an artificial intelligence model. Specifically, all meeting minutes data is integrated into one large text file, and a generative AI model such as GPT-2 is trained based on this. A training text dataset is created, and after learning, the trained model and tokenizer are saved. This trained model is used to generate responses to user questions.

[0688] 3. Initializing the chat bot using the trained model

[0689] The server loads the trained AI model and tokenizer and initializes a new chat bot instance, which is responsible for generating responses to user queries.

[0690] 4. User Q&A

[0691] A user uses a device (such as a smartphone or tablet) to input a question to the chat bot. The input question is sent to the server via the device. For example, a user might ask, "Please tell me the details of yesterday's troubleshooting." This question is sent from the device to the chat bot, and the chat bot generates a response based on the collected meeting minutes data. The user is provided with a response such as, "As for the details of yesterday's troubleshooting, a problem occurred with device A, and a software update was implemented as a solution."

[0692] To realize this system, the server uses the following hardware and software: The hardware includes the server device, in-factory collection equipment, and user terminals. The software includes a generative AI model such as GPT-2, a directory scanning program, a tokenizer, and a chat bot instance.

[0693] As a concrete example, consider the case where a factory engineer asks, "Please tell me what my boss said in the meeting yesterday." Examples of prompt sentences are as follows:

[0694] Prompt: "Tell me what your boss said in the meeting yesterday."

[0695] Based on this prompt, the trained model references detailed meeting minutes data to generate what the supervisor said in the meeting. For example, it can provide a response such as, "The supervisor presented a new schedule for equipment maintenance and instructed that it be carried out within one week." This system enables users to quickly and accurately obtain the information they need.

[0696] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0697] Step 1: Collecting meeting minutes data

[0698] The server stores meeting minutes data automatically collected from machines in the factory in a specific directory. The machines in the factory generate work reports and meeting minutes data, which are then uploaded to the server in real time. The input is the meeting minutes data, and the output is data integrated into a directory or database. Specifically, the server scans the directory to detect new files and centrally manages their contents.

[0699] Step 2: Integrating and learning from meeting minutes data

[0700] The server combines the collected meeting minutes data into one large text file and uses it to train an AI model. Specifically, it uses the combined text data to train the GPT-2 model. The input is the combined text data, and the output is a trained AI model and tokenizer. The server tokenizes the text data, feeds it to the model, and saves the learning results.

[0701] Step 3: Initialize the chat bot

[0702] The server loads the trained artificial intelligence model and tokenizer and initializes a new chat bot instance. The input is the trained model and tokenizer, and the output is the initialized chat bot instance. Specifically, the server loads the saved model and tokenizer into memory and prepares the dialogue system.

[0703] Step 4: Enter user questions

[0704] A user uses a device (such as a smartphone or tablet) to input a question to the chat bot. This input is sent to the server through the device. The input is the user's question, and the output is a question request to the server. Specifically, the user types a question through the UI, which is then sent to the server via an HTTP request or similar.

[0705] Step 5: Question processing and response generation

[0706] The server generates an appropriate response based on the user's question using a trained artificial intelligence model. The input is the user's question, and the output is the generated response. The server tokenizes the question, inputs it into the model, and decodes and formats the generated text.

[0707] Step 6: View the response

[0708] The terminal receives the response sent from the server and displays it to the user. The input is the response from the server, and the output is the text displayed to the user. Specifically, the terminal parses the response from the server and displays it on the screen.

[0709] This series of steps allows users to quickly and accurately obtain the information they need in real time, for example, by providing an appropriate response to a prompt such as "Please tell me what my boss said in the meeting yesterday."

[0710] 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.

[0711] This invention combines a system that collects data from company meeting minutes and generates responses to user questions using an artificial intelligence model with an emotion engine that recognizes user emotions. This system is mainly composed of three main components: a server, a terminal, and a user.

[0712] 1. Collection of meeting minutes data

[0713] The server scans specific directories and databases where the company's meeting minutes data is stored and collects all meeting minutes files. At that time, it reads all text files in the specified directory and stores their contents in a list or array. This aggregates scattered meeting minutes data in one place.

[0714] 2. Training the AI ​​model

[0715] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model, combining the individual meeting minutes texts and merging them into one large text file, which is then used in the subsequent training process.

[0716] Next, train an artificial intelligence model using a generative model (e.g., GPT-2). Load the generative model and tokenizer, create a training dataset, and train the model based on the set parameters. After training is complete, save the trained model and tokenizer.

[0717] 3. Initializing the chat bot using the trained model

[0718] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which is responsible for generating responses to user queries.

[0719] 4. User Q&A

[0720] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[0721] 5. Use of Emotion Engine

[0722] The server passes the query received from the user to the emotion engine, which analyzes the user's emotion. The emotion engine analyzes the query and recognizes the emotion the user is feeling (e.g., joy, sadness, anger, surprise, etc.).

[0723] 6. Emotion-based response generation

[0724] The chat bot on the server uses the trained artificial intelligence model to generate responses to user queries, taking into account the emotions identified by the emotion engine. For example, if the user is angry, a more careful and polite response will be generated.

[0725] 7. Displaying the Response

[0726] The server uses a tokenizer to decode the generated response and convert it into a natural language response, such as "Project X was launched in April of last year and achieved a major milestone in June."

[0727] 8. Providing a Response

[0728] The device displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for. Furthermore, the response provided takes into consideration the user's feelings, improving the user experience.

[0729] In this way, the present invention helps office-based business people to efficiently search through minutes data and quickly obtain the information they need, while providing responses that take into consideration the user's feelings.

[0730] The processing flow will be explained below.

[0731] Step 1:

[0732] The server scans a specific directory or database where the company's meeting minutes data is stored and collects all meeting minutes files, which involves reading all text files in the directory and storing their contents in a list or array.

[0733] Step 2:

[0734] The server consolidates the collected meeting minutes data into a single text file, sequentially reads each meeting minute, and combines them into a single large text file, which is then used to train the AI ​​model.

[0735] Step 3:

[0736] The server loads a generative model (e.g., GPT-2) and a tokenizer, creates a training dataset, trains the model using training text files, and saves the trained model and tokenizer, ready for subsequent query-answer processing.

[0737] Step 4:

[0738] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which responds to user questions.

[0739] Step 5:

[0740] The user uses the terminal to input a question to the chat bot, for example, "Please tell me how project X progressed last year." This input is then sent from the terminal to the server.

[0741] Step 6:

[0742] Before parsing the received query, the server passes it to an emotion engine to analyze the user's emotions. The emotion engine identifies the user's emotions (e.g., joy, anger, sadness, surprise, etc.) based on the text input.

[0743] Step 7:

[0744] The chat bot on the server uses the trained AI model to generate responses to queries, taking into account the emotions identified by the emotion engine. For example, if the user expresses anger, the response will be more careful and polite.

[0745] Step 8:

[0746] The server uses a tokenizer to decode the generated response and convert it into a natural language response, for example, generating text like "Project X started in April of last year and achieved a major milestone in June."

[0747] Step 9:

[0748] The device displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for. Emotionally sensitive responses improve the user experience.

[0749] Example 2

[0750] 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."

[0751] Conventional information search systems that use internal company meeting minutes data have the problem that they respond to user questions in a formulaic manner and are unable to provide flexible responses that take the user's emotions and situation into consideration, resulting in a poor user experience.

[0752] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes means for collecting minutes data, means for training an artificial intelligence model using the minutes data, means for initializing a chat interface using the trained artificial intelligence model and a tokenizer, means for analyzing emotions in response to a user query using an emotion engine, means for generating a response taking emotions into consideration, and means for displaying the generated response on a user terminal. This makes it possible to provide a response that takes the user's emotions into consideration.

[0753] "Minutes data" refers to a text file or document that records the contents of a conference or meeting.

[0754] An "artificial intelligence model" is an algorithm or system that uses data to learn and generate responses to user queries.

[0755] A "tokenizer" is a tool for converting text data into a format that can be processed by a model.

[0756] A "chat interface" is an interactive user interface that allows users to enter questions and receive responses.

[0757] An "emotion engine" is a system or software that analyzes emotions from a user's text and labels it based on those emotions.

[0758] A "query" is a question or request that a user enters into a system.

[0759] A "user terminal" is a device that a user directly operates and interacts with the system. For example, a PC or smartphone is an example.

[0760] This invention is a system that collects meeting minutes data and generates responses to user questions using an artificial intelligence model. The system also incorporates an emotion engine that recognizes user emotions. The system mainly consists of three main components: a server, a terminal, and a user.

[0761] Use of collected meeting transcript data

[0762] The server scans specific directories and databases within the company to collect all meeting minutes files. It reads all text files in the specified directory and stores their contents in a list or array. For example, it scans the directory / data / meeting_minutes and collects text files. This aggregates scattered meeting minutes data into one place.

[0763] The learning process of artificial intelligence models

[0764] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model. It combines the individual meeting minutes texts and merges them into one large text file. It processes this merged text file using a tokenizer to create the training dataset for the model. The generative model used is GPT-2, and it loads this model and tokenizer and trains the model based on the set parameters. After training is complete, it saves the trained model and tokenizer.

[0765] Chat BOT initialization

[0766] The server initializes the chat bot using the saved trained model and tokenizer. It loads the model and tokenizer into memory and creates a new chat bot instance, which is responsible for generating responses to user queries.

[0767] User questions and answers

[0768] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[0769] Using the Emotion Engine

[0770] The server passes the query received from the user to the emotion engine, which analyzes the user's emotion. The emotion engine analyzes the query and recognizes the emotion the user is feeling (e.g., joy, sadness, anger, surprise, etc.).

[0771] Emotion-based response generation

[0772] The chat bot on the server uses the trained artificial intelligence model to generate responses to user queries, taking into account the emotions identified by the emotion engine. For example, if the user is angry, a more careful and polite response will be generated.

[0773] Viewing and Providing Responses

[0774] The server uses a tokenizer to decode the generated response and convert it into a natural language response. For example, the server might generate an answer such as, "Project X was launched in April of last year and achieved a major milestone in June." The device then displays this generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for.

[0775] In this way, the present invention helps office-based business people to efficiently search through minutes data and quickly obtain the information they need, while providing responses that take into consideration the user's feelings.

[0776] Specific examples

[0777] Example question: "How did project X go last year?"

[0778] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0779] Step 1:

[0780] The server scans a specific directory or database where meeting minutes data is stored and collects all meeting minutes files. As input, it receives a directory path (e.g., / data / meeting_minutes) and stores all text files in the directory in a list or array. Specifically, it uses the Python os module to detect text files in the directory, reads them, and adds them to a list. As output, it obtains a list of the paths to the text files.

[0781] Step 2:

[0782] The server combines the collected minutes data into a single large text file to create a training dataset for the AI ​​model. As input, it receives the path list of the minutes files obtained in step 1 and combines them into a single text file. Specifically, it opens each text file, concatenates its contents into a single string, and writes it to a new file. As output, it saves the combined training dataset as a single text file.

[0783] Step 3:

[0784] The server uses the integrated text file to train an artificial intelligence model (generative model). As input, it receives the integrated text file and training parameters (e.g., number of epochs, batch size, learning rate). Specifically, it uses Hugging Face's Transformers library to load a tokenizer and generative model (e.g., GPT-2), tokenizes the text data with the tokenizer, and trains the model. As output, it saves the trained model and tokenizer.

[0785] Step 4:

[0786] The server initializes the chat bot using the saved trained model and tokenizer. As input, it receives the trained model and tokenizer obtained in step 3 and loads them into memory. Specifically, it loads the model and tokenizer using Hugging Face's Transformers library and creates a new chat bot instance. As output, it obtains the initialized chat bot instance.

[0787] Step 5:

[0788] The user uses the terminal to input a question to the chat bot and send it. As input, the bot receives the query that the user input into the terminal (e.g., "Please tell me the progress of project X last year."). As a specific operation, the bot sends the user's input to the server via the send button or input completion event. As output, the user's query is sent to the server.

[0789] Step 6:

[0790] The server passes the received query to the emotion engine and analyzes the user's emotion. The server receives the user's query as input and passes it to the emotion engine. Specifically, it uses an emotion analysis library (e.g., TextBlob) to analyze the emotion of the query and obtains an emotion label (e.g., joy, sadness, anger, surprise, etc.). The output is the emotion label and the analysis result.

[0791] Step 7:

[0792] The chat bot on the server generates a response to the user's query while taking into account the emotions identified by the emotion engine. As input, it receives the user's query and emotion label. Specifically, it uses the generative AI model to create a prompt sentence based on the emotion label and generate a response. As output, it obtains the generated response text.

[0793] Step 8:

[0794] The server decodes the generated response using a tokenizer and converts it into a natural language response. As input, it receives the generated response text. Specifically, it uses the generative AI model's tokenizer to decode the response text and converts it into a natural language response. As output, it obtains a natural language response.

[0795] Step 9:

[0796] The terminal displays the generated response to the user. As input, it receives a natural language response and displays it to the user. As a specific operation, it adds the natural language response to an area that displays responses through a user interface. As output, it allows the user to confirm the response.

[0797] (Application example 2)

[0798] 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."

[0799] Physical stores require a means to respond appropriately and quickly to customer questions. Furthermore, responses that do not take into account the user's emotions present a risk of lowering customer satisfaction. Furthermore, responses that take emotions into account require the assistance of a human operator, which is costly and labor-intensive. To address these issues, the present invention aims to provide a system that automates customer service in physical stores, understands emotions, and responds appropriately.

[0800] 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 minutes data, means for training an artificial intelligence model using the minutes data, means for generating responses to questions from users using the trained artificial intelligence model, means for analyzing the user's emotions, and means for adjusting the response based on the emotion analysis results. This enables quick and emotional responses to customer questions in physical stores.

[0801] "Minutes data" is document data that records the contents of meetings and discussions.

[0802] An "artificial intelligence model" is a program that uses machine learning algorithms to learn patterns from data and make inferences and predictions.

[0803] "User" is a concept that refers to a person or organization that uses a system.

[0804] An "emotion engine" is software that analyzes and recognizes emotions from input data such as text and voice.

[0805] A "server" is a part of a computer system that provides services to other computers over a network.

[0806] "Emotion analysis result" is user emotion information output as a result of analysis by the emotion engine.

[0807] The "means for generating a response" refers to the process of using an artificial intelligence model to generate an appropriate answer to a question from a user.

[0808] A "collection method" is a process or method for gathering specific data from a designated location.

[0809] This invention is a system for providing fast and emotionally sensitive responses to customer questions in a brick-and-mortar store. The system includes three main components: a server, a terminal, and a user.

[0810] server

[0811] The server has the following roles:

[0812] 1. Collection of meeting minutes data: The server collects meeting minutes data from the specified directory or database, thereby consolidating scattered meeting minutes data in one place.

[0813] 2. Training the AI ​​model: The server uses the collected meeting minutes data to train a generative model (e.g., GPT-2). The meeting minutes data is combined into one large text file, and this file is used to train the AI ​​model.

[0814] 3. Sentiment Analysis: Pass the user query to the sentiment engine to analyze the sentiment. For example, use the TextBlob library to analyze the sentiment of the text.

[0815] 4. Response generation and adjustment: Based on the accumulated data and sentiment analysis results, a generative model is used to generate responses to user questions and adjust them to take sentiment into account.

[0816] Terminal

[0817] The terminal functions as follows:

[0818] 1. Providing a user interface: An interface is provided for users to input questions. This is typically implemented as a smartphone app.

[0819] 2. Question submission: The system has the function of submitting questions entered by the user to the server, which analyzes the submitted data and generates an appropriate response.

[0820] 3. Display Response: Display the response received from the server to the user.

[0821] User

[0822] The user is a customer who uses this system.

[0823] 1. Entering a question: The user enters a question using a terminal. For example, the user enters a specific question such as "Can I return this product?"

[0824] 2. Receiving a response: The user can check the response displayed on the terminal and get the necessary information immediately.

[0825] Usage example

[0826] Specific use cases include the following scenarios:

[0827] Example 1:

[0828] User Input: "Can I return this item?"

[0829] Emotion analysis result: Neutral

[0830] Generated response: "We're glad to help. This item can be returned within 30 days of purchase."

[0831] Prompt Sentence Examples

[0832] Below are some example prompts to input to a generative AI model:

[0833] A customer asked: "Can I return this item?"

[0834] Sentiment analysis was performed using TextBlob, and the user's sentiment score was 0.0.

[0835] The response generated using the GPT-2 model is shown below:

[0836] "We're glad we could help. This item can be returned within 30 days of purchase."

[0837] In this way, the present invention realizes a system that improves the efficiency of customer service in physical stores and provides quick responses that take emotions into consideration.

[0838] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0839] Step 1:

[0840] The server collects meeting minutes data from a specified directory or database. It receives the path of the target directory as input. The server reads all text files in the directory and stores their contents in a list. This aggregates the scattered meeting minutes data in one place, and outputs one large text data.

[0841] Step 2:

[0842] The server trains an AI model using the collected meeting minutes data. As input, it receives the integrated meeting minutes text data. The server loads a generative model such as GPT-2 and a tokenizer to create a training dataset. It trains the model using this dataset and saves the trained model and tokenizer. As output, it obtains the trained AI model and tokenizer.

[0843] Step 3:

[0844] The server initializes the chat bot using the trained artificial intelligence model and tokenizer. As input, it loads the saved trained model and tokenizer. The server reads them into memory and creates a new chat bot instance. As output, it obtains the initialized chat bot instance.

[0845] Step 4:

[0846] The user inputs a question using a terminal and sends it to the server. As input, a specific question text (e.g., "Can I return this product?") is obtained. The terminal sends this input to the server. As output, the question text sent to the server is obtained.

[0847] Step 5:

[0848] The server passes the question text received from the user to the emotion engine to analyze the sentiment. The server receives the question text as input. The server calculates the sentiment score of the text using an emotion analysis library such as TextBlob. The server receives the sentiment score as output.

[0849] Step 6:

[0850] The server uses a trained artificial intelligence model to generate a response to the question, taking into account the sentiment analysis results. The input is the question text and a sentiment score. The server inputs the question text into the generative model and tailors the response based on the sentiment score. The output is a tailored natural language response.

[0851] Step 7:

[0852] The server sends the generated response to the terminal. As input, it receives a generated natural language response. The server sends this response to the terminal. As output, it receives a natural language response sent to the terminal.

[0853] Step 8:

[0854] The terminal displays the response received from the server to the user. As input, it receives a natural language response sent from the server. The terminal displays this response to the user. As output, it receives a response displayed on the user's terminal screen.

[0855] 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.

[0856] 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.

[0857] 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.

[0858] [Fourth embodiment]

[0859] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[0860] 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.

[0861] 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).

[0862] 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.

[0863] 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.

[0864] 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).

[0865] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0866] 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.

[0867] 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.

[0868] 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.

[0869] 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.

[0870] 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.

[0871] 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."

[0872] This system collects data from company meeting minutes and generates responses to user questions using an artificial intelligence model. This system is composed of three main components: a server, a terminal, and a user.

[0873] 1. Collection of meeting minutes data

[0874] The server has the ability to scan specific directories and databases where the company's meeting minutes data is stored and collect all meeting minutes files, thereby consolidating scattered meeting minutes data in one place.

[0875] Specifically, the server reads all text files in a specified directory and stores their contents in a list or array, which is later used to train an artificial intelligence model.

[0876] 2. Training the AI ​​model

[0877] The server uses the collected meeting minutes data to train an AI model. Specifically, it combines the text-based meeting minutes data into one large text file and uses it to train a generative model such as GPT-2.

[0878] It creates a training text dataset, and after training, saves the trained model and tokenizer, which are then used to generate answers to user questions.

[0879] 3. Initializing the chat bot using the trained model

[0880] The server loads the trained AI model and tokenizer and initializes a new chat bot instance, which is responsible for generating responses to user queries.

[0881] 4. User Q&A

[0882] The user inputs a question to the chat bot using a terminal. This input is sent to the server via the terminal. The chat bot on the server uses a trained artificial intelligence model to generate an appropriate response to the question. The generated response is then sent back to the terminal and displayed to the user.

[0883] As a concrete example, consider a case where a user asks, "Please tell me about the progress of Project X last year." This question is sent from the device to a chat bot, which generates a response based on the collected meeting minutes data. The bot then provides the user with an answer such as, "Project X started in April of last year and achieved major milestones in June. The main challenges were A and B, and C and D were proposed as solutions for each."

[0884] In this way, the present invention helps office workers to efficiently search through minutes data and quickly obtain the information they need.

[0885] The processing flow will be explained below.

[0886] Step 1:

[0887] The server scans specific directories and databases where the company's meeting minutes data is stored and collects all meeting minutes files. At that time, it reads all text files in the specified directory and stores their contents in a list or array. This aggregates scattered meeting minutes data in one place.

[0888] Step 2:

[0889] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model. Specifically, it combines the individual meeting minutes texts and integrates them into one large text file. This single text file is then used in the subsequent learning process.

[0890] Step 3:

[0891] The server trains an artificial intelligence model using a generative model (e.g., GPT-2). It loads the generative model and tokenizer, creates a training dataset, and trains the model based on the configured parameters. After training is complete, it saves the trained model and tokenizer.

[0892] Step 4:

[0893] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which is responsible for generating responses to user queries.

[0894] Step 5:

[0895] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[0896] Step 6:

[0897] The chat bot on the server processes the received query and generates an appropriate response using an artificial intelligence model. The query is encoded with a tokenizer, and the generated tokens are input into the model to generate a predicted response.

[0898] Step 7:

[0899] The server uses a tokenizer to decode the generated response and convert it into a natural language response, such as "Project X was launched in April of last year and achieved a major milestone in June."

[0900] Step 8:

[0901] The terminal displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for.

[0902] Example 1

[0903] 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."

[0904] Conventional meeting minutes data management systems make it difficult to efficiently find the necessary information from the vast amount of data. Manual search and information extraction are time-consuming, hindering productivity. Furthermore, the lack of advanced technology to generate appropriate responses makes it difficult for users to quickly obtain useful information.

[0905] 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.

[0906] In this invention, the server includes a means for scanning specific company directories and databases to collect meeting minutes data, a means for integrating the collected meeting minutes data into a single large text file and training a generative AI model using the text, and a means for loading the trained model and tokenizer and initializing a chat bot instance, thereby enabling the server to generate prompt and appropriate responses to user questions.

[0907] A "server" is a computer system that provides and processes information via a communication network.

[0908] A "directory" is a hierarchical unit that lists files and folders in a computer.

[0909] A "database" is a system that systematically stores and manages large amounts of data, allowing it to be searched and retrieved quickly.

[0910] "Minutes data" is document data that records the contents of meetings and discussions.

[0911] A "text file" is a file format for storing text data, and is generally saved as plain text.

[0912] A "generative AI model" is an artificial intelligence model that generates new text or information based on input data.

[0913] A "trained model" is an artificial intelligence model that has been fully trained using a specific dataset.

[0914] A "tokenizer" is a tool that divides text data into small units (tokens) such as words and phrases.

[0915] A "chat bot instance" is an instance of a chat bot that has been initialized to simulate a conversation with a user.

[0916] "User" means an individual or organization that uses the system.

[0917] This invention is a system that collects internal meeting minutes data and generates responses to user questions using a generative AI model. The system mainly consists of three main components: a server, a terminal, and a user.

[0918] Hardware and Software Configuration

[0919] server

[0920] The server is responsible for data collection, AI model training, and chat bot initialization and operation. Specifically, it uses programming languages ​​and libraries such as Python, TensorFlow, and PyTorch. The server also operates chat bot instances using web frameworks such as Flask and Django. MySQL or PostgreSQL are commonly used as databases.

[0921] Terminal

[0922] The device is a computer, tablet, smartphone, etc. that provides an interface for users to enter questions. The device is connected to the server via a web browser. The user interface is developed using HTML, CSS, JavaScript, etc.

[0923] User

[0924] A user is an individual or organization that uses the system to enter questions and receive responses. Users access the system through a terminal and obtain the information they need.

[0925] Data processing and calculation

[0926] The server scans designated directories and databases to collect meeting minutes data stored within the company. The collected meeting minutes data is consolidated into one large text file. This consolidated text file is then used to train a generative AI model, such as GPT-2. The trained model is then used to generate responses to user questions.

[0927] The terminal transmits the user's input to the server and displays the responses received from the server.

[0928] For example, if a user asks, "Please tell me about the progress of Project X last year," this input is sent from the terminal to the server. The chat bot on the server generates a response based on the collected meeting minutes data and provides the user with an answer such as, "Project X started in April of last year and achieved a major milestone in June. The main challenges were A and B, and C and D were proposed as solutions for each."

[0929] Prompt Sentence Examples

[0930] Below are some example prompts to input to a generative AI model:

[0931] Search for "Progress of Project X last year" in your company's meeting minutes data.

[0932] In this way, the present invention helps office workers to efficiently search through minutes data and quickly obtain the information they need.

[0933] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0934] Step 1:

[0935] The server scans specific directories and databases within the company to collect meeting minutes data. During this process, the server searches for all text and PDF files in the specified directory and retrieves their contents. The collected files are saved in temporary storage on the server. For example, files such as "meeting_20230310.txt" and "strategy_session.pdf" are retrieved from the " / corporate / minutes / " directory. The input for this step is the directory path, and the output is a list of the collected meeting minutes files.

[0936] Step 2:

[0937] The server combines the collected minutes files into one large text file. It reads each file and concatenates their contents into a single text file called "combined_minutes.txt." For example, it combines the text content extracted from "meeting_20230310.txt" and "strategy_session.pdf." The input for this step is a list of minutes files, and the output is a combined text file.

[0938] Step 3:

[0939] The server uses the combined text files to train a generative AI model (e.g., GPT-2). It trains the model using Python scripts and libraries (TensorFlow, PyTorch). During training, it uses the text data "combined_minutes.txt" as input and generates a trained model and tokenizer "trained_gpt2_model.pt" and "tokenizer.json" as outputs. Specifically, it uses a GPU to process each batch of data and optimize the model parameters.

[0940] Step 4:

[0941] The server loads the trained model and tokenizer and initializes a new chat bot instance. These are loaded on the server using a web framework such as Flask or Django. For example, start a Flask application, load "trained_gpt2_model.pt" and "tokenizer.json", and create a chat bot instance. The input of this step is the trained model and tokenizer, and the output is an initialized chat bot instance.

[0942] Step 5:

[0943] A user uses a terminal to input a question to the chat bot. For example, using a web browser, the user inputs a question such as "How did project X progress last year?" This input is sent to the server via the terminal. The input of this step is the user's question, and the output is a request to send to the server.

[0944] Step 6:

[0945] The chat bot on the server receives the user's question and generates an appropriate response using the trained model. As a specific example, the model generates text based on "Progress of Project X last year" and outputs the following response: "Project X started in April of last year and achieved major milestones in June. The main challenges were A and B, and C and D were proposed as solutions for each." The input of this step is the user's question, and the output is the generated response.

[0946] Step 7:

[0947] The server sends the generated response back to the device. The server sends the response text as an HTTP response. For example, it packs the generated response in JSON format and sends it back to the browser. The input of this step is the generated response, and the output is the response to the device.

[0948] Step 8:

[0949] The terminal receives the response from the server and displays it to the user. A concrete example is a web browser, which displays the received text on the screen for the user to view. The input to this step is the response from the server, and the output is what is displayed to the user.

[0950] (Application example 1)

[0951] 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."

[0952] With conventional methods for managing meeting minutes data, it was difficult to efficiently integrate data scattered across multiple locations, search it, and quickly obtain the necessary information. Furthermore, there was no environment in place for collecting work reports and meeting minutes data within the factory in real time and allowing on-site staff to access it immediately. This resulted in problems such as reduced work efficiency and delays in troubleshooting.

[0953] 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.

[0954] In this invention, the server includes means for collecting minutes data, means for training an artificial intelligence model using the minutes data, means for generating responses to questions from users using the trained artificial intelligence model, means for the user to input the questions using a terminal device and receive the responses, and means for automatically collecting the minutes data using machines in a factory and uploading it to the server. This enables efficient collection and integration of scattered minutes data, allowing users to quickly obtain the information they need in real time.

[0955] "Minutes data" is a text document that records the contents of meetings, work reports, and the like.

[0956] A "collection means" is a system component that scans a specific directory or database and aggregates the minutes data in one place.

[0957] The "artificial intelligence model" is a machine learning technology that learns from collected meeting minutes data and generates appropriate responses to questions from users.

[0958] "Training methods" are the processes and techniques used to train artificial intelligence models using collected meeting minutes data.

[0959] The "means for generating a response" is a system component that uses a trained artificial intelligence model to generate an appropriate answer to a question entered by a user.

[0960] A "terminal device" is a hardware device (e.g., smartphone, tablet, PC) through which a user accesses the system, enters questions, and receives responses.

[0961] The "machines in the factory" refer to devices and robots installed in the factory, which are devices that automatically collect meeting minutes data and upload it to the server.

[0962] A "server" is a computer system that centrally manages the collection of minutes data, learning of artificial intelligence models, and response generation.

[0963] The present invention is a system that collects meeting minutes data and generates responses to user questions using an artificial intelligence model. This system is mainly composed of three elements: a server, a terminal, and a user.

[0964] 1. Collection of meeting minutes data

[0965] The server has the function of scanning specific directories and databases and aggregating meeting minutes data automatically collected from machines in the factory into one place. For example, every time a meeting or work report is held in the factory, the data is uploaded to the server. This process allows scattered meeting minutes data to be collected in a centralized manner.

[0966] 2. Training the AI ​​model

[0967] The server uses the collected meeting minutes data to train an artificial intelligence model. Specifically, all meeting minutes data is integrated into one large text file, and a generative AI model such as GPT-2 is trained based on this. A training text dataset is created, and after learning, the trained model and tokenizer are saved. This trained model is used to generate responses to user questions.

[0968] 3. Initializing the chat bot using the trained model

[0969] The server loads the trained AI model and tokenizer and initializes a new chat bot instance, which is responsible for generating responses to user queries.

[0970] 4. User Q&A

[0971] A user uses a device (such as a smartphone or tablet) to input a question to the chat bot. The input question is sent to the server via the device. For example, a user might ask, "Please tell me the details of yesterday's troubleshooting." This question is sent from the device to the chat bot, and the chat bot generates a response based on the collected meeting minutes data. The user is provided with a response such as, "As for the details of yesterday's troubleshooting, a problem occurred with device A, and a software update was implemented as a solution."

[0972] To realize this system, the server uses the following hardware and software: The hardware includes the server device, in-factory collection equipment, and user terminals. The software includes a generative AI model such as GPT-2, a directory scanning program, a tokenizer, and a chat bot instance.

[0973] As a concrete example, consider the case where a factory engineer asks, "Please tell me what my boss said in the meeting yesterday." Examples of prompt sentences are as follows:

[0974] Prompt: "Tell me what your boss said in the meeting yesterday."

[0975] Based on this prompt, the trained model references detailed meeting minutes data to generate what the supervisor said in the meeting. For example, it can provide a response such as, "The supervisor presented a new schedule for equipment maintenance and instructed that it be carried out within one week." This system enables users to quickly and accurately obtain the information they need.

[0976] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0977] Step 1: Collecting meeting minutes data

[0978] The server stores meeting minutes data automatically collected from machines in the factory in a specific directory. The machines in the factory generate work reports and meeting minutes data, which are then uploaded to the server in real time. The input is the meeting minutes data, and the output is data integrated into a directory or database. Specifically, the server scans the directory to detect new files and centrally manages their contents.

[0979] Step 2: Integrating and learning from meeting minutes data

[0980] The server combines the collected meeting minutes data into one large text file and uses it to train an AI model. Specifically, it uses the combined text data to train the GPT-2 model. The input is the combined text data, and the output is a trained AI model and tokenizer. The server tokenizes the text data, feeds it to the model, and saves the learning results.

[0981] Step 3: Initialize the chat bot

[0982] The server loads the trained artificial intelligence model and tokenizer and initializes a new chat bot instance. The input is the trained model and tokenizer, and the output is the initialized chat bot instance. Specifically, the server loads the saved model and tokenizer into memory and prepares the dialogue system.

[0983] Step 4: Enter user questions

[0984] A user uses a device (such as a smartphone or tablet) to input a question to the chat bot. This input is sent to the server through the device. The input is the user's question, and the output is a question request to the server. Specifically, the user types a question through the UI, which is then sent to the server via an HTTP request or similar.

[0985] Step 5: Question processing and response generation

[0986] The server generates an appropriate response based on the user's question using a trained artificial intelligence model. The input is the user's question, and the output is the generated response. The server tokenizes the question, inputs it into the model, and decodes and formats the generated text.

[0987] Step 6: View the response

[0988] The terminal receives the response sent from the server and displays it to the user. The input is the response from the server, and the output is the text displayed to the user. Specifically, the terminal parses the response from the server and displays it on the screen.

[0989] This series of steps allows users to quickly and accurately obtain the information they need in real time, for example, by providing an appropriate response to a prompt such as "Please tell me what my boss said in the meeting yesterday."

[0990] 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.

[0991] This invention combines a system that collects data from company meeting minutes and generates responses to user questions using an artificial intelligence model with an emotion engine that recognizes user emotions. This system is mainly composed of three main components: a server, a terminal, and a user.

[0992] 1. Collection of meeting minutes data

[0993] The server scans specific directories and databases where the company's meeting minutes data is stored and collects all meeting minutes files. At that time, it reads all text files in the specified directory and stores their contents in a list or array. This aggregates scattered meeting minutes data in one place.

[0994] 2. Training the AI ​​model

[0995] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model, combining the individual meeting minutes texts and merging them into one large text file, which is then used in the subsequent training process.

[0996] Next, train an artificial intelligence model using a generative model (e.g., GPT-2). Load the generative model and tokenizer, create a training dataset, and train the model based on the set parameters. After training is complete, save the trained model and tokenizer.

[0997] 3. Initializing the chat bot using the trained model

[0998] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which is responsible for generating responses to user queries.

[0999] 4. User Q&A

[1000] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[1001] 5. Use of Emotion Engine

[1002] The server passes the query received from the user to the emotion engine, which analyzes the user's emotion. The emotion engine analyzes the query and recognizes the emotion the user is feeling (e.g., joy, sadness, anger, surprise, etc.).

[1003] 6. Emotion-based response generation

[1004] The chat bot on the server uses the trained artificial intelligence model to generate responses to user queries, taking into account the emotions identified by the emotion engine. For example, if the user is angry, a more careful and polite response will be generated.

[1005] 7. Displaying the Response

[1006] The server uses a tokenizer to decode the generated response and convert it into a natural language response, such as "Project X was launched in April of last year and achieved a major milestone in June."

[1007] 8. Providing a Response

[1008] The device displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for. Furthermore, the response provided takes into consideration the user's feelings, improving the user experience.

[1009] In this way, the present invention helps office-based business people to efficiently search through minutes data and quickly obtain the information they need, while providing responses that take into consideration the user's feelings.

[1010] The processing flow will be explained below.

[1011] Step 1:

[1012] The server scans a specific directory or database where the company's meeting minutes data is stored and collects all meeting minutes files, which involves reading all text files in the directory and storing their contents in a list or array.

[1013] Step 2:

[1014] The server consolidates the collected meeting minutes data into a single text file, sequentially reads each meeting minute, and combines them into a single large text file, which is then used to train the AI ​​model.

[1015] Step 3:

[1016] The server loads a generative model (e.g., GPT-2) and a tokenizer, creates a training dataset, trains the model using training text files, and saves the trained model and tokenizer, ready for subsequent query-answer processing.

[1017] Step 4:

[1018] The server initializes the chat bot using the saved trained model and tokenizer, which involves loading the model and tokenizer into memory and creating a new chat bot instance, which responds to user questions.

[1019] Step 5:

[1020] The user uses the terminal to input a question to the chat bot, for example, "Please tell me how project X progressed last year." This input is then sent from the terminal to the server.

[1021] Step 6:

[1022] Before parsing the received query, the server passes it to an emotion engine to analyze the user's emotions. The emotion engine identifies the user's emotions (e.g., joy, anger, sadness, surprise, etc.) based on the text input.

[1023] Step 7:

[1024] The chat bot on the server uses the trained AI model to generate responses to queries, taking into account the emotions identified by the emotion engine. For example, if the user expresses anger, the response will be more careful and polite.

[1025] Step 8:

[1026] The server uses a tokenizer to decode the generated response and convert it into a natural language response, for example, generating text like "Project X started in April of last year and achieved a major milestone in June."

[1027] Step 9:

[1028] The device displays the generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for. Emotionally sensitive responses improve the user experience.

[1029] Example 2

[1030] 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."

[1031] Conventional information search systems that use internal company meeting minutes data have the problem that they respond to user questions in a formulaic manner and are unable to provide flexible responses that take the user's emotions and situation into consideration, resulting in a poor user experience.

[1032] The identification process by the identification processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means. In this invention, the server includes means for collecting minutes data, means for training an artificial intelligence model using the minutes data, means for initializing a chat interface using the trained artificial intelligence model and a tokenizer, means for analyzing emotions in response to a user query using an emotion engine, means for generating a response taking emotions into consideration, and means for displaying the generated response on a user terminal. This makes it possible to provide a response that takes the user's emotions into consideration.

[1033] "Minutes data" refers to a text file or document that records the contents of a conference or meeting.

[1034] An "artificial intelligence model" is an algorithm or system that uses data to learn and generate responses to user queries.

[1035] A "tokenizer" is a tool for converting text data into a format that can be processed by a model.

[1036] A "chat interface" is an interactive user interface that allows users to enter questions and receive responses.

[1037] An "emotion engine" is a system or software that analyzes emotions from a user's text and labels it based on those emotions.

[1038] A "query" is a question or request that a user enters into a system.

[1039] A "user terminal" is a device that a user directly operates and interacts with the system. For example, a PC or smartphone is an example.

[1040] This invention is a system that collects meeting minutes data and generates responses to user questions using an artificial intelligence model. The system also incorporates an emotion engine that recognizes user emotions. The system mainly consists of three main components: a server, a terminal, and a user.

[1041] Use of collected meeting transcript data

[1042] The server scans specific directories and databases within the company to collect all meeting minutes files. It reads all text files in the specified directory and stores their contents in a list or array. For example, it scans the directory / data / meeting_minutes and collects text files. This aggregates scattered meeting minutes data into one place.

[1043] The learning process of artificial intelligence models

[1044] The server uses the collected meeting minutes data to create a training dataset for the AI ​​model. It combines the individual meeting minutes texts and merges them into one large text file. It processes this merged text file using a tokenizer to create the training dataset for the model. The generative model used is GPT-2, and it loads this model and tokenizer and trains the model based on the set parameters. After training is complete, it saves the trained model and tokenizer.

[1045] Chat BOT initialization

[1046] The server initializes the chat bot using the saved trained model and tokenizer. It loads the model and tokenizer into memory and creates a new chat bot instance, which is responsible for generating responses to user queries.

[1047] User questions and answers

[1048] The user uses the terminal to input a question to the chat bot and send it. For example, they input a query such as "Please tell me the progress of project X last year." This input is sent from the terminal to the server.

[1049] Using the Emotion Engine

[1050] The server passes the query received from the user to the emotion engine, which analyzes the user's emotion. The emotion engine analyzes the query and recognizes the emotion the user is feeling (e.g., joy, sadness, anger, surprise, etc.).

[1051] Emotion-based response generation

[1052] The chat bot on the server uses the trained artificial intelligence model to generate responses to user queries, taking into account the emotions identified by the emotion engine. For example, if the user is angry, a more careful and polite response will be generated.

[1053] Viewing and Providing Responses

[1054] The server uses a tokenizer to decode the generated response and convert it into a natural language response. For example, the server might generate an answer such as, "Project X was launched in April of last year and achieved a major milestone in June." The device then displays this generated response to the user, allowing the user to quickly and accurately obtain the information they were looking for.

[1055] In this way, the present invention helps office-based business people to efficiently search through minutes data and quickly obtain the information they need, while providing responses that take into consideration the user's feelings.

[1056] Specific examples

[1057] Example question: "How did project X go last year?"

[1058] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1059] Step 1:

[1060] The server scans a specific directory or database where meeting minutes data is stored and collects all meeting minutes files. As input, it receives a directory path (e.g., / data / meeting_minutes) and stores all text files in the directory in a list or array. Specifically, it uses the Python os module to detect text files in the directory, reads them, and adds them to a list. As output, it obtains a list of the paths to the text files.

[1061] Step 2:

[1062] The server combines the collected minutes data into a single large text file to create a training dataset for the AI ​​model. As input, it receives the path list of the minutes files obtained in step 1 and combines them into a single text file. Specifically, it opens each text file, concatenates its contents into a single string, and writes it to a new file. As output, it saves the combined training dataset as a single text file.

[1063] Step 3:

[1064] The server uses the integrated text file to train an artificial intelligence model (generative model). As input, it receives the integrated text file and training parameters (e.g., number of epochs, batch size, learning rate). Specifically, it uses Hugging Face's Transformers library to load a tokenizer and generative model (e.g., GPT-2), tokenizes the text data with the tokenizer, and trains the model. As output, it saves the trained model and tokenizer.

[1065] Step 4:

[1066] The server initializes the chat bot using the saved trained model and tokenizer. As input, it receives the trained model and tokenizer obtained in step 3 and loads them into memory. Specifically, it loads the model and tokenizer using Hugging Face's Transformers library and creates a new chat bot instance. As output, it obtains the initialized chat bot instance.

[1067] Step 5:

[1068] The user uses the terminal to input a question to the chat bot and send it. As input, the bot receives the query that the user input into the terminal (e.g., "Please tell me the progress of project X last year."). As a specific operation, the bot sends the user's input to the server via the send button or input completion event. As output, the user's query is sent to the server.

[1069] Step 6:

[1070] The server passes the received query to the emotion engine and analyzes the user's emotion. The server receives the user's query as input and passes it to the emotion engine. Specifically, it uses an emotion analysis library (e.g., TextBlob) to analyze the emotion of the query and obtains an emotion label (e.g., joy, sadness, anger, surprise, etc.). The output is the emotion label and the analysis result.

[1071] Step 7:

[1072] The chat bot on the server generates a response to the user's query while taking into account the emotions identified by the emotion engine. As input, it receives the user's query and emotion label. Specifically, it uses the generative AI model to create a prompt sentence based on the emotion label and generate a response. As output, it obtains the generated response text.

[1073] Step 8:

[1074] The server decodes the generated response using a tokenizer and converts it into a natural language response. As input, it receives the generated response text. Specifically, it uses the generative AI model's tokenizer to decode the response text and converts it into a natural language response. As output, it obtains a natural language response.

[1075] Step 9:

[1076] The terminal displays the generated response to the user. As input, it receives a natural language response and displays it to the user. As a specific operation, it adds the natural language response to an area that displays responses through a user interface. As output, it allows the user to confirm the response.

[1077] (Application example 2)

[1078] 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."

[1079] Physical stores require a means to respond appropriately and quickly to customer questions. Furthermore, responses that do not take into account the user's emotions present a risk of lowering customer satisfaction. Furthermore, responses that take emotions into account require the assistance of a human operator, which is costly and labor-intensive. To address these issues, the present invention aims to provide a system that automates customer service in physical stores, understands emotions, and responds appropriately.

[1080] 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 minutes data, means for training an artificial intelligence model using the minutes data, means for generating responses to questions from users using the trained artificial intelligence model, means for analyzing the user's emotions, and means for adjusting the response based on the emotion analysis results. This enables quick and emotional responses to customer questions in physical stores.

[1081] "Minutes data" is document data that records the contents of meetings and discussions.

[1082] An "artificial intelligence model" is a program that uses machine learning algorithms to learn patterns from data and make inferences and predictions.

[1083] "User" is a concept that refers to a person or organization that uses a system.

[1084] An "emotion engine" is software that analyzes and recognizes emotions from input data such as text and voice.

[1085] A "server" is a part of a computer system that provides services to other computers over a network.

[1086] "Emotion analysis result" is user emotion information output as a result of analysis by the emotion engine.

[1087] The "means for generating a response" refers to the process of using an artificial intelligence model to generate an appropriate answer to a question from a user.

[1088] A "collection method" is a process or method for gathering specific data from a designated location.

[1089] This invention is a system for providing fast and emotionally sensitive responses to customer questions in a brick-and-mortar store. The system includes three main components: a server, a terminal, and a user.

[1090] server

[1091] The server has the following roles:

[1092] 1. Collection of meeting minutes data: The server collects meeting minutes data from the specified directory or database, thereby consolidating scattered meeting minutes data in one place.

[1093] 2. Training the AI ​​model: The server uses the collected meeting minutes data to train a generative model (e.g., GPT-2). The meeting minutes data is combined into one large text file, and this file is used to train the AI ​​model.

[1094] 3. Sentiment Analysis: Pass the user query to the sentiment engine to analyze the sentiment. For example, use the TextBlob library to analyze the sentiment of the text.

[1095] 4. Response generation and adjustment: Based on the accumulated data and sentiment analysis results, a generative model is used to generate responses to user questions and adjust them to take sentiment into account.

[1096] Terminal

[1097] The terminal functions as follows:

[1098] 1. Providing a user interface: An interface is provided for users to input questions. This is typically implemented as a smartphone app.

[1099] 2. Question submission: The system has the function of submitting questions entered by the user to the server, which analyzes the submitted data and generates an appropriate response.

[1100] 3. Display Response: Display the response received from the server to the user.

[1101] User

[1102] The user is a customer who uses this system.

[1103] 1. Entering a question: The user enters a question using a terminal. For example, the user enters a specific question such as "Can I return this product?"

[1104] 2. Receiving a response: The user can check the response displayed on the terminal and get the necessary information immediately.

[1105] Usage example

[1106] Specific use cases include the following scenarios:

[1107] Example 1:

[1108] User Input: "Can I return this item?"

[1109] Emotion analysis result: Neutral

[1110] Generated response: "We're glad to help. This item can be returned within 30 days of purchase."

[1111] Prompt Sentence Examples

[1112] Below are some example prompts to input to a generative AI model:

[1113] A customer asked: "Can I return this item?"

[1114] Sentiment analysis was performed using TextBlob, and the user's sentiment score was 0.0.

[1115] The response generated using the GPT-2 model is shown below:

[1116] "We're glad we could help. This item can be returned within 30 days of purchase."

[1117] In this way, the present invention realizes a system that improves the efficiency of customer service in physical stores and provides quick responses that take emotions into consideration.

[1118] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1119] Step 1:

[1120] The server collects meeting minutes data from a specified directory or database. It receives the path of the target directory as input. The server reads all text files in the directory and stores their contents in a list. This aggregates the scattered meeting minutes data in one place, and outputs one large text data.

[1121] Step 2:

[1122] The server trains an AI model using the collected meeting minutes data. As input, it receives the integrated meeting minutes text data. The server loads a generative model such as GPT-2 and a tokenizer to create a training dataset. It trains the model using this dataset and saves the trained model and tokenizer. As output, it obtains the trained AI model and tokenizer.

[1123] Step 3:

[1124] The server initializes the chat bot using the trained artificial intelligence model and tokenizer. As input, it loads the saved trained model and tokenizer. The server reads them into memory and creates a new chat bot instance. As output, it obtains the initialized chat bot instance.

[1125] Step 4:

[1126] The user inputs a question using a terminal and sends it to the server. As input, a specific question text (e.g., "Can I return this product?") is obtained. The terminal sends this input to the server. As output, the question text sent to the server is obtained.

[1127] Step 5:

[1128] The server passes the question text received from the user to the emotion engine to analyze the sentiment. The server receives the question text as input. The server calculates the sentiment score of the text using an emotion analysis library such as TextBlob. The server receives the sentiment score as output.

[1129] Step 6:

[1130] The server uses a trained artificial intelligence model to generate a response to the question, taking into account the sentiment analysis results. The input is the question text and a sentiment score. The server inputs the question text into the generative model and tailors the response based on the sentiment score. The output is a tailored natural language response.

[1131] Step 7:

[1132] The server sends the generated response to the terminal. As input, it receives a generated natural language response. The server sends this response to the terminal. As output, it receives a natural language response sent to the terminal.

[1133] Step 8:

[1134] The terminal displays the response received from the server to the user. As input, it receives a natural language response sent from the server. The terminal displays this response to the user. As output, it receives a response displayed on the user's terminal screen.

[1135] 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.

[1136] 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.

[1137] 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.

[1138] 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.

[1139] 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.

[1140] 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.

[1141] 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).

[1142] 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.

[1143] 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."

[1144] 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.

[1145] 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).

[1146] 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.

[1147] 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.

[1148] 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.

[1149] 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.

[1150] 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.

[1151] 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.

[1152] 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.

[1153] 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.

[1154] 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.

[1155] 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.

[1156] The following is further disclosed regarding the above embodiment.

[1157] (Claim 1)

[1158] a means of collecting meeting minutes data;

[1159] means for training an artificial intelligence model using the minutes data;

[1160] means for generating responses to questions posed by a user using the trained artificial intelligence model;

[1161] A system including:

[1162] (Claim 2)

[1163] The system of claim 1 , wherein the artificial intelligence model is a generative model.

[1164] (Claim 3)

[1165] 2. The system according to claim 1, wherein the minutes data is a document stored in text format.

[1166] "Example 1"

[1167] (Claim 1)

[1168] A means of scanning specific internal directories and databases to collect meeting minutes data;

[1169] A method to integrate the collected minutes data into one large text file and use that text to train a generative AI model.

[1170] A means to load the trained model and tokenizer and initialize a chat BOT instance.

[1171] a means for receiving user input, generating a response to the user's question using the trained generative AI model, and providing the response to the user;

[1172] A system including:

[1173] (Claim 2)

[1174] 2. The system of claim 1, wherein the generative AI model is a GPT-2 model.

[1175] (Claim 3)

[1176] 2. The system according to claim 1, wherein the minutes data is a document stored in text format.

[1177] "Application Example 1"

[1178] (Claim 1)

[1179] a means of collecting meeting minutes data;

[1180] means for training an artificial intelligence model using the minutes data;

[1181] means for generating responses to questions posed by a user using the trained artificial intelligence model;

[1182] means for the user to input the question using a terminal device and receive the response;

[1183] A means for automatically collecting the minutes data by a machine in the factory and uploading it to a server;

[1184] A system including:

[1185] (Claim 2)

[1186] The system of claim 1 , wherein the artificial intelligence model is a generative model.

[1187] (Claim 3)

[1188] 2. The system according to claim 1, wherein the minutes data is a document stored in text format.

[1189] "Example 2: Combining Emotion Engines"

[1190] (Claim 1)

[1191] a means of collecting meeting minutes data;

[1192] means for training an artificial intelligence model using the minutes data;

[1193] a means for initializing a chat interface using the trained artificial intelligence model and the tokenizer;

[1194] means for analyzing emotions in response to a user query using an emotion engine;

[1195] a means for generating a response taking into account emotions;

[1196] means for displaying the generated response on a user terminal;

[1197] A system including:

[1198] (Claim 2)

[1199] The system of claim 1 , wherein the artificial intelligence model is a generative model.

[1200] (Claim 3)

[1201] 2. The system according to claim 1, wherein the minutes data is a document stored in text format.

[1202] "Application example 2 when combining emotion engines"

[1203] (Claim 1)

[1204] a means of collecting meeting minutes data;

[1205] means for training an artificial intelligence model using the minutes data;

[1206] means for generating responses to questions posed by a user using the trained artificial intelligence model;

[1207] means for analyzing user emotions;

[1208] means for adjusting a response based on the emotion analysis result;

[1209] A system including:

[1210] (Claim 2)

[1211] The system of claim 1 , wherein the artificial intelligence model is a generative model.

[1212] (Claim 3)

[1213] 2. The system according to claim 1, wherein the minutes data is a document stored in text format. [Explanation of symbols]

[1214] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. a means of collecting meeting minutes data; means for training an artificial intelligence model using the minutes data; means for generating responses to questions posed by a user using the trained artificial intelligence model; A system including:

2. The system of claim 1 , wherein the artificial intelligence model is a generative model.

3. The system according to claim 1 , wherein the minutes data is a document stored in a text format.

Images