System

Abstract

A system is provided.SOLUTION: A system comprising: means for collecting and storing in a database instruction manuals and frequently asked question data for an information handling terminal; means for analyzing the collected data using a generative model and training the model; means for receiving user entered queries and passing them to the trained model; means for returning responses generated from the model to the user; and means for recording the queries and generated responses and periodically evaluating the performance of the 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] In existing sales activities, sales representatives spend a lot of time and effort responding to customer inquiries about how to operate mobile devices or how to use corporate websites. This slows down sales activities, reduces work efficiency, and often makes it difficult for customers to obtain appropriate answers. The objective of this invention is to provide a system that solves these problems and allows customers to obtain quick and accurate answers while reducing the burden on sales representatives. [Means for solving the problem]

[0005] The present invention provides a system including: a means for collecting instruction manuals and frequently asked questions (FAQ) data for information terminals and storing them in a database; a means for analyzing the collected data using a generative model and training the model; a means for receiving inquiries entered by users and passing them to the trained model; a means for returning responses generated by the model to the user; and a means for recording inquiries and generated responses and periodically evaluating the performance of the model. Furthermore, by providing a means for collecting additional data and retraining if the model's performance falls below a predetermined standard, and a means for evaluating user feedback, the accuracy of the model can be continuously improved. This allows sales representatives to focus on their core tasks without spending time researching operation methods, and allows customers to receive prompt and appropriate support.

[0006] "Information terminals handled" refers to electronic devices such as mobile terminals, tablets, and smartphones from various manufacturers and brands, and includes information on how to operate them and their functions.

[0007] A "generative model" refers to an algorithm that applies machine learning and artificial intelligence techniques to extract patterns and knowledge from large amounts of data and generate new information and responses.

[0008] "Database" means a structured collection of data that allows information to be stored efficiently and securely, and quickly searched and retrieved when needed.

[0009] "Training" refers to the process of training an algorithm using large amounts of example data so that the generative model learns appropriate patterns and rules.

[0010] An "inquiry" refers to a question or request made by a user to a system to learn specific information or a method of operation.

[0011] A "response" refers to an answer or instruction that a generative model generates in response to a user's query.

[0012] "Recording" refers to the act of storing data from inquiries received and responses generated by a system for later reference and analysis.

[0013] "Performance evaluation" refers to a set of measures for measuring the quality and accuracy of the responses produced by a generative model and determining whether improvement is needed.

[0014] "Retraining" refers to the process of adding new data and retraining a generative model in order to improve the model's accuracy.

[0015] "Feedback" refers to the act of users providing ratings and comments on the responses provided, based on which the system is improved. [Brief explanation of the drawings]

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

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

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

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

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

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

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

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

[0024] [First embodiment]

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

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

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

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

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

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

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

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

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

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

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

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

[0037] The present invention is an operation method support system that uses a generative model to provide convenience to both salespeople and customers. The overall flow of this system will be explained from the perspectives of the server, terminals, and users.

[0038] 1. Data collection and learning phase

[0039] server

[0040] The server first collects instruction manuals and FAQ (frequently asked questions) data for all information devices handled from the Internet and internal databases, and then centrally manages them. Since this data is used as learning material for the model, it is cleaned in advance to remove duplicate data and unnecessary information. After data cleaning is complete, the data is organized and saved.

[0041] The server then trains the pre-processed data with a generative model using machine learning algorithms, which then serves as the basis for quickly and accurately responding to future queries.

[0042] 2. Inquiry Receipt and Response Phase

[0043] User

[0044] When a user has a question about an operation method or a specific function, they input their query into the system through a dedicated user interface, which may be provided in the form of a web browser, a mobile application, or a help desk.

[0045] Terminal

[0046] User inquiries received via the user interface are sent from the terminal to the server. The terminal converts the inquiry content into a format such as JSON and sends it to the server's data reception API.

[0047] server

[0048] The server analyzes the received query and passes it to a trained generative model. The generative model generates an appropriate answer based on the query, constructs a response, and returns it to the server. The server then sends this response back to the user, allowing the user to receive a quick and accurate answer.

[0049] 3. Performance evaluation and improvement phase

[0050] server

[0051] The server records all queries and responses in log files. Periodically, it analyzes these logs to evaluate the model's performance. This evaluation involves scoring the generated responses to ensure their accuracy. Based on the evaluation results, if the model's accuracy falls below a predetermined standard, it may automatically initiate a process to collect additional data and retrain.

[0052] User

[0053] Additionally, users can enter feedback on the answers they provide, which is sent to the server and used as data for future model improvements.

[0054] Specific examples

[0055] For example, suppose a user asks, "How do I set up Wi-Fi?" The flow is shown below.

[0056] 1. User enters a query

[0057] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[0058] 2. The device sends a query to the server

[0059] The browser on the user's device sends this query to the server's API.

[0060] 3. The server receives the query and passes it to the generative model.

[0061] The server analyzes the received query and inputs it into the generative model.

[0062] 4. The generative model generates a response

[0063] The generative model retrieves information about "how to set up Wi-Fi" from a database and generates appropriate instructions.

[0064] 5. The server sends the response back to the user

[0065] The server generates a response and sends it back to the user, who displays it in their browser.

[0066] 6. The user receives the response and performs the action

[0067] The user follows the displayed steps to set up Wi-Fi.

[0068] This allows users to receive prompt and appropriate responses, and frees sales representatives from having to deal with inquiries about operation methods, allowing them to focus on more important tasks.

[0069] The processing flow will be explained below.

[0070] Step 1:

[0071] The server collects instruction manuals and FAQ data from the information terminals it handles and stores them in a database, including scraping data from the Internet and retrieving it from its internal database.

[0072] Step 2:

[0073] The server preprocesses the collected data, specifically by performing text cleaning and data normalization to remove duplicate data and unnecessary information.

[0074] Step 3:

[0075] The server trains a generative model based on the preprocessed data, using machine learning algorithms to extract the necessary knowledge from the data and train the model.

[0076] Step 4:

[0077] The server stores the trained generative model and prepares it for generating responses to queries.

[0078] Step 5:

[0079] The user inputs an inquiry about operation methods through a dedicated user interface, for example, "Please tell me how to set up Wi-Fi."

[0080] Step 6:

[0081] The terminal converts the user's inquiry into JSON format or similar and sends it to the server's API. This transmission generally uses the HTTPS protocol.

[0082] Step 7:

[0083] The server parses the received query, performs any necessary preprocessing, and then passes it to the generative model. Parsing includes tokenizing and encoding the text.

[0084] Step 8:

[0085] The server receives the response from the generative model and transforms it into a format that is easy for the user to understand, for example, organizing the steps as a numbered list.

[0086] Step 9:

[0087] The server sends the converted response back to the user via the API to the device, again using a secure protocol.

[0088] Step 10:

[0089] The device displays the response received from the server and provides it to the user, either on a web page or within an application.

[0090] Step 11:

[0091] Performs actions based on the answers the user provides, for example, walking through Wi-Fi setup steps to resolve the issue.

[0092] Step 12:

[0093] The server logs user queries and the responses it generates, providing data for later performance evaluation and model improvement.

[0094] Step 13:

[0095] The server periodically analyzes the log data and evaluates the performance of the generative model using metrics such as accuracy and response time.

[0096] Step 14:

[0097] If the accuracy of the model falls below a predetermined standard, the server collects new data and retrains the model, ensuring that it always provides highly accurate responses.

[0098] Step 15:

[0099] The user inputs feedback on the provided responses and sends it to the server via the terminal, which is used to improve the system.

[0100] As a result, this system can reduce the burden on sales staff and improve customer satisfaction.

[0101] Example 1

[0102] 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."

[0103] Many modern information terminals have complex operation methods and multifunctional settings, which often lead to users becoming confused and not receiving appropriate support. Sales representatives are also overwhelmed by responding to a variety of inquiries, preventing them from concentrating on their core duties. Furthermore, traditional FAQs and instruction manuals only provide static information, making it difficult to provide dynamic and prompt support.

[0104] 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.

[0105] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for cleaning and preprocessing the collected data, means for training the preprocessed data using a generative model, means for receiving inquiries entered by users, converting the inquiry content into JSON format, and sending the JSON format to the server, means for inputting the user's inquiry into the trained generative model and obtaining a response generated from the model, means for returning the obtained response to the user interface, and means for recording the inquiries and the generated responses and periodically evaluating the performance of the model. This allows users to receive prompt and appropriate support, and frees sales representatives from responding to inquiries so that they can focus on their core business.

[0106] An "information terminal" is a device or system that provides information about an electronic device or system operated by a user.

[0107] An "instruction manual" is a document that describes the basic operation methods, setting procedures, precautions, etc. of an information terminal.

[0108] "Frequently asked questions data" is data that compiles questions frequently asked by users and the answers to those questions.

[0109] A "database" is a system for efficiently storing, managing, and searching large amounts of data.

[0110] "Data cleanup" is the process of removing redundancies and noise from collected data and preparing it in a format suitable for analysis.

[0111] "Preprocessing" is a series of steps to convert data into a format suitable for a generative model.

[0112] A "generative model" is a model that uses machine learning algorithms to generate optimal answers from given data.

[0113] "Training" is the process of optimizing the parameters of a generative model using collected data to train the model.

[0114] A "user interface" is an interactive screen or application that allows a user to access and query a system.

[0115] "JSON format" is a data format that structures data in text format and makes it easily exchangeable.

[0116] A "server" is a computer system that processes data and provides services over a network.

[0117] An "inquiry" is a question or request made by a user to the system.

[0118] A "response" is an answer that a system provides to a user's query.

[0119] "Recording" means saving the contents of inquiries and responses in a log file or the like.

[0120] "Performance evaluation" is the process of evaluating the accuracy and appropriateness of the responses of a generative model.

[0121] This invention is an operation method support system that uses a generative AI model to provide convenience to both sales representatives and customers. In this system, the server, terminal, and user each play specific roles, and provide quick and accurate answers to user inquiries.

[0122] Data collection and preprocessing

[0123] server

[0124] The server collects the user manuals and FAQ data for the information terminals from the internet and an internal database. It also cleans the data, removing duplicate data and noise. Specifically, it uses Python scripts and BeautifulSoup to scrape the web, and the Pandas library to process the data. The cleaned data is then stored using a large-scale data processing framework (e.g., Hadoop).

[0125] Training generative AI models

[0126] server

[0127] The server trains a generative model using the collected and preprocessed data. TensorFlow or PyTorch is used for this training. The training data is split using a tokenizer and input to a neural network. A deep learning framework is used to optimize the model parameters. In particular, the Transformers architecture is commonly used.

[0128] Receiving inquiries via a user interface

[0129] User

[0130] Users input questions about operations and specific functions through a user interface, which is provided by a web browser or a mobile application, such as Google Chrome or Safari.

[0131] Specific examples

[0132] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[0133] Analysis of inquiries

[0134] Terminal

[0135] The device converts the user's inquiry into JSON format and sends it to the server's receiving API. For example, it generates JSON data using a JavaScript or Python library and sends it to the server via an HTTP request. Browsers used include Firefox and Google Chrome.

[0136] Answer generation using generative AI models

[0137] server

[0138] The server inputs the received query into a generative model. This process involves encoding the query and converting it into a format suitable for the generative model. The generative AI model (e.g., GPT-3) generates an appropriate answer based on the query. The generated answer is then formatted.

[0139] Specific examples

[0140] The server uses the generative model to generate instructions on how to set up Wi-Fi.

[0141] Returning answers to users

[0142] server

[0143] The server generates an answer and sends it back to the user. The response data is organized in JSON format and sent to the device in an HTTP response.

[0144] Terminal

[0145] The device analyzes the returned data and displays it in a user interface, using HTML and JavaScript to display the answers in the browser.

[0146] Performance evaluation and feedback collection

[0147] server

[0148] The server logs all queries and responses and periodically analyzes them. The log data is evaluated using Python and R to measure the accuracy of the responses. Users can enter feedback on the answers provided, and this data is also collected.

[0149] This process allows users to receive prompt and appropriate answers, and frees up sales representatives from the burden of handling inquiries. The system continuously collects data to improve the model's performance, providing a high-quality user experience.

[0150] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0151] Step 1:

[0152] Data collection and preprocessing

[0153] The server collects the user manuals and frequently asked questions (FAQ data) for the information terminals from the internet and an internal database. The input uses the URL of the online user manual and the access information for the internal database. Specifically, it performs web scraping using a Python script and BeautifulSoup to extract the data. After the data is collected, it is formatted using Pandas to remove duplicate data and noise. The clean data is generated as output and stored in Hadoop or another storage system.

[0154] Step 2:

[0155] Training generative AI models

[0156] The server trains a generative model using the collected and preprocessed data. It uses clean training data as input. Specifically, it tokenizes the training data using the TensorFlow or PyTorch framework and feeds it to a neural network. The selected deep learning model (e.g., Transformers architecture) learns patterns in the data and finds optimal parameters. The output is a trained generative model.

[0157] Step 3:

[0158] Receiving inquiries via a user interface

[0159] The user inputs questions about operation methods or specific functions through the user interface. The user's question is used as input. Specifically, the user enters a question into an inquiry form in a web browser such as Google Chrome or Safari and clicks the submit button. The user's question data is sent to the device as output.

[0160] Step 4:

[0161] Analysis of inquiries

[0162] The terminal converts the user's inquiry into JSON format and sends it to the server's receiving API. The user's question received from the web browser is used as input. Specifically, it generates JSON data using JavaScript or Python libraries and sends it to the server via an HTTP request. The JSON-formatted inquiry data is sent to the server as output.

[0163] Step 5:

[0164] Answer generation using generative AI models

[0165] The server inputs the received query content into the generative model. Parsed JSON data is used as input. Specifically, the server encodes the query content and converts it into a format suitable for the generative model. The generative AI model (e.g., GPT-3) generates an appropriate answer based on the query content. The output is the generated response text.

[0166] Step 6:

[0167] Returning answers to users

[0168] The server returns the generated answer to the user. The generated response text is used as input. Specifically, the server formats the response data in JSON format and sends it to the terminal as an HTTP response. The terminal receives the returned data and displays it in the user interface. As output, the final response is displayed on the user's browser.

[0169] Step 7:

[0170] Performance evaluation and feedback collection

[0171] The server logs all queries and responses and periodically analyzes them. The query and response log data is used as input. Specific operations include evaluating the log data using Python or R to measure the model's performance. Users can provide feedback on the answers provided, and this data is also collected. The output is the model's performance evaluation results and user feedback.

[0172] This processing step allows the system to respond quickly and accurately to user inquiries and to continually collect data to improve performance.

[0173] (Application example 1)

[0174] 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."

[0175] Conventional systems have struggled to respond quickly and accurately to user inquiries about content search and playback settings. Content distribution services, in particular, require appropriate recommendations based on users' viewing history and recent popular content, but there has been a lack of efficient ways to provide this information. Furthermore, systems lack the ability to automatically update data and retrain when system performance deteriorates.

[0176] 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.

[0177] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving queries entered by users and passing them to the trained model, means for returning responses generated from the model to the users, means for recording queries and the generated responses and periodically evaluating the performance of the model, and means for generating content recommendations based on the user's viewing history and recent popular content in content distribution, thereby enabling users to receive appropriate recommendations and quick responses regarding content they want to view.

[0178] The term "information terminals" refers to various electronic devices and content playback terminals used by users.

[0179] "Instruction manual" refers to a document that contains detailed information about the operation and functions of the information terminal.

[0180] "Frequently asked questions data" refers to documents or data that systematically compile answers to questions and problems that users frequently have.

[0181] A "database" refers to a system for efficiently storing, searching, and managing collected information and data.

[0182] "Generative model" refers to a system trained using machine learning algorithms to generate appropriate responses to user queries.

[0183] "Training" refers to the process of using collected data to train a generative model and improve its accuracy.

[0184] "Content distribution" refers to the service of providing digital media such as movies, TV shows, and music to users via the Internet.

[0185] "Viewing history" refers to a record of content that a user has viewed in the past.

[0186] "Popular content" refers to digital media that is viewed by many users and has received high ratings.

[0187] "Content recommendation" refers to suggesting the next piece of media to watch based on the user's viewing history and popular content.

[0188] This invention applies a content distribution service to an operation method support system that uses a generative model to provide convenience to both sales representatives and customers. The details of this system will be explained in three phases: data collection and learning phase, inquiry reception and response phase, and performance evaluation and improvement phase.

[0189] 1. Data collection and learning phase

[0190] server

[0191] The server first collects instruction manuals and FAQs (frequently asked questions) for the information terminals (various electronic devices and content playback devices used by users) from the Internet and an internal database, and then centrally manages them. Since this data is used as learning material for the model, it is cleaned in advance to remove duplicate data and unnecessary information. Python-based scripts and data processing libraries (Pandas, NumPy, etc.) are used for data cleaning.

[0192] The server then trains the preprocessed data using a generative model (e.g., GPT-4), which lays the foundation for the model to respond quickly and accurately to user queries.

[0193] 2. Inquiry Receipt and Response Phase

[0194] User

[0195] Users access the "Content Help AI" app on their smartphone or head-mounted display (HMD) and input a query by voice or text, such as "What movie would you recommend I watch next?"

[0196] Terminal

[0197] The device (smartphone or HMD) converts the input voice data into text (using voice recognition software) and converts it into JSON format, which is then sent to the server.

[0198] server

[0199] The server analyzes the received query and passes it to the generative model, which generates an appropriate response from the database based on the prompt "Please provide information about recommended movies that the user should watch." For example, the prompt might look like this:

[0200] "What movie should users watch next? Create a recommendation list that takes into account recent popular movies and the user's viewing history."

[0201] After the generative model generates a response, the server sends the response back to the user device and displays it on the user's smartphone or HMD.

[0202] 3. Performance evaluation and improvement phase

[0203] server

[0204] The server records all queries and the generated responses in a log file, allowing the system's performance to be evaluated periodically. If performance falls below a predetermined standard, the server collects additional data and retrains the model.

[0205] User

[0206] Additionally, users can provide feedback on the answers they provide, which is also sent to the server and used as data for future model improvements.

[0207] As a concrete example, when a user asks, "What movie would you recommend I watch next?" the system follows these steps:

[0208] 1. The user uses the app to enter a question by voice.

[0209] 2. The device converts the voice into text and sends it to the server.

[0210] 3. The server inputs a prompt sentence into the generative model.

[0211] 4. The generative model generates a list of appropriate movie recommendations.

[0212] 5. The server generates a response and sends it back to the user, who displays it in the app.

[0213] This process allows users to get relevant and fast recommendations for content they might want to watch.

[0214] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0215] Step 1:

[0216] User enters a query

[0217] Users use the "Content Help AI" app on their smartphones or head-mounted displays (HMDs) to input questions about the content they want to watch. For example, they might say, "What movie would you recommend I watch next?" This input data is in audio format.

[0218] Step 2:

[0219] The device converts the voice data into text

[0220] The device uses speech recognition software to convert the voice data into text data. The input to this process is the user's voice data, and the output is the question in text form.

[0221] Step 3:

[0222] The device converts the text data into JSON format and sends it to the server.

[0223] The terminal converts text data into JSON format. In this conversion process, the text data is converted into a structured JSON object. The converted JSON data is sent to the server. The input of this process is text data, and the output is JSON format data.

[0224] Step 4:

[0225] The server receives and parses the JSON data.

[0226] The server receives and analyzes the JSON data sent from the terminal. In this analysis process, the JSON object is analyzed to extract the user's query. The input of this process is the JSON data, and the output is the extracted data of the query.

[0227] Step 5:

[0228] The server inputs a prompt to the generative model.

[0229] The server inputs a prompt sentence into the generative model based on the extracted query content. An example of a prompt sentence is, "Please tell me what movie the user should watch next. Please create a recommendation list taking into account recent popular movies and the user's viewing history." The inputs to this process are the extracted data and the prompt sentence, and the output is a response generated by the generative model.

[0230] Step 6:

[0231] The generative model generates the response

[0232] A generative model (e.g., GPT-4) generates a response based on the prompt sentence. This response generation process creates an appropriate recommendation list based on the viewing history and popular content stored in the database. The input of this process is the prompt sentence, and the output is the generated response data.

[0233] Step 7:

[0234] The server generates a response and sends it back to the user.

[0235] The server receives the response generated from the generative model and sends it back to the user terminal. In this process, the generated response data is converted into an appropriate format and sent to the terminal. The input of this process is the generated response data, and the output is the data sent to the user terminal.

[0236] Step 8:

[0237] The terminal displays the response data to the user

[0238] The device analyzes the response data received from the server and displays it to the user. This display process converts the generated response into a format that is easy for the user to understand and displays it within the app. The input of this process is the response data received from the server, and the output is the information displayed on the user's screen.

[0239] In this way, users can get appropriate and quick recommendations and operating procedures for the content they want to watch.

[0240] 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.

[0241] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. The overall flow of this system will be specifically explained from the perspectives of the server, terminal, and user.

[0242] 1. Data collection and learning phase

[0243] server

[0244] The server first collects instruction manuals and FAQ (frequently asked questions) data for all information devices it handles from the Internet and an internal database, and then centrally manages them. This data is used as learning material for the generative model. The collected data is pre-cleaned to remove duplicate data and unnecessary information. After data cleaning is complete, the data is stored in the database.

[0245] The server then trains the pre-processed data with a generative model using machine learning algorithms, which then serves as the basis for quickly and accurately responding to future queries.

[0246] 2. Query Receipt and Response Generation Phase

[0247] User

[0248] If a user has a question about how to operate the system or a specific function, they can enter a query into the system through a dedicated user interface, such as "How do I set up Wi-Fi?"

[0249] Terminal

[0250] User inquiries received via the user interface are sent by the terminal to the server. Specifically, the inquiry content is converted into JSON format or similar and sent to the server's API.

[0251] server

[0252] The server analyzes the received query, performs necessary preprocessing, and then passes it to the generative model, which generates an appropriate answer based on the query and returns the response to the server.

[0253] 3. Emotion recognition and response adaptation using an emotion engine

[0254] server

[0255] Before sending the generated response, the server passes the query content to an emotion engine to recognize the user's emotion. The emotion engine uses natural language processing technology to analyze emotions from the user's text data, and determines whether the user is feeling anger, sadness, joy, anticipation, or other emotions.

[0256] Based on the emotions recognized by the emotion engine, the server will adapt the tone and content of the response. For example, if the user is angry, the tone of the response will be more polite. It also includes the ability to escalate to a human operator if necessary.

[0257] 4. Sending the Response

[0258] server

[0259] Once the tailored response is complete, the server sends it to the terminal for transmission back to the user, using a secure protocol.

[0260] Terminal

[0261] The terminal displays the response received from the server on the user interface, allowing the user to perform operations based on the response.

[0262] 5. Performance evaluation and improvement

[0263] server

[0264] The server logs user queries and generated responses. This accumulates data for later performance evaluation and model improvement. It periodically analyzes the log data to evaluate the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[0265] User

[0266] The user inputs feedback on the provided responses and sends it to the server via the terminal, which helps improve the system.

[0267] Specific examples

[0268] For example, if a user asks, "How do I set up Wi-Fi?", the flow is as follows:

[0269] 1. User enters a query

[0270] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[0271] 2. The device sends a query to the server

[0272] The browser on the user's device sends this query to the server's API.

[0273] 3. The server receives the query and passes it to the generative model.

[0274] The server analyzes the received query and inputs it into the generative model.

[0275] 4. The generative model generates a response

[0276] The generative model retrieves information about "how to set up Wi-Fi" from a database and generates appropriate instructions.

[0277] 5. The server passes the query to the emotion engine

[0278] The server passes the query content to the emotion engine to recognize the user's emotion.

[0279] 6. Emotion engine recognizes user emotions

[0280] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[0281] 7. The server adjusts the response

[0282] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[0283] 8. The server sends the response to the device

[0284] The server sends the tailored response to the terminal for transmission back to the user.

[0285] 9. The device displays the response

[0286] The terminal displays the response on the user interface, and the user performs an operation based on the response.

[0287] This system allows users to receive quick and sensitive responses, freeing sales representatives from having to deal with inquiries about how to use the system, allowing them to focus on more important tasks.

[0288] The processing flow will be explained below.

[0289] Step 1:

[0290] The server collects the instruction manuals and FAQ data of the information terminals it handles and stores them in a database. This is done by scraping data from the internet and retrieving data from the company's internal database. The collected data is pre-processed to remove duplicate and irrelevant data.

[0291] Step 2:

[0292] The server trains a generative model based on the preprocessed data, using machine learning algorithms to analyze the data and train the model to generate appropriate responses to queries.

[0293] Step 3:

[0294] The server uses an emotion engine to train the system to recognize user emotions based on the text data contained in the user's inquiry. The emotion engine uses natural language processing technology to build an emotion analysis model.

[0295] Step 4:

[0296] The user inputs an inquiry about operation methods through a dedicated user interface. For example, the user inputs a specific question such as "How do I set up Wi-Fi?"

[0297] Step 5:

[0298] The terminal converts the query entered by the user into an appropriate format, such as JSON, and sends it to the server's API using the HTTPS protocol.

[0299] Step 6:

[0300] The server parses the received query and performs any necessary preprocessing, including tokenizing and encoding the text, before feeding the preprocessed data into the generative model.

[0301] Step 7:

[0302] The server receives the response from the generative model and inputs it into the emotion engine to analyze the user's emotion. The emotion engine recognizes the user's emotion and returns the result to the server.

[0303] Step 8:

[0304] The server adapts the tone and content of the response based on the results of the emotion engine. For example, if the user is angry, the tone of the response will be more polite. It will also escalate the situation to a human operator if necessary.

[0305] Step 9:

[0306] The server sends the tailored response to the terminal for transmission back to the user, using a secure communications protocol.

[0307] Step 10:

[0308] The terminal displays the response received from the server on the user interface, and the user checks the displayed response and performs an operation based on it.

[0309] Step 11:

[0310] Perform actions based on the responses the user provides, for example, stepping through Wi-Fi setup procedures to resolve the issue.

[0311] Step 12:

[0312] The server logs all user queries and generated responses, providing data for later performance evaluation and model improvement.

[0313] Step 13:

[0314] The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine, including response accuracy, emotion recognition precision, and response time.

[0315] Step 14:

[0316] If the server finds that the model's performance falls below a predetermined standard, it collects new data and retrains the model, using the most recent query data and user feedback.

[0317] Step 15:

[0318] The user inputs feedback on the provided responses and sends it to the server via the terminal, which is used to improve the performance of the system.

[0319] As a result, this system can reduce the burden on sales staff and improve customer satisfaction.

[0320] Example 2

[0321] 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."

[0322] In modern information systems, it is difficult for users to obtain prompt and appropriate responses when inquiring about operation methods or specific functions. Furthermore, responses that take emotion into consideration are not provided, resulting in a poor user experience. Furthermore, it is difficult to generate adaptive responses based on the content of the inquiry, and there is a need for improved user satisfaction. Therefore, a system that solves these issues and provides efficient and emotion-sensitive responses is needed.

[0323] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0324] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving inquiries entered by a user and passing them to the trained model, means for analyzing the content of the user's inquiry to generate a prompt, means for analyzing the content of the inquiry and performing necessary preprocessing, means for passing the generated response to an emotion engine to recognize the user's emotion, means for adapting the tone and content of the response based on the recognized emotion, means for escalating the response to a human operator as necessary, means for returning the response generated from the model to the user, and means for recording the inquiries and the generated responses in a log and periodically evaluating the performance of the model. This makes it possible to provide quick and appropriate responses that take the user's emotions into consideration and improve user satisfaction.

[0325] "Information handling terminal" refers to all electronic devices that users can use to operate and configure.

[0326] An "instruction manual" refers to a document that describes how to use and set up the information terminal, as well as precautions to take.

[0327] "FAQ data" refers to data that compiles answers to questions and inquiries frequently asked by users.

[0328] A "database" refers to a system for efficiently storing, searching, and managing large amounts of data.

[0329] A "generative model" refers to an algorithm or machine learning model that automatically generates sentences or responses based on data.

[0330] "User" refers to an individual or organization that uses the system or service.

[0331] "Inquiry" refers to a question or concern a user has about a system or service.

[0332] A "prompt" refers to an input sentence that prompts a generative model to generate an appropriate response.

[0333] An "emotion engine" refers to a system or technology that analyzes emotions from a user's text data and recognizes their emotional state.

[0334] An "operator" refers to a human worker in charge of operating and managing a system or service.

[0335] "Log" refers to data that records system operation history and events.

[0336] "Performance evaluation" refers to methods for evaluating the accuracy of systems and models, response time, and emotion recognition accuracy.

[0337] "Preprocessing" refers to the process of converting data into the required format or removing noise before inputting it into a generative model.

[0338] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. The overall flow of this system will be specifically explained from the perspectives of the server, terminal, and user.

[0339] Data collection and learning phase

[0340] server

[0341] The server first collects the user's manual and FAQ data for the information terminal from the internet or an internal database. Web scraping technology (e.g., BeautifulSoup or Selenium) is used for collection. The collected data is centrally managed in CSV or JSON format. Next, data cleaning is performed, using regular expressions and database queries (e.g., SQL) to remove duplicate data and unnecessary information. The cleaned data is stored in a dedicated database. The server then trains the preprocessed data using a generative model (e.g., GPT-3). Machine learning frameworks such as TensorFlow and PyTorch are used for training. The trained model serves as the basis for responding quickly and accurately to future inquiries.

[0342] Query Receipt and Response Generation Phase

[0343] User

[0344] Users enter a query through a dedicated web interface, for example, "How do I set up Wi-Fi?"

[0345] Terminal

[0346] The user's device uses JavaScript or Ajax to convert the inquiry content into JSON format and sends it to the server's API endpoint.

[0347] server

[0348] The server parses the received JSON-formatted query, performs any necessary preprocessing, and then passes it to the generative model. The generative model generates an appropriate answer based on the query content. For example, the generative model retrieves information about "how to set up Wi-Fi" from a database and generates the appropriate steps.

[0349] Emotion engine recognizes emotions and adapts responses

[0350] server

[0351] The generated response is passed to an emotion engine (e.g., Google Cloud Natural Language API) to analyze the user's emotions. Emotions such as anger, sadness, and joy are detected from the text data. The emotion engine returns the analysis results to the server. The server then adjusts the tone and content of the response based on the analysis results. For example, if the user expresses anger, the tone of the response needs to be changed to be more polite. It also includes a means to escalate the response to a human operator if necessary.

[0352] Sending a Response

[0353] server

[0354] The server then converts the adjusted response back into JSON format and sends it to the device using a secure protocol such as HTTPS.

[0355] Terminal

[0356] The terminal displays the received response on the user interface, and the user can proceed with the operation by checking the displayed response.

[0357] Performance evaluation and improvement

[0358] server

[0359] The server records user queries and generated responses in a log file. This accumulates data for later performance evaluation and model improvement. The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[0360] User

[0361] Users can input their satisfaction with the responses and feedback through a dedicated UI, which is then sent to the server via their device. This feedback is used to improve the system.

[0362] Specific examples

[0363] For example, if a user asks, "How do I set up Wi-Fi?", the specific operation is as follows.

[0364] 1. User enters a query

[0365] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[0366] 2. The device sends a query to the server

[0367] This query is converted to JSON format using JavaScript and sent to the server's API endpoint via Ajax.

[0368] 3. The server receives and analyzes the query

[0369] Analyze the received JSON data using Python and perform the necessary preprocessing.

[0370] 4. The generative model generates a response

[0371] The preprocessed data is fed into a generative model (e.g., GPT-3) to obtain information about "how to set up Wi-Fi" and generate appropriate instructions.

[0372] 5. The server passes the query to the emotion engine

[0373] The generated response is passed to an emotion engine to recognize the user's emotion.

[0374] 6. Emotion engine recognizes user emotions

[0375] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[0376] 7. The server adjusts the response

[0377] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[0378] 8. The server sends the response to the device

[0379] The adjusted response is converted back to JSON format and sent to the terminal.

[0380] 9. The device displays the response

[0381] The received response is displayed in the user interface, and the user performs an action based on it.

[0382] Prompt Sentence Examples

[0383] "Please tell me how to set up Wi-Fi. And please provide appropriate emotion-aware responses to my queries."

[0384] This system allows users to receive prompt and sensitive answers, eliminating any concerns or questions they may have about the system.

[0385] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0386] Program processing steps

[0387] Step 1: Collect and store data

[0388] server

[0389] The server collects instruction manuals and FAQ data from the internet and internal databases. Specifically, it uses web scraping technology (e.g., BeautifulSoup or Selenium) to extract the necessary data from HTML pages such as instruction manuals. The collected data is saved in CSV or JSON format. It is then stored in a database so that it can be used for later processing.

[0390] Input: instruction manual, FAQ data

[0391] Output: Clean data stored in a database

[0392] Step 2: Cleaning the data

[0393] server

[0394] The server cleans the collected data, using regular expressions and database queries (e.g., SQL) to remove duplicate data and unnecessary information. The cleaned data is then stored in a dedicated database. This prepares the data for use in training the generative model.

[0395] Input: Collected data in the database

[0396] Output: Data in a clean database

[0397] Step 3: Train the model

[0398] server

[0399] The server trains a generative model (e.g., GPT-3) on clean data using a machine learning framework such as TensorFlow or PyTorch. The trained model then becomes the basis for responding to queries.

[0400] Input: Data in a clean database

[0401] Output: A trained generative model

[0402] Step 4: Receiving a user query

[0403] User

[0404] Users enter queries through a dedicated web interface, for example, "How do I set up Wi-Fi?"

[0405] Terminal

[0406] The terminal uses JavaScript or Ajax to convert the query content into JSON format and send it to the server's API endpoint.

[0407] Input: User's inquiry

[0408] Output: Query sent to the server in JSON format

[0409] Step 5: Parsing and preprocessing the query

[0410] server

[0411] The server parses the received JSON query and converts it into an appropriate format. Specifically, it uses Python to parse the JSON data and performs the necessary preprocessing before passing it to the generative model.

[0412] Input: JSON format query

[0413] Output: Preprocessed query data

[0414] Step 6: Generate a response

[0415] server

[0416] The server then passes the preprocessed query to the generative model, which then generates the appropriate response. For example, the generative model retrieves information about "how to set up Wi-Fi" from a database and generates the appropriate instructions.

[0417] Input: Preprocessed query data

[0418] Output: The generated response

[0419] Step 7: Emotion Recognition

[0420] server

[0421] The generated response is passed to an emotion engine (e.g., Google Cloud Natural Language API) to analyze the user's emotions. Emotions such as anger, sadness, and joy are detected from the text data. The emotion engine returns the analysis results to the server.

[0422] Input: Generated response

[0423] Output: User sentiment analysis results

[0424] Step 8: Adjusting the response

[0425] server

[0426] The server adjusts the tone and content of the response based on the results of the emotion engine. For example, if the user is expressing anger, the tone of the response should be changed to be more polite. It also includes a means to escalate the response to a human operator if necessary.

[0427] Input: User sentiment analysis results

[0428] Output: Adjusted response

[0429] Step 9: Sending a Response

[0430] server

[0431] The server then converts the adjusted response back into JSON format and sends it to the device using a secure protocol such as HTTPS.

[0432] Input: Adjusted response

[0433] Output: JSON formatted response sent to the terminal

[0434] Step 10: View the response

[0435] Terminal

[0436] The terminal displays the received response on the user interface, allowing the user to confirm the displayed response and proceed with the operation.

[0437] Input: The response sent by the server

[0438] Output: The response displayed in the user interface

[0439] Step 11: Logging queries and responses

[0440] server

[0441] The server records user queries and generated responses in a log file, which provides data for later performance evaluation and model improvement.

[0442] Input: The query and generated response

[0443] Output: Data recorded in the log file

[0444] Step 12: Evaluate the model's performance

[0445] server

[0446] The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine, including response accuracy, response time, and emotion recognition accuracy. If necessary, it collects new data and retrains the model.

[0447] Input: Log data

[0448] Output: Performance evaluation results and retrained model

[0449] Step 13: Gather user feedback

[0450] User

[0451] Users input their satisfaction and feedback regarding the provided responses through a dedicated UI and send it to the server via their device.

[0452] server

[0453] The server analyzes the collected feedback and uses it to improve the system.

[0454] Input: User feedback

[0455] Output: Improved system

[0456] Through these steps, users can receive quick and emotionally sensitive answers.

[0457] (Application example 2)

[0458] 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."

[0459] Conventional content distribution services have the problem of reducing user satisfaction because they recommend content without considering the user's emotional state. Furthermore, they often return a uniform response to user inquiries, failing to consider the user's emotions. This results in issues such as a decrease in service usage and an increased likelihood that users will switch to other services.

[0460] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0461] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving inquiries entered by users and passing them to the trained model, means for returning responses generated from the model to the users, means for recording the inquiries and the generated responses and periodically evaluating the performance of the model, emotion recognition means for analyzing emotions from the user's inquiry text and adjusting the response content, and means for recommending appropriate content according to the user's emotions in a content distribution service, thereby enabling responses and content recommendations that take user emotions into consideration.

[0462] "Handling information terminal" refers to a device used by a user to perform operations or make inquiries.

[0463] An "instruction manual" is a document or material that explains how to use an information terminal and its functions.

[0464] "FAQ data" is data that compiles information about questions and doubts that users frequently have.

[0465] A "generative model" is an algorithm or machine learning model that generates responses to queries based on collected data.

[0466] "Training means" refers to the method or process for training a generative model using collected data.

[0467] A "user query" is a question or request made by a user to a system for information or support.

[0468] A "response" is an answer or instruction that a generative model generates in response to a user query.

[0469] A "database" is a system for systematically storing and managing collected instruction manuals and frequently asked questions.

[0470] "Emotion recognition means" refers to the technology or function for analyzing and extracting emotions from the user's inquiry text.

[0471] A "content distribution service" is a service that provides users with content such as videos, music, and articles.

[0472] "Feedback means" refers to a method or function for collecting user evaluations and opinions on responses and using them to improve the system.

[0473] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. This system performs the following processes from the perspectives of the server, terminal, and user.

[0474] 1. Data collection and learning phase

[0475] server

[0476] First, the server collects instruction manuals and frequently asked questions data and stores them in a database. The collected data is pre-cleaned to remove duplicate data and unnecessary information. After data cleaning is complete, the data is stored in the database. Next, the server analyzes the pre-processed data using a generative model to train the model. A specific machine learning algorithm is used for this training. Once the model is trained, it serves as the basis for responding quickly and accurately to future inquiries.

[0477] 2. Query Receipt and Response Generation Phase

[0478] User

[0479] If a user has a question about how to operate the system or a specific function, they can input a query into the system through a dedicated user interface, such as "What movies do you recommend?"

[0480] Terminal

[0481] User inquiries received via the user interface are sent by the terminal to the server. Specifically, the inquiry content is converted into JSON format or similar and sent to the server's API.

[0482] server

[0483] The server analyzes the received query, performs necessary preprocessing, and then passes it to the generative model, which generates an appropriate answer based on the query and returns the response to the server.

[0484] 3. Emotion recognition and response adaptation using an emotion engine

[0485] server

[0486] Before sending the generated response, the server passes the query content to an emotion engine to recognize the user's emotion. The emotion engine uses natural language processing technology to analyze emotions from the user's text data. As a result, it determines whether the user is feeling anger, sadness, joy, anticipation, or other emotions. Based on the emotion recognized by the emotion engine, the server adapts the tone and content of the response. For example, if the user is expressing anger, the tone of the response will be adjusted to be more polite. It also includes a function to escalate to a human operator if necessary.

[0487] 4. Sending the Response

[0488] server

[0489] Once the tailored response is complete, the server sends it to the terminal for transmission back to the user, using a secure protocol.

[0490] Terminal

[0491] The terminal displays the response received from the server on the user interface, allowing the user to perform operations based on the response.

[0492] 5. Performance evaluation and improvement

[0493] server

[0494] The server logs user queries and generated responses. This accumulates data for later performance evaluation and model improvement. It periodically analyzes the log data to evaluate the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[0495] User

[0496] The user inputs feedback on the provided responses and sends it to the server via the terminal, which helps improve the system.

[0497] Specific examples

[0498] For example, if a user inquires, "What movies do you recommend?", the flow is as follows:

[0499] 1. User enters a query

[0500] The user enters a question.

[0501] 2. The device sends a query to the server

[0502] The user terminal sends this query to the server's API.

[0503] 3. The server receives the query and passes it to the generative model.

[0504] The server analyzes the received query and inputs it into the generative model.

[0505] 4. The generative model generates a response

[0506] The generative model retrieves information about "movie recommendation" from the database and generates appropriate steps.

[0507] 5. The server passes the query to the emotion engine

[0508] The server passes the query content to the emotion engine to recognize the user's emotion.

[0509] 6. Emotion engine recognizes user emotions

[0510] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[0511] 7. The server adjusts the response

[0512] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[0513] 8. The server sends the response to the device

[0514] The server sends the tailored response to the terminal for transmission back to the user.

[0515] 9. The device displays the response

[0516] The terminal displays the response on the user interface, and the user performs an operation based on the response.

[0517] This system allows users to receive prompt and emotionally sensitive answers, improving their satisfaction with the service.

[0518] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0519] Step 1:

[0520] User enters a query

[0521] A user types a question into a smartphone application, for example, "What movies do you recommend?" This query is entered through a user interface, and the entered text is stored on the device as raw text.

[0522] Input: User query text (e.g., "What movies do you recommend?")

[0523] Output: Raw query text

[0524] Step 2:

[0525] The device sends a query to the server

[0526] The terminal converts user inquiries received via the user interface into JSON format and sends it to the server's API, allowing the server to receive the inquiry content.

[0527] Input: Raw query text

[0528] Output: Data converted to JSON format (e.g., {"query": "What movies do you recommend?"})

[0529] Step 3:

[0530] The server receives the query and passes it to the generative model.

[0531] The server parses the received JSON data and extracts the query content, which is then passed to the generative model, which then generates an appropriate response based on this data.

[0532] Input: Query data in JSON format

[0533] Output: Generated response (e.g., "I recommend 'Hero Movie' as an uplifting action movie.")

[0534] Step 4:

[0535] The generative model generates the response

[0536] The generative model generates an appropriate answer based on the query. This process involves retrieving relevant information from a database and performing data processing and calculations to generate the appropriate response.

[0537] Input: The query passed to the generative model

[0538] Output: Correct response

[0539] Step 5:

[0540] The server passes the query to the emotion engine

[0541] The server passes the generated response and the original query text to the emotion engine to recognize the user's emotion. The emotion engine uses natural language processing techniques to analyze the user's emotional state from the text.

[0542] Input: Generated response and original query text

[0543] Output: Emotion recognition result (e.g. positive, negative, neutral)

[0544] Step 6:

[0545] Emotion engine recognizes user emotions

[0546] The emotion engine analyzes the user's emotion from the input text and returns the state to the server. In this step, data processing and calculations are performed to extract the user's emotion using an analysis algorithm and return the result.

[0547] Input: Query text and generated response

[0548] Output: Emotion recognition result (e.g., positive)

[0549] Step 7:

[0550] The server adjusts the response

[0551] Based on the results of the emotion engine, the server adjusts the generated response, ensuring that the response is of the appropriate tone and content depending on the user's emotional state, and escalating to a human operator if necessary.

[0552] Input: Generated response sentences and emotion recognition results

[0553] Output: Tailored response (e.g., "I recommend 'Hero Movie' as an uplifting action movie. I'm glad you liked it!")

[0554] Step 8:

[0555] The server sends a response to the device

[0556] The server sends the tailored response to the terminal for transmission back to the user, using a secure protocol.

[0557] Input: Adjusted response sentence

[0558] Output: Response data sent to the terminal

[0559] Step 9:

[0560] The terminal displays the response

[0561] The terminal displays the response received from the server on the user interface, allowing the user to perform operations and make selections based on the response.

[0562] Input: Response data sent from the server

[0563] Output: The response text displayed in the user interface

[0564] 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.

[0565] 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.

[0566] 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.

[0567] [Second embodiment]

[0568] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0569] 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.

[0570] 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).

[0571] 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.

[0572] 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.

[0573] 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).

[0574] 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.

[0575] 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.

[0576] 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.

[0577] 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.

[0578] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0579] 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."

[0580] The present invention is an operation method support system that uses a generative model to provide convenience to both salespeople and customers. The overall flow of this system will be explained from the perspectives of the server, terminals, and users.

[0581] 1. Data collection and learning phase

[0582] server

[0583] The server first collects instruction manuals and FAQ (frequently asked questions) data for all information devices handled from the Internet and internal databases, and then centrally manages them. Since this data is used as learning material for the model, it is cleaned in advance to remove duplicate data and unnecessary information. After data cleaning is complete, the data is organized and saved.

[0584] The server then trains the pre-processed data with a generative model using machine learning algorithms, which then serves as the basis for quickly and accurately responding to future queries.

[0585] 2. Inquiry Receipt and Response Phase

[0586] User

[0587] When a user has a question about an operation method or a specific function, they input their query into the system through a dedicated user interface, which may be provided in the form of a web browser, a mobile application, or a help desk.

[0588] Terminal

[0589] User inquiries received via the user interface are sent from the terminal to the server. The terminal converts the inquiry content into a format such as JSON and sends it to the server's data reception API.

[0590] server

[0591] The server analyzes the received query and passes it to a trained generative model. The generative model generates an appropriate answer based on the query, constructs a response, and returns it to the server. The server then sends this response back to the user, allowing the user to receive a quick and accurate answer.

[0592] 3. Performance evaluation and improvement phase

[0593] server

[0594] The server records all queries and responses in log files. Periodically, it analyzes these logs to evaluate the model's performance. This evaluation involves scoring the generated responses to ensure their accuracy. Based on the evaluation results, if the model's accuracy falls below a predetermined standard, it may automatically initiate a process to collect additional data and retrain.

[0595] User

[0596] Additionally, users can enter feedback on the answers they provide, which is sent to the server and used as data for future model improvements.

[0597] Specific examples

[0598] For example, suppose a user asks, "How do I set up Wi-Fi?" The flow is shown below.

[0599] 1. User enters a query

[0600] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[0601] 2. The device sends a query to the server

[0602] The browser on the user's device sends this query to the server's API.

[0603] 3. The server receives the query and passes it to the generative model.

[0604] The server analyzes the received query and inputs it into the generative model.

[0605] 4. The generative model generates a response

[0606] The generative model retrieves information about "how to set up Wi-Fi" from a database and generates appropriate instructions.

[0607] 5. The server sends the response back to the user

[0608] The server generates a response and sends it back to the user, who displays it in their browser.

[0609] 6. The user receives the response and performs the action

[0610] The user follows the displayed steps to set up Wi-Fi.

[0611] This allows users to receive prompt and appropriate responses, and frees sales representatives from having to deal with inquiries about operation methods, allowing them to focus on more important tasks.

[0612] The processing flow will be explained below.

[0613] Step 1:

[0614] The server collects instruction manuals and FAQ data from the information terminals it handles and stores them in a database, including scraping data from the Internet and retrieving it from its internal database.

[0615] Step 2:

[0616] The server preprocesses the collected data, specifically by performing text cleaning and data normalization to remove duplicate data and unnecessary information.

[0617] Step 3:

[0618] The server trains a generative model based on the preprocessed data, using machine learning algorithms to extract the necessary knowledge from the data and train the model.

[0619] Step 4:

[0620] The server stores the trained generative model and prepares it for generating responses to queries.

[0621] Step 5:

[0622] The user inputs an inquiry about operation methods through a dedicated user interface, for example, "Please tell me how to set up Wi-Fi."

[0623] Step 6:

[0624] The terminal converts the user's inquiry into JSON format or similar and sends it to the server's API. This transmission generally uses the HTTPS protocol.

[0625] Step 7:

[0626] The server parses the received query, performs any necessary preprocessing, and then passes it to the generative model. Parsing includes tokenizing and encoding the text.

[0627] Step 8:

[0628] The server receives the response from the generative model and transforms it into a format that is easy for the user to understand, for example, organizing the steps as a numbered list.

[0629] Step 9:

[0630] The server sends the converted response back to the user via the API to the device, again using a secure protocol.

[0631] Step 10:

[0632] The device displays the response received from the server and provides it to the user, either on a web page or within an application.

[0633] Step 11:

[0634] Performs actions based on the answers the user provides, for example, walking through Wi-Fi setup steps to resolve the issue.

[0635] Step 12:

[0636] The server logs user queries and the responses it generates, providing data for later performance evaluation and model improvement.

[0637] Step 13:

[0638] The server periodically analyzes the log data and evaluates the performance of the generative model using metrics such as accuracy and response time.

[0639] Step 14:

[0640] If the accuracy of the model falls below a predetermined standard, the server collects new data and retrains the model, ensuring that it always provides highly accurate responses.

[0641] Step 15:

[0642] The user inputs feedback on the provided responses and sends it to the server via the terminal, which is used to improve the system.

[0643] As a result, this system can reduce the burden on sales staff and improve customer satisfaction.

[0644] Example 1

[0645] 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."

[0646] Many modern information terminals have complex operation methods and multifunctional settings, which often lead to users becoming confused and not receiving appropriate support. Sales representatives are also overwhelmed by responding to a variety of inquiries, preventing them from concentrating on their core duties. Furthermore, traditional FAQs and instruction manuals only provide static information, making it difficult to provide dynamic and prompt support.

[0647] 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.

[0648] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for cleaning and preprocessing the collected data, means for training the preprocessed data using a generative model, means for receiving inquiries entered by users, converting the inquiry content into JSON format, and sending the JSON format to the server, means for inputting the user's inquiry into the trained generative model and obtaining a response generated from the model, means for returning the obtained response to the user interface, and means for recording the inquiries and the generated responses and periodically evaluating the performance of the model. This allows users to receive prompt and appropriate support, and frees sales representatives from responding to inquiries so that they can focus on their core business.

[0649] An "information terminal" is a device or system that provides information about an electronic device or system operated by a user.

[0650] An "instruction manual" is a document that describes the basic operation methods, setting procedures, precautions, etc. of an information terminal.

[0651] "Frequently asked questions data" is data that compiles questions frequently asked by users and the answers to those questions.

[0652] A "database" is a system for efficiently storing, managing, and searching large amounts of data.

[0653] "Data cleanup" is the process of removing redundancies and noise from collected data and preparing it in a format suitable for analysis.

[0654] "Preprocessing" is a series of steps to convert data into a format suitable for a generative model.

[0655] A "generative model" is a model that uses machine learning algorithms to generate optimal answers from given data.

[0656] "Training" is the process of optimizing the parameters of a generative model using collected data to train the model.

[0657] A "user interface" is an interactive screen or application that allows a user to access and query a system.

[0658] "JSON format" is a data format that structures data in text format and makes it easily exchangeable.

[0659] A "server" is a computer system that processes data and provides services over a network.

[0660] An "inquiry" is a question or request made by a user to the system.

[0661] A "response" is an answer that a system provides to a user's query.

[0662] "Recording" means saving the contents of inquiries and responses in a log file or the like.

[0663] "Performance evaluation" is the process of evaluating the accuracy and appropriateness of the responses of a generative model.

[0664] This invention is an operation method support system that uses a generative AI model to provide convenience to both sales representatives and customers. In this system, the server, terminal, and user each play specific roles, and provide quick and accurate answers to user inquiries.

[0665] Data collection and preprocessing

[0666] server

[0667] The server collects the user manuals and FAQ data for the information terminals from the internet and an internal database. It also cleans the data, removing duplicate data and noise. Specifically, it uses Python scripts and BeautifulSoup to scrape the web, and the Pandas library to process the data. The cleaned data is then stored using a large-scale data processing framework (e.g., Hadoop).

[0668] Training generative AI models

[0669] server

[0670] The server trains a generative model using the collected and preprocessed data. TensorFlow or PyTorch is used for this training. The training data is split using a tokenizer and input to a neural network. A deep learning framework is used to optimize the model parameters. In particular, the Transformers architecture is commonly used.

[0671] Receiving inquiries via a user interface

[0672] User

[0673] Users input questions about operations and specific functions through a user interface, which is provided by a web browser or a mobile application, such as Google Chrome or Safari.

[0674] Specific examples

[0675] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[0676] Analysis of inquiries

[0677] Terminal

[0678] The device converts the user's inquiry into JSON format and sends it to the server's receiving API. For example, it generates JSON data using a JavaScript or Python library and sends it to the server via an HTTP request. Browsers used include Firefox and Google Chrome.

[0679] Answer generation using generative AI models

[0680] server

[0681] The server inputs the received query into a generative model. This process involves encoding the query and converting it into a format suitable for the generative model. The generative AI model (e.g., GPT-3) generates an appropriate answer based on the query. The generated answer is then formatted.

[0682] Specific examples

[0683] The server uses the generative model to generate instructions on how to set up Wi-Fi.

[0684] Returning answers to users

[0685] server

[0686] The server generates an answer and sends it back to the user. The response data is organized in JSON format and sent to the device in an HTTP response.

[0687] Terminal

[0688] The device analyzes the returned data and displays it in a user interface, using HTML and JavaScript to display the answers in the browser.

[0689] Performance evaluation and feedback collection

[0690] server

[0691] The server logs all queries and responses and periodically analyzes them. The log data is evaluated using Python and R to measure the accuracy of the responses. Users can enter feedback on the answers provided, and this data is also collected.

[0692] This process allows users to receive prompt and appropriate answers, and frees up sales representatives from the burden of handling inquiries. The system continuously collects data to improve the model's performance, providing a high-quality user experience.

[0693] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0694] Step 1:

[0695] Data collection and preprocessing

[0696] The server collects the user manuals and frequently asked questions (FAQ data) for the information terminals from the internet and an internal database. The input uses the URL of the online user manual and the access information for the internal database. Specifically, it performs web scraping using a Python script and BeautifulSoup to extract the data. After the data is collected, it is formatted using Pandas to remove duplicate data and noise. The clean data is generated as output and stored in Hadoop or another storage system.

[0697] Step 2:

[0698] Training generative AI models

[0699] The server trains a generative model using the collected and preprocessed data. It uses clean training data as input. Specifically, it tokenizes the training data using the TensorFlow or PyTorch framework and feeds it to a neural network. The selected deep learning model (e.g., Transformers architecture) learns patterns in the data and finds optimal parameters. The output is a trained generative model.

[0700] Step 3:

[0701] Receiving inquiries via a user interface

[0702] The user inputs questions about operation methods or specific functions through the user interface. The user's question is used as input. Specifically, the user enters a question into an inquiry form in a web browser such as Google Chrome or Safari and clicks the submit button. The user's question data is sent to the device as output.

[0703] Step 4:

[0704] Analysis of inquiries

[0705] The terminal converts the user's inquiry into JSON format and sends it to the server's receiving API. The user's question received from the web browser is used as input. Specifically, it generates JSON data using JavaScript or Python libraries and sends it to the server via an HTTP request. The JSON-formatted inquiry data is sent to the server as output.

[0706] Step 5:

[0707] Answer generation using generative AI models

[0708] The server inputs the received query content into the generative model. Parsed JSON data is used as input. Specifically, the server encodes the query content and converts it into a format suitable for the generative model. The generative AI model (e.g., GPT-3) generates an appropriate answer based on the query content. The output is the generated response text.

[0709] Step 6:

[0710] Returning answers to users

[0711] The server returns the generated answer to the user. The generated response text is used as input. Specifically, the server formats the response data in JSON format and sends it to the terminal as an HTTP response. The terminal receives the returned data and displays it in the user interface. As output, the final response is displayed on the user's browser.

[0712] Step 7:

[0713] Performance evaluation and feedback collection

[0714] The server logs all queries and responses and periodically analyzes them. The query and response log data is used as input. Specific operations include evaluating the log data using Python or R to measure the model's performance. Users can provide feedback on the answers provided, and this data is also collected. The output is the model's performance evaluation results and user feedback.

[0715] This processing step allows the system to respond quickly and accurately to user inquiries and to continually collect data to improve performance.

[0716] (Application example 1)

[0717] 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."

[0718] Conventional systems have struggled to respond quickly and accurately to user inquiries about content search and playback settings. Content distribution services, in particular, require appropriate recommendations based on users' viewing history and recent popular content, but there has been a lack of efficient ways to provide this information. Furthermore, systems lack the ability to automatically update data and retrain when system performance deteriorates.

[0719] 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.

[0720] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving queries entered by users and passing them to the trained model, means for returning responses generated from the model to the users, means for recording queries and the generated responses and periodically evaluating the performance of the model, and means for generating content recommendations based on the user's viewing history and recent popular content in content distribution, thereby enabling users to receive appropriate recommendations and quick responses regarding content they want to view.

[0721] The term "information terminals" refers to various electronic devices and content playback terminals used by users.

[0722] "Instruction manual" refers to a document that contains detailed information about the operation and functions of the information terminal.

[0723] "Frequently asked questions data" refers to documents or data that systematically compile answers to questions and problems that users frequently have.

[0724] A "database" refers to a system for efficiently storing, searching, and managing collected information and data.

[0725] "Generative model" refers to a system trained using machine learning algorithms to generate appropriate responses to user queries.

[0726] "Training" refers to the process of using collected data to train a generative model and improve its accuracy.

[0727] "Content distribution" refers to the service of providing digital media such as movies, TV shows, and music to users via the Internet.

[0728] "Viewing history" refers to a record of content that a user has viewed in the past.

[0729] "Popular content" refers to digital media that is viewed by many users and has received high ratings.

[0730] "Content recommendation" refers to suggesting the next piece of media to watch based on the user's viewing history and popular content.

[0731] This invention applies a content distribution service to an operation method support system that uses a generative model to provide convenience to both sales representatives and customers. The details of this system will be explained in three phases: data collection and learning phase, inquiry reception and response phase, and performance evaluation and improvement phase.

[0732] 1. Data collection and learning phase

[0733] server

[0734] The server first collects instruction manuals and FAQs (frequently asked questions) for the information terminals (various electronic devices and content playback devices used by users) from the Internet and an internal database, and then centrally manages them. Since this data is used as learning material for the model, it is cleaned in advance to remove duplicate data and unnecessary information. Python-based scripts and data processing libraries (Pandas, NumPy, etc.) are used for data cleaning.

[0735] The server then trains the preprocessed data using a generative model (e.g., GPT-4), which lays the foundation for the model to respond quickly and accurately to user queries.

[0736] 2. Inquiry Receipt and Response Phase

[0737] User

[0738] Users access the "Content Help AI" app on their smartphone or head-mounted display (HMD) and input a query by voice or text, such as "What movie would you recommend I watch next?"

[0739] Terminal

[0740] The device (smartphone or HMD) converts the input voice data into text (using voice recognition software) and converts it into JSON format, which is then sent to the server.

[0741] server

[0742] The server analyzes the received query and passes it to the generative model, which generates an appropriate response from the database based on the prompt "Please provide information about recommended movies that the user should watch." For example, the prompt might look like this:

[0743] "What movie should users watch next? Create a recommendation list that takes into account recent popular movies and the user's viewing history."

[0744] After the generative model generates a response, the server sends the response back to the user device and displays it on the user's smartphone or HMD.

[0745] 3. Performance evaluation and improvement phase

[0746] server

[0747] The server records all queries and the generated responses in a log file, allowing the system's performance to be evaluated periodically. If performance falls below a predetermined standard, the server collects additional data and retrains the model.

[0748] User

[0749] Additionally, users can provide feedback on the answers they provide, which is also sent to the server and used as data for future model improvements.

[0750] As a concrete example, when a user asks, "What movie would you recommend I watch next?" the system follows these steps:

[0751] 1. The user uses the app to enter a question by voice.

[0752] 2. The device converts the voice into text and sends it to the server.

[0753] 3. The server inputs a prompt sentence into the generative model.

[0754] 4. The generative model generates a list of appropriate movie recommendations.

[0755] 5. The server generates a response and sends it back to the user, who displays it in the app.

[0756] This process allows users to get relevant and fast recommendations for content they might want to watch.

[0757] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0758] Step 1:

[0759] User enters a query

[0760] Users use the "Content Help AI" app on their smartphones or head-mounted displays (HMDs) to input questions about the content they want to watch. For example, they might say, "What movie would you recommend I watch next?" This input data is in audio format.

[0761] Step 2:

[0762] The device converts the voice data into text

[0763] The device uses speech recognition software to convert the voice data into text data. The input to this process is the user's voice data, and the output is the question in text form.

[0764] Step 3:

[0765] The device converts the text data into JSON format and sends it to the server.

[0766] The terminal converts text data into JSON format. In this conversion process, the text data is converted into a structured JSON object. The converted JSON data is sent to the server. The input of this process is text data, and the output is JSON format data.

[0767] Step 4:

[0768] The server receives and parses the JSON data.

[0769] The server receives and analyzes the JSON data sent from the terminal. In this analysis process, the JSON object is analyzed to extract the user's query. The input of this process is the JSON data, and the output is the extracted data of the query.

[0770] Step 5:

[0771] The server inputs a prompt to the generative model.

[0772] The server inputs a prompt sentence into the generative model based on the extracted query content. An example of a prompt sentence is, "Please tell me what movie the user should watch next. Please create a recommendation list taking into account recent popular movies and the user's viewing history." The inputs to this process are the extracted data and the prompt sentence, and the output is a response generated by the generative model.

[0773] Step 6:

[0774] The generative model generates the response

[0775] A generative model (e.g., GPT-4) generates a response based on the prompt sentence. This response generation process creates an appropriate recommendation list based on the viewing history and popular content stored in the database. The input of this process is the prompt sentence, and the output is the generated response data.

[0776] Step 7:

[0777] The server generates a response and sends it back to the user.

[0778] The server receives the response generated from the generative model and sends it back to the user terminal. In this process, the generated response data is converted into an appropriate format and sent to the terminal. The input of this process is the generated response data, and the output is the data sent to the user terminal.

[0779] Step 8:

[0780] The terminal displays the response data to the user

[0781] The device analyzes the response data received from the server and displays it to the user. This display process converts the generated response into a format that is easy for the user to understand and displays it within the app. The input of this process is the response data received from the server, and the output is the information displayed on the user's screen.

[0782] In this way, users can get appropriate and quick recommendations and operating procedures for the content they want to watch.

[0783] 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.

[0784] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. The overall flow of this system will be specifically explained from the perspectives of the server, terminal, and user.

[0785] 1. Data collection and learning phase

[0786] server

[0787] The server first collects instruction manuals and FAQ (frequently asked questions) data for all information devices it handles from the Internet and an internal database, and then centrally manages them. This data is used as learning material for the generative model. The collected data is pre-cleaned to remove duplicate data and unnecessary information. After data cleaning is complete, the data is stored in the database.

[0788] The server then trains the pre-processed data with a generative model using machine learning algorithms, which then serves as the basis for quickly and accurately responding to future queries.

[0789] 2. Query Receipt and Response Generation Phase

[0790] User

[0791] If a user has a question about how to operate the system or a specific function, they can enter a query into the system through a dedicated user interface, such as "How do I set up Wi-Fi?"

[0792] Terminal

[0793] User inquiries received via the user interface are sent by the terminal to the server. Specifically, the inquiry content is converted into JSON format or similar and sent to the server's API.

[0794] server

[0795] The server analyzes the received query, performs necessary preprocessing, and then passes it to the generative model, which generates an appropriate answer based on the query and returns the response to the server.

[0796] 3. Emotion recognition and response adaptation using an emotion engine

[0797] server

[0798] Before sending the generated response, the server passes the query content to an emotion engine to recognize the user's emotion. The emotion engine uses natural language processing technology to analyze emotions from the user's text data, and determines whether the user is feeling anger, sadness, joy, anticipation, or other emotions.

[0799] Based on the emotions recognized by the emotion engine, the server will adapt the tone and content of the response. For example, if the user is angry, the tone of the response will be more polite. It also includes the ability to escalate to a human operator if necessary.

[0800] 4. Sending the Response

[0801] server

[0802] Once the tailored response is complete, the server sends it to the terminal for transmission back to the user, using a secure protocol.

[0803] Terminal

[0804] The terminal displays the response received from the server on the user interface, allowing the user to perform operations based on the response.

[0805] 5. Performance evaluation and improvement

[0806] server

[0807] The server logs user queries and generated responses. This accumulates data for later performance evaluation and model improvement. It periodically analyzes the log data to evaluate the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[0808] User

[0809] The user inputs feedback on the provided responses and sends it to the server via the terminal, which helps improve the system.

[0810] Specific examples

[0811] For example, if a user asks, "How do I set up Wi-Fi?", the flow is as follows:

[0812] 1. User enters a query

[0813] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[0814] 2. The device sends a query to the server

[0815] The browser on the user's device sends this query to the server's API.

[0816] 3. The server receives the query and passes it to the generative model.

[0817] The server analyzes the received query and inputs it into the generative model.

[0818] 4. The generative model generates a response

[0819] The generative model retrieves information about "how to set up Wi-Fi" from a database and generates appropriate instructions.

[0820] 5. The server passes the query to the emotion engine

[0821] The server passes the query content to the emotion engine to recognize the user's emotion.

[0822] 6. Emotion engine recognizes user emotions

[0823] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[0824] 7. The server adjusts the response

[0825] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[0826] 8. The server sends the response to the device

[0827] The server sends the tailored response to the terminal for transmission back to the user.

[0828] 9. The device displays the response

[0829] The terminal displays the response on the user interface, and the user performs an operation based on the response.

[0830] This system allows users to receive quick and sensitive responses, freeing sales representatives from having to deal with inquiries about how to use the system, allowing them to focus on more important tasks.

[0831] The processing flow will be explained below.

[0832] Step 1:

[0833] The server collects the instruction manuals and FAQ data of the information terminals it handles and stores them in a database. This is done by scraping data from the internet and retrieving data from the company's internal database. The collected data is pre-processed to remove duplicate and irrelevant data.

[0834] Step 2:

[0835] The server trains a generative model based on the preprocessed data, using machine learning algorithms to analyze the data and train the model to generate appropriate responses to queries.

[0836] Step 3:

[0837] The server uses an emotion engine to train the system to recognize user emotions based on the text data contained in the user's inquiry. The emotion engine uses natural language processing technology to build an emotion analysis model.

[0838] Step 4:

[0839] The user inputs an inquiry about operation methods through a dedicated user interface. For example, the user inputs a specific question such as "How do I set up Wi-Fi?"

[0840] Step 5:

[0841] The terminal converts the query entered by the user into an appropriate format, such as JSON, and sends it to the server's API using the HTTPS protocol.

[0842] Step 6:

[0843] The server parses the received query and performs any necessary preprocessing, including tokenizing and encoding the text, before feeding the preprocessed data into the generative model.

[0844] Step 7:

[0845] The server receives the response from the generative model and inputs it into the emotion engine to analyze the user's emotion. The emotion engine recognizes the user's emotion and returns the result to the server.

[0846] Step 8:

[0847] The server adapts the tone and content of the response based on the results of the emotion engine. For example, if the user is angry, the tone of the response will be more polite. It will also escalate the situation to a human operator if necessary.

[0848] Step 9:

[0849] The server sends the tailored response to the terminal for transmission back to the user, using a secure communications protocol.

[0850] Step 10:

[0851] The terminal displays the response received from the server on the user interface, and the user checks the displayed response and performs an operation based on it.

[0852] Step 11:

[0853] Perform actions based on the responses the user provides, for example, stepping through Wi-Fi setup procedures to resolve the issue.

[0854] Step 12:

[0855] The server logs all user queries and generated responses, providing data for later performance evaluation and model improvement.

[0856] Step 13:

[0857] The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine, including response accuracy, emotion recognition precision, and response time.

[0858] Step 14:

[0859] If the server finds that the model's performance falls below a predetermined standard, it collects new data and retrains the model, using the most recent query data and user feedback.

[0860] Step 15:

[0861] The user inputs feedback on the provided responses and sends it to the server via the terminal, which is used to improve the performance of the system.

[0862] As a result, this system can reduce the burden on sales staff and improve customer satisfaction.

[0863] Example 2

[0864] 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."

[0865] In modern information systems, it is difficult for users to obtain prompt and appropriate responses when inquiring about operation methods or specific functions. Furthermore, responses that take emotion into consideration are not provided, resulting in a poor user experience. Furthermore, it is difficult to generate adaptive responses based on the content of the inquiry, and there is a need for improved user satisfaction. Therefore, a system that solves these issues and provides efficient and emotion-sensitive responses is needed.

[0866] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0867] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving inquiries entered by a user and passing them to the trained model, means for analyzing the content of the user's inquiry to generate a prompt, means for analyzing the content of the inquiry and performing necessary preprocessing, means for passing the generated response to an emotion engine to recognize the user's emotion, means for adapting the tone and content of the response based on the recognized emotion, means for escalating the response to a human operator as necessary, means for returning the response generated from the model to the user, and means for recording the inquiries and the generated responses in a log and periodically evaluating the performance of the model. This makes it possible to provide quick and appropriate responses that take the user's emotions into consideration and improve user satisfaction.

[0868] "Information handling terminal" refers to all electronic devices that users can use to operate and configure.

[0869] An "instruction manual" refers to a document that describes how to use and set up the information terminal, as well as precautions to take.

[0870] "FAQ data" refers to data that compiles answers to questions and inquiries frequently asked by users.

[0871] A "database" refers to a system for efficiently storing, searching, and managing large amounts of data.

[0872] A "generative model" refers to an algorithm or machine learning model that automatically generates sentences or responses based on data.

[0873] "User" refers to an individual or organization that uses the system or service.

[0874] "Inquiry" refers to a question or concern a user has about a system or service.

[0875] A "prompt" refers to an input sentence that prompts a generative model to generate an appropriate response.

[0876] An "emotion engine" refers to a system or technology that analyzes emotions from a user's text data and recognizes their emotional state.

[0877] An "operator" refers to a human worker in charge of operating and managing a system or service.

[0878] "Log" refers to data that records system operation history and events.

[0879] "Performance evaluation" refers to methods for evaluating the accuracy of systems and models, response time, and emotion recognition accuracy.

[0880] "Preprocessing" refers to the process of converting data into the required format or removing noise before inputting it into a generative model.

[0881] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. The overall flow of this system will be specifically explained from the perspectives of the server, terminal, and user.

[0882] Data collection and learning phase

[0883] server

[0884] The server first collects the user's manual and FAQ data for the information terminal from the internet or an internal database. Web scraping technology (e.g., BeautifulSoup or Selenium) is used for collection. The collected data is centrally managed in CSV or JSON format. Next, data cleaning is performed, using regular expressions and database queries (e.g., SQL) to remove duplicate data and unnecessary information. The cleaned data is stored in a dedicated database. The server then trains the preprocessed data using a generative model (e.g., GPT-3). Machine learning frameworks such as TensorFlow and PyTorch are used for training. The trained model serves as the basis for responding quickly and accurately to future inquiries.

[0885] Query Receipt and Response Generation Phase

[0886] User

[0887] Users enter a query through a dedicated web interface, for example, "How do I set up Wi-Fi?"

[0888] Terminal

[0889] The user's device uses JavaScript or Ajax to convert the inquiry content into JSON format and sends it to the server's API endpoint.

[0890] server

[0891] The server parses the received JSON-formatted query, performs any necessary preprocessing, and then passes it to the generative model. The generative model generates an appropriate answer based on the query content. For example, the generative model retrieves information about "how to set up Wi-Fi" from a database and generates the appropriate steps.

[0892] Emotion engine recognizes emotions and adapts responses

[0893] server

[0894] The generated response is passed to an emotion engine (e.g., Google Cloud Natural Language API) to analyze the user's emotions. Emotions such as anger, sadness, and joy are detected from the text data. The emotion engine returns the analysis results to the server. The server then adjusts the tone and content of the response based on the analysis results. For example, if the user expresses anger, the tone of the response needs to be changed to be more polite. It also includes a means to escalate the response to a human operator if necessary.

[0895] Sending a Response

[0896] server

[0897] The server then converts the adjusted response back into JSON format and sends it to the device using a secure protocol such as HTTPS.

[0898] Terminal

[0899] The terminal displays the received response on the user interface, and the user can proceed with the operation by checking the displayed response.

[0900] Performance evaluation and improvement

[0901] server

[0902] The server records user queries and generated responses in a log file. This accumulates data for later performance evaluation and model improvement. The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[0903] User

[0904] Users can input their satisfaction with the responses and feedback through a dedicated UI, which is then sent to the server via their device. This feedback is used to improve the system.

[0905] Specific examples

[0906] For example, if a user asks, "How do I set up Wi-Fi?", the specific operation is as follows.

[0907] 1. User enters a query

[0908] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[0909] 2. The device sends a query to the server

[0910] This query is converted to JSON format using JavaScript and sent to the server's API endpoint via Ajax.

[0911] 3. The server receives and analyzes the query

[0912] Analyze the received JSON data using Python and perform the necessary preprocessing.

[0913] 4. The generative model generates a response

[0914] The preprocessed data is fed into a generative model (e.g., GPT-3) to obtain information about "how to set up Wi-Fi" and generate appropriate instructions.

[0915] 5. The server passes the query to the emotion engine

[0916] The generated response is passed to an emotion engine to recognize the user's emotion.

[0917] 6. Emotion engine recognizes user emotions

[0918] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[0919] 7. The server adjusts the response

[0920] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[0921] 8. The server sends the response to the device

[0922] The adjusted response is converted back to JSON format and sent to the terminal.

[0923] 9. The device displays the response

[0924] The received response is displayed in the user interface, and the user performs an action based on it.

[0925] Prompt Sentence Examples

[0926] "Please tell me how to set up Wi-Fi. And please provide appropriate emotion-aware responses to my queries."

[0927] This system allows users to receive prompt and sensitive answers, eliminating any concerns or questions they may have about the system.

[0928] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0929] Program processing steps

[0930] Step 1: Collect and store data

[0931] server

[0932] The server collects instruction manuals and FAQ data from the internet and internal databases. Specifically, it uses web scraping technology (e.g., BeautifulSoup or Selenium) to extract the necessary data from HTML pages such as instruction manuals. The collected data is saved in CSV or JSON format. It is then stored in a database so that it can be used for later processing.

[0933] Input: instruction manual, FAQ data

[0934] Output: Clean data stored in a database

[0935] Step 2: Cleaning the data

[0936] server

[0937] The server cleans the collected data, using regular expressions and database queries (e.g., SQL) to remove duplicate data and unnecessary information. The cleaned data is then stored in a dedicated database. This prepares the data for use in training the generative model.

[0938] Input: Collected data in the database

[0939] Output: Data in a clean database

[0940] Step 3: Train the model

[0941] server

[0942] The server trains a generative model (e.g., GPT-3) on clean data using a machine learning framework such as TensorFlow or PyTorch. The trained model then becomes the basis for responding to queries.

[0943] Input: Data in a clean database

[0944] Output: A trained generative model

[0945] Step 4: Receiving a user query

[0946] User

[0947] Users enter queries through a dedicated web interface, for example, "How do I set up Wi-Fi?"

[0948] Terminal

[0949] The terminal uses JavaScript or Ajax to convert the query content into JSON format and send it to the server's API endpoint.

[0950] Input: User's inquiry

[0951] Output: Query sent to the server in JSON format

[0952] Step 5: Parsing and preprocessing the query

[0953] server

[0954] The server parses the received JSON query and converts it into an appropriate format. Specifically, it uses Python to parse the JSON data and performs the necessary preprocessing before passing it to the generative model.

[0955] Input: JSON format query

[0956] Output: Preprocessed query data

[0957] Step 6: Generate a response

[0958] server

[0959] The server then passes the preprocessed query to the generative model, which then generates the appropriate response. For example, the generative model retrieves information about "how to set up Wi-Fi" from a database and generates the appropriate instructions.

[0960] Input: Preprocessed query data

[0961] Output: The generated response

[0962] Step 7: Emotion Recognition

[0963] server

[0964] The generated response is passed to an emotion engine (e.g., Google Cloud Natural Language API) to analyze the user's emotions. Emotions such as anger, sadness, and joy are detected from the text data. The emotion engine returns the analysis results to the server.

[0965] Input: Generated response

[0966] Output: User sentiment analysis results

[0967] Step 8: Adjusting the response

[0968] server

[0969] The server adjusts the tone and content of the response based on the results of the emotion engine. For example, if the user is expressing anger, the tone of the response should be changed to be more polite. It also includes a means to escalate the response to a human operator if necessary.

[0970] Input: User sentiment analysis results

[0971] Output: Adjusted response

[0972] Step 9: Sending a Response

[0973] server

[0974] The server then converts the adjusted response back into JSON format and sends it to the device using a secure protocol such as HTTPS.

[0975] Input: Adjusted response

[0976] Output: JSON formatted response sent to the terminal

[0977] Step 10: View the response

[0978] Terminal

[0979] The terminal displays the received response on the user interface, allowing the user to confirm the displayed response and proceed with the operation.

[0980] Input: The response sent by the server

[0981] Output: The response displayed in the user interface

[0982] Step 11: Logging queries and responses

[0983] server

[0984] The server records user queries and generated responses in a log file, which provides data for later performance evaluation and model improvement.

[0985] Input: The query and generated response

[0986] Output: Data recorded in the log file

[0987] Step 12: Evaluate the model's performance

[0988] server

[0989] The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine, including response accuracy, response time, and emotion recognition accuracy. If necessary, it collects new data and retrains the model.

[0990] Input: Log data

[0991] Output: Performance evaluation results and retrained model

[0992] Step 13: Gather user feedback

[0993] User

[0994] Users input their satisfaction and feedback regarding the provided responses through a dedicated UI and send it to the server via their device.

[0995] server

[0996] The server analyzes the collected feedback and uses it to improve the system.

[0997] Input: User feedback

[0998] Output: Improved system

[0999] Through these steps, users can receive quick and emotionally sensitive answers.

[1000] (Application example 2)

[1001] 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."

[1002] Conventional content distribution services have the problem of reducing user satisfaction because they recommend content without considering the user's emotional state. Furthermore, they often return a uniform response to user inquiries, failing to consider the user's emotions. This results in issues such as a decrease in service usage and an increased likelihood that users will switch to other services.

[1003] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1004] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving inquiries entered by users and passing them to the trained model, means for returning responses generated from the model to the users, means for recording the inquiries and the generated responses and periodically evaluating the performance of the model, emotion recognition means for analyzing emotions from the user's inquiry text and adjusting the response content, and means for recommending appropriate content according to the user's emotions in a content distribution service, thereby enabling responses and content recommendations that take user emotions into consideration.

[1005] "Handling information terminal" refers to a device used by a user to perform operations or make inquiries.

[1006] An "instruction manual" is a document or material that explains how to use an information terminal and its functions.

[1007] "FAQ data" is data that compiles information about questions and doubts that users frequently have.

[1008] A "generative model" is an algorithm or machine learning model that generates responses to queries based on collected data.

[1009] "Training means" refers to the method or process for training a generative model using collected data.

[1010] A "user query" is a question or request made by a user to a system for information or support.

[1011] A "response" is an answer or instruction that a generative model generates in response to a user query.

[1012] A "database" is a system for systematically storing and managing collected instruction manuals and frequently asked questions.

[1013] "Emotion recognition means" refers to the technology or function for analyzing and extracting emotions from the user's inquiry text.

[1014] A "content distribution service" is a service that provides users with content such as videos, music, and articles.

[1015] "Feedback means" refers to a method or function for collecting user evaluations and opinions on responses and using them to improve the system.

[1016] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. This system performs the following processes from the perspectives of the server, terminal, and user.

[1017] 1. Data collection and learning phase

[1018] server

[1019] First, the server collects instruction manuals and frequently asked questions data and stores them in a database. The collected data is pre-cleaned to remove duplicate data and unnecessary information. After data cleaning is complete, the data is stored in the database. Next, the server analyzes the pre-processed data using a generative model to train the model. A specific machine learning algorithm is used for this training. Once the model is trained, it serves as the basis for responding quickly and accurately to future inquiries.

[1020] 2. Query Receipt and Response Generation Phase

[1021] User

[1022] If a user has a question about how to operate the system or a specific function, they can input a query into the system through a dedicated user interface, such as "What movies do you recommend?"

[1023] Terminal

[1024] User inquiries received via the user interface are sent by the terminal to the server. Specifically, the inquiry content is converted into JSON format or similar and sent to the server's API.

[1025] server

[1026] The server analyzes the received query, performs necessary preprocessing, and then passes it to the generative model, which generates an appropriate answer based on the query and returns the response to the server.

[1027] 3. Emotion recognition and response adaptation using an emotion engine

[1028] server

[1029] Before sending the generated response, the server passes the query content to an emotion engine to recognize the user's emotion. The emotion engine uses natural language processing technology to analyze emotions from the user's text data. As a result, it determines whether the user is feeling anger, sadness, joy, anticipation, or other emotions. Based on the emotion recognized by the emotion engine, the server adapts the tone and content of the response. For example, if the user is expressing anger, the tone of the response will be adjusted to be more polite. It also includes a function to escalate to a human operator if necessary.

[1030] 4. Sending the Response

[1031] server

[1032] Once the tailored response is complete, the server sends it to the terminal for transmission back to the user, using a secure protocol.

[1033] Terminal

[1034] The terminal displays the response received from the server on the user interface, allowing the user to perform operations based on the response.

[1035] 5. Performance evaluation and improvement

[1036] server

[1037] The server logs user queries and generated responses. This accumulates data for later performance evaluation and model improvement. It periodically analyzes the log data to evaluate the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[1038] User

[1039] The user inputs feedback on the provided responses and sends it to the server via the terminal, which helps improve the system.

[1040] Specific examples

[1041] For example, if a user inquires, "What movies do you recommend?", the flow is as follows:

[1042] 1. User enters a query

[1043] The user enters a question.

[1044] 2. The device sends a query to the server

[1045] The user terminal sends this query to the server's API.

[1046] 3. The server receives the query and passes it to the generative model.

[1047] The server analyzes the received query and inputs it into the generative model.

[1048] 4. The generative model generates a response

[1049] The generative model retrieves information about "movie recommendation" from the database and generates appropriate steps.

[1050] 5. The server passes the query to the emotion engine

[1051] The server passes the query content to the emotion engine to recognize the user's emotion.

[1052] 6. Emotion engine recognizes user emotions

[1053] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[1054] 7. The server adjusts the response

[1055] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[1056] 8. The server sends the response to the device

[1057] The server sends the tailored response to the terminal for transmission back to the user.

[1058] 9. The device displays the response

[1059] The terminal displays the response on the user interface, and the user performs an operation based on the response.

[1060] This system allows users to receive prompt and emotionally sensitive answers, improving their satisfaction with the service.

[1061] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1062] Step 1:

[1063] User enters a query

[1064] A user types a question into a smartphone application, for example, "What movies do you recommend?" This query is entered through a user interface, and the entered text is stored on the device as raw text.

[1065] Input: User query text (e.g., "What movies do you recommend?")

[1066] Output: Raw query text

[1067] Step 2:

[1068] The device sends a query to the server

[1069] The terminal converts user inquiries received via the user interface into JSON format and sends it to the server's API, allowing the server to receive the inquiry content.

[1070] Input: Raw query text

[1071] Output: Data converted to JSON format (e.g., {"query": "What movies do you recommend?"})

[1072] Step 3:

[1073] The server receives the query and passes it to the generative model.

[1074] The server parses the received JSON data and extracts the query content, which is then passed to the generative model, which then generates an appropriate response based on this data.

[1075] Input: Query data in JSON format

[1076] Output: Generated response (e.g., "I recommend 'Hero Movie' as an uplifting action movie.")

[1077] Step 4:

[1078] The generative model generates the response

[1079] The generative model generates an appropriate answer based on the query. This process involves retrieving relevant information from a database and performing data processing and calculations to generate the appropriate response.

[1080] Input: The query passed to the generative model

[1081] Output: Correct response

[1082] Step 5:

[1083] The server passes the query to the emotion engine

[1084] The server passes the generated response and the original query text to the emotion engine to recognize the user's emotion. The emotion engine uses natural language processing techniques to analyze the user's emotional state from the text.

[1085] Input: Generated response and original query text

[1086] Output: Emotion recognition result (e.g. positive, negative, neutral)

[1087] Step 6:

[1088] Emotion engine recognizes user emotions

[1089] The emotion engine analyzes the user's emotion from the input text and returns the state to the server. In this step, data processing and calculations are performed to extract the user's emotion using an analysis algorithm and return the result.

[1090] Input: Query text and generated response

[1091] Output: Emotion recognition result (e.g., positive)

[1092] Step 7:

[1093] The server adjusts the response

[1094] Based on the results of the emotion engine, the server adjusts the generated response, ensuring that the response is of the appropriate tone and content depending on the user's emotional state, and escalating to a human operator if necessary.

[1095] Input: Generated response sentences and emotion recognition results

[1096] Output: Tailored response (e.g., "I recommend 'Hero Movie' as an uplifting action movie. I'm glad you liked it!")

[1097] Step 8:

[1098] The server sends a response to the device

[1099] The server sends the tailored response to the terminal for transmission back to the user, using a secure protocol.

[1100] Input: Adjusted response sentence

[1101] Output: Response data sent to the terminal

[1102] Step 9:

[1103] The terminal displays the response

[1104] The terminal displays the response received from the server on the user interface, allowing the user to perform operations and make selections based on the response.

[1105] Input: Response data sent from the server

[1106] Output: The response text displayed in the user interface

[1107] 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.

[1108] 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.

[1109] 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.

[1110] [Third embodiment]

[1111] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1112] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[1113] 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).

[1114] 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.

[1115] 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.

[1116] 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).

[1117] 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.

[1118] 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.

[1119] 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.

[1120] 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.

[1121] 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.

[1122] 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."

[1123] The present invention is an operation method support system that uses a generative model to provide convenience to both salespeople and customers. The overall flow of this system will be explained from the perspectives of the server, terminals, and users.

[1124] 1. Data collection and learning phase

[1125] server

[1126] The server first collects instruction manuals and FAQ (frequently asked questions) data for all information devices handled from the Internet and internal databases, and then centrally manages them. Since this data is used as learning material for the model, it is cleaned in advance to remove duplicate data and unnecessary information. After data cleaning is complete, the data is organized and saved.

[1127] The server then trains the pre-processed data with a generative model using machine learning algorithms, which then serves as the basis for quickly and accurately responding to future queries.

[1128] 2. Inquiry Receipt and Response Phase

[1129] User

[1130] When a user has a question about an operation method or a specific function, they input their query into the system through a dedicated user interface, which may be provided in the form of a web browser, a mobile application, or a help desk.

[1131] Terminal

[1132] User inquiries received via the user interface are sent from the terminal to the server. The terminal converts the inquiry content into a format such as JSON and sends it to the server's data reception API.

[1133] server

[1134] The server analyzes the received query and passes it to a trained generative model. The generative model generates an appropriate answer based on the query, constructs a response, and returns it to the server. The server then sends this response back to the user, allowing the user to receive a quick and accurate answer.

[1135] 3. Performance evaluation and improvement phase

[1136] server

[1137] The server records all queries and responses in log files. Periodically, it analyzes these logs to evaluate the model's performance. This evaluation involves scoring the generated responses to ensure their accuracy. Based on the evaluation results, if the model's accuracy falls below a predetermined standard, it may automatically initiate a process to collect additional data and retrain.

[1138] User

[1139] Additionally, users can enter feedback on the answers they provide, which is sent to the server and used as data for future model improvements.

[1140] Specific examples

[1141] For example, suppose a user asks, "How do I set up Wi-Fi?" The flow is shown below.

[1142] 1. User enters a query

[1143] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[1144] 2. The device sends a query to the server

[1145] The browser on the user's device sends this query to the server's API.

[1146] 3. The server receives the query and passes it to the generative model.

[1147] The server analyzes the received query and inputs it into the generative model.

[1148] 4. The generative model generates a response

[1149] The generative model retrieves information about "how to set up Wi-Fi" from a database and generates appropriate instructions.

[1150] 5. The server sends the response back to the user

[1151] The server generates a response and sends it back to the user, who displays it in their browser.

[1152] 6. The user receives the response and performs the action

[1153] The user follows the displayed steps to set up Wi-Fi.

[1154] This allows users to receive prompt and appropriate responses, and frees sales representatives from having to deal with inquiries about operation methods, allowing them to focus on more important tasks.

[1155] The processing flow will be explained below.

[1156] Step 1:

[1157] The server collects instruction manuals and FAQ data from the information terminals it handles and stores them in a database, including scraping data from the Internet and retrieving it from its internal database.

[1158] Step 2:

[1159] The server preprocesses the collected data, specifically by performing text cleaning and data normalization to remove duplicate data and unnecessary information.

[1160] Step 3:

[1161] The server trains a generative model based on the preprocessed data, using machine learning algorithms to extract the necessary knowledge from the data and train the model.

[1162] Step 4:

[1163] The server stores the trained generative model and prepares it for generating responses to queries.

[1164] Step 5:

[1165] The user inputs an inquiry about operation methods through a dedicated user interface, for example, "Please tell me how to set up Wi-Fi."

[1166] Step 6:

[1167] The terminal converts the user's inquiry into JSON format or similar and sends it to the server's API. This transmission generally uses the HTTPS protocol.

[1168] Step 7:

[1169] The server parses the received query, performs any necessary preprocessing, and then passes it to the generative model. Parsing includes tokenizing and encoding the text.

[1170] Step 8:

[1171] The server receives the response from the generative model and transforms it into a format that is easy for the user to understand, for example, organizing the steps as a numbered list.

[1172] Step 9:

[1173] The server sends the converted response back to the user via the API to the device, again using a secure protocol.

[1174] Step 10:

[1175] The device displays the response received from the server and provides it to the user, either on a web page or within an application.

[1176] Step 11:

[1177] Performs actions based on the answers the user provides, for example, walking through Wi-Fi setup steps to resolve the issue.

[1178] Step 12:

[1179] The server logs user queries and the responses it generates, providing data for later performance evaluation and model improvement.

[1180] Step 13:

[1181] The server periodically analyzes the log data and evaluates the performance of the generative model using metrics such as accuracy and response time.

[1182] Step 14:

[1183] If the accuracy of the model falls below a predetermined standard, the server collects new data and retrains the model, ensuring that it always provides highly accurate responses.

[1184] Step 15:

[1185] The user inputs feedback on the provided responses and sends it to the server via the terminal, which is used to improve the system.

[1186] As a result, this system can reduce the burden on sales staff and improve customer satisfaction.

[1187] Example 1

[1188] 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."

[1189] Many modern information terminals have complex operation methods and multifunctional settings, which often lead to users becoming confused and not receiving appropriate support. Sales representatives are also overwhelmed by responding to a variety of inquiries, preventing them from concentrating on their core duties. Furthermore, traditional FAQs and instruction manuals only provide static information, making it difficult to provide dynamic and prompt support.

[1190] 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.

[1191] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for cleaning and preprocessing the collected data, means for training the preprocessed data using a generative model, means for receiving inquiries entered by users, converting the inquiry content into JSON format, and sending the JSON format to the server, means for inputting the user's inquiry into the trained generative model and obtaining a response generated from the model, means for returning the obtained response to the user interface, and means for recording the inquiries and the generated responses and periodically evaluating the performance of the model. This allows users to receive prompt and appropriate support, and frees sales representatives from responding to inquiries so that they can focus on their core business.

[1192] An "information terminal" is a device or system that provides information about an electronic device or system operated by a user.

[1193] An "instruction manual" is a document that describes the basic operation methods, setting procedures, precautions, etc. of an information terminal.

[1194] "Frequently asked questions data" is data that compiles questions frequently asked by users and the answers to those questions.

[1195] A "database" is a system for efficiently storing, managing, and searching large amounts of data.

[1196] "Data cleanup" is the process of removing redundancies and noise from collected data and preparing it in a format suitable for analysis.

[1197] "Preprocessing" is a series of steps to convert data into a format suitable for a generative model.

[1198] A "generative model" is a model that uses machine learning algorithms to generate optimal answers from given data.

[1199] "Training" is the process of optimizing the parameters of a generative model using collected data to train the model.

[1200] A "user interface" is an interactive screen or application that allows a user to access and query a system.

[1201] "JSON format" is a data format that structures data in text format and makes it easily exchangeable.

[1202] A "server" is a computer system that processes data and provides services over a network.

[1203] An "inquiry" is a question or request made by a user to the system.

[1204] A "response" is an answer that a system provides to a user's query.

[1205] "Recording" means saving the contents of inquiries and responses in a log file or the like.

[1206] "Performance evaluation" is the process of evaluating the accuracy and appropriateness of the responses of a generative model.

[1207] This invention is an operation method support system that uses a generative AI model to provide convenience to both sales representatives and customers. In this system, the server, terminal, and user each play specific roles, and provide quick and accurate answers to user inquiries.

[1208] Data collection and preprocessing

[1209] server

[1210] The server collects the user manuals and FAQ data for the information terminals from the internet and an internal database. It also cleans the data, removing duplicate data and noise. Specifically, it uses Python scripts and BeautifulSoup to scrape the web, and the Pandas library to process the data. The cleaned data is then stored using a large-scale data processing framework (e.g., Hadoop).

[1211] Training generative AI models

[1212] server

[1213] The server trains a generative model using the collected and preprocessed data. TensorFlow or PyTorch is used for this training. The training data is split using a tokenizer and input to a neural network. A deep learning framework is used to optimize the model parameters. In particular, the Transformers architecture is commonly used.

[1214] Receiving inquiries via a user interface

[1215] User

[1216] Users input questions about operations and specific functions through a user interface, which is provided by a web browser or a mobile application, such as Google Chrome or Safari.

[1217] Specific examples

[1218] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[1219] Analysis of inquiries

[1220] Terminal

[1221] The device converts the user's inquiry into JSON format and sends it to the server's receiving API. For example, it generates JSON data using a JavaScript or Python library and sends it to the server via an HTTP request. Browsers used include Firefox and Google Chrome.

[1222] Answer generation using generative AI models

[1223] server

[1224] The server inputs the received query into a generative model. This process involves encoding the query and converting it into a format suitable for the generative model. The generative AI model (e.g., GPT-3) generates an appropriate answer based on the query. The generated answer is then formatted.

[1225] Specific examples

[1226] The server uses the generative model to generate instructions on how to set up Wi-Fi.

[1227] Returning answers to users

[1228] server

[1229] The server generates an answer and sends it back to the user. The response data is organized in JSON format and sent to the device in an HTTP response.

[1230] Terminal

[1231] The device analyzes the returned data and displays it in a user interface, using HTML and JavaScript to display the answers in the browser.

[1232] Performance evaluation and feedback collection

[1233] server

[1234] The server logs all queries and responses and periodically analyzes them. The log data is evaluated using Python and R to measure the accuracy of the responses. Users can enter feedback on the answers provided, and this data is also collected.

[1235] This process allows users to receive prompt and appropriate answers, and frees up sales representatives from the burden of handling inquiries. The system continuously collects data to improve the model's performance, providing a high-quality user experience.

[1236] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1237] Step 1:

[1238] Data collection and preprocessing

[1239] The server collects the user manuals and frequently asked questions (FAQ data) for the information terminals from the internet and an internal database. The input uses the URL of the online user manual and the access information for the internal database. Specifically, it performs web scraping using a Python script and BeautifulSoup to extract the data. After the data is collected, it is formatted using Pandas to remove duplicate data and noise. The clean data is generated as output and stored in Hadoop or another storage system.

[1240] Step 2:

[1241] Training generative AI models

[1242] The server trains a generative model using the collected and preprocessed data. It uses clean training data as input. Specifically, it tokenizes the training data using the TensorFlow or PyTorch framework and feeds it to a neural network. The selected deep learning model (e.g., Transformers architecture) learns patterns in the data and finds optimal parameters. The output is a trained generative model.

[1243] Step 3:

[1244] Receiving inquiries via a user interface

[1245] The user inputs questions about operation methods or specific functions through the user interface. The user's question is used as input. Specifically, the user enters a question into an inquiry form in a web browser such as Google Chrome or Safari and clicks the submit button. The user's question data is sent to the device as output.

[1246] Step 4:

[1247] Analysis of inquiries

[1248] The terminal converts the user's inquiry into JSON format and sends it to the server's receiving API. The user's question received from the web browser is used as input. Specifically, it generates JSON data using JavaScript or Python libraries and sends it to the server via an HTTP request. The JSON-formatted inquiry data is sent to the server as output.

[1249] Step 5:

[1250] Answer generation using generative AI models

[1251] The server inputs the received query content into the generative model. Parsed JSON data is used as input. Specifically, the server encodes the query content and converts it into a format suitable for the generative model. The generative AI model (e.g., GPT-3) generates an appropriate answer based on the query content. The output is the generated response text.

[1252] Step 6:

[1253] Returning answers to users

[1254] The server returns the generated answer to the user. The generated response text is used as input. Specifically, the server formats the response data in JSON format and sends it to the terminal as an HTTP response. The terminal receives the returned data and displays it in the user interface. As output, the final response is displayed on the user's browser.

[1255] Step 7:

[1256] Performance evaluation and feedback collection

[1257] The server logs all queries and responses and periodically analyzes them. The query and response log data is used as input. Specific operations include evaluating the log data using Python or R to measure the model's performance. Users can provide feedback on the answers provided, and this data is also collected. The output is the model's performance evaluation results and user feedback.

[1258] This processing step allows the system to respond quickly and accurately to user inquiries and to continually collect data to improve performance.

[1259] (Application example 1)

[1260] 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."

[1261] Conventional systems have struggled to respond quickly and accurately to user inquiries about content search and playback settings. Content distribution services, in particular, require appropriate recommendations based on users' viewing history and recent popular content, but there has been a lack of efficient ways to provide this information. Furthermore, systems lack the ability to automatically update data and retrain when system performance deteriorates.

[1262] 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.

[1263] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving queries entered by users and passing them to the trained model, means for returning responses generated from the model to the users, means for recording queries and the generated responses and periodically evaluating the performance of the model, and means for generating content recommendations based on the user's viewing history and recent popular content in content distribution, thereby enabling users to receive appropriate recommendations and quick responses regarding content they want to view.

[1264] The term "information terminals" refers to various electronic devices and content playback terminals used by users.

[1265] "Instruction manual" refers to a document that contains detailed information about the operation and functions of the information terminal.

[1266] "Frequently asked questions data" refers to documents or data that systematically compile answers to questions and problems that users frequently have.

[1267] A "database" refers to a system for efficiently storing, searching, and managing collected information and data.

[1268] "Generative model" refers to a system trained using machine learning algorithms to generate appropriate responses to user queries.

[1269] "Training" refers to the process of using collected data to train a generative model and improve its accuracy.

[1270] "Content distribution" refers to the service of providing digital media such as movies, TV shows, and music to users via the Internet.

[1271] "Viewing history" refers to a record of content that a user has viewed in the past.

[1272] "Popular content" refers to digital media that is viewed by many users and has received high ratings.

[1273] "Content recommendation" refers to suggesting the next piece of media to watch based on the user's viewing history and popular content.

[1274] This invention applies a content distribution service to an operation method support system that uses a generative model to provide convenience to both sales representatives and customers. The details of this system will be explained in three phases: data collection and learning phase, inquiry reception and response phase, and performance evaluation and improvement phase.

[1275] 1. Data collection and learning phase

[1276] server

[1277] The server first collects instruction manuals and FAQs (frequently asked questions) for the information terminals (various electronic devices and content playback devices used by users) from the Internet and an internal database, and then centrally manages them. Since this data is used as learning material for the model, it is cleaned in advance to remove duplicate data and unnecessary information. Python-based scripts and data processing libraries (Pandas, NumPy, etc.) are used for data cleaning.

[1278] The server then trains the preprocessed data using a generative model (e.g., GPT-4), which lays the foundation for the model to respond quickly and accurately to user queries.

[1279] 2. Inquiry Receipt and Response Phase

[1280] User

[1281] Users access the "Content Help AI" app on their smartphone or head-mounted display (HMD) and input a query by voice or text, such as "What movie would you recommend I watch next?"

[1282] Terminal

[1283] The device (smartphone or HMD) converts the input voice data into text (using voice recognition software) and converts it into JSON format, which is then sent to the server.

[1284] server

[1285] The server analyzes the received query and passes it to the generative model, which generates an appropriate response from the database based on the prompt "Please provide information about recommended movies that the user should watch." For example, the prompt might look like this:

[1286] "What movie should users watch next? Create a recommendation list that takes into account recent popular movies and the user's viewing history."

[1287] After the generative model generates a response, the server sends the response back to the user device and displays it on the user's smartphone or HMD.

[1288] 3. Performance evaluation and improvement phase

[1289] server

[1290] The server records all queries and the generated responses in a log file, allowing the system's performance to be evaluated periodically. If performance falls below a predetermined standard, the server collects additional data and retrains the model.

[1291] User

[1292] Additionally, users can provide feedback on the answers they provide, which is also sent to the server and used as data for future model improvements.

[1293] As a concrete example, when a user asks, "What movie would you recommend I watch next?" the system follows these steps:

[1294] 1. The user uses the app to enter a question by voice.

[1295] 2. The device converts the voice into text and sends it to the server.

[1296] 3. The server inputs a prompt sentence into the generative model.

[1297] 4. The generative model generates a list of appropriate movie recommendations.

[1298] 5. The server generates a response and sends it back to the user, who displays it in the app.

[1299] This process allows users to get relevant and fast recommendations for content they might want to watch.

[1300] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1301] Step 1:

[1302] User enters a query

[1303] Users use the "Content Help AI" app on their smartphones or head-mounted displays (HMDs) to input questions about the content they want to watch. For example, they might say, "What movie would you recommend I watch next?" This input data is in audio format.

[1304] Step 2:

[1305] The device converts the voice data into text

[1306] The device uses speech recognition software to convert the voice data into text data. The input to this process is the user's voice data, and the output is the question in text form.

[1307] Step 3:

[1308] The device converts the text data into JSON format and sends it to the server.

[1309] The terminal converts text data into JSON format. In this conversion process, the text data is converted into a structured JSON object. The converted JSON data is sent to the server. The input of this process is text data, and the output is JSON format data.

[1310] Step 4:

[1311] The server receives and parses the JSON data.

[1312] The server receives and analyzes the JSON data sent from the terminal. In this analysis process, the JSON object is analyzed to extract the user's query. The input of this process is the JSON data, and the output is the extracted data of the query.

[1313] Step 5:

[1314] The server inputs a prompt to the generative model.

[1315] The server inputs a prompt sentence into the generative model based on the extracted query content. An example of a prompt sentence is, "Please tell me what movie the user should watch next. Please create a recommendation list taking into account recent popular movies and the user's viewing history." The inputs to this process are the extracted data and the prompt sentence, and the output is a response generated by the generative model.

[1316] Step 6:

[1317] The generative model generates the response

[1318] A generative model (e.g., GPT-4) generates a response based on the prompt sentence. This response generation process creates an appropriate recommendation list based on the viewing history and popular content stored in the database. The input of this process is the prompt sentence, and the output is the generated response data.

[1319] Step 7:

[1320] The server generates a response and sends it back to the user.

[1321] The server receives the response generated from the generative model and sends it back to the user terminal. In this process, the generated response data is converted into an appropriate format and sent to the terminal. The input of this process is the generated response data, and the output is the data sent to the user terminal.

[1322] Step 8:

[1323] The terminal displays the response data to the user

[1324] The device analyzes the response data received from the server and displays it to the user. This display process converts the generated response into a format that is easy for the user to understand and displays it within the app. The input of this process is the response data received from the server, and the output is the information displayed on the user's screen.

[1325] In this way, users can get appropriate and quick recommendations and operating procedures for the content they want to watch.

[1326] 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.

[1327] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. The overall flow of this system will be specifically explained from the perspectives of the server, terminal, and user.

[1328] 1. Data collection and learning phase

[1329] server

[1330] The server first collects instruction manuals and FAQ (frequently asked questions) data for all information devices it handles from the Internet and an internal database, and then centrally manages them. This data is used as learning material for the generative model. The collected data is pre-cleaned to remove duplicate data and unnecessary information. After data cleaning is complete, the data is stored in the database.

[1331] The server then trains the pre-processed data with a generative model using machine learning algorithms, which then serves as the basis for quickly and accurately responding to future queries.

[1332] 2. Query Receipt and Response Generation Phase

[1333] User

[1334] If a user has a question about how to operate the system or a specific function, they can enter a query into the system through a dedicated user interface, such as "How do I set up Wi-Fi?"

[1335] Terminal

[1336] User inquiries received via the user interface are sent by the terminal to the server. Specifically, the inquiry content is converted into JSON format or similar and sent to the server's API.

[1337] server

[1338] The server analyzes the received query, performs necessary preprocessing, and then passes it to the generative model, which generates an appropriate answer based on the query and returns the response to the server.

[1339] 3. Emotion recognition and response adaptation using an emotion engine

[1340] server

[1341] Before sending the generated response, the server passes the query content to an emotion engine to recognize the user's emotion. The emotion engine uses natural language processing technology to analyze emotions from the user's text data, and determines whether the user is feeling anger, sadness, joy, anticipation, or other emotions.

[1342] Based on the emotions recognized by the emotion engine, the server will adapt the tone and content of the response. For example, if the user is angry, the tone of the response will be more polite. It also includes the ability to escalate to a human operator if necessary.

[1343] 4. Sending the Response

[1344] server

[1345] Once the tailored response is complete, the server sends it to the terminal for transmission back to the user, using a secure protocol.

[1346] Terminal

[1347] The terminal displays the response received from the server on the user interface, allowing the user to perform operations based on the response.

[1348] 5. Performance evaluation and improvement

[1349] server

[1350] The server logs user queries and generated responses. This accumulates data for later performance evaluation and model improvement. It periodically analyzes the log data to evaluate the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[1351] User

[1352] The user inputs feedback on the provided responses and sends it to the server via the terminal, which helps improve the system.

[1353] Specific examples

[1354] For example, if a user asks, "How do I set up Wi-Fi?", the flow is as follows:

[1355] 1. User enters a query

[1356] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[1357] 2. The device sends a query to the server

[1358] The browser on the user's device sends this query to the server's API.

[1359] 3. The server receives the query and passes it to the generative model.

[1360] The server analyzes the received query and inputs it into the generative model.

[1361] 4. The generative model generates a response

[1362] The generative model retrieves information about "how to set up Wi-Fi" from a database and generates appropriate instructions.

[1363] 5. The server passes the query to the emotion engine

[1364] The server passes the query content to the emotion engine to recognize the user's emotion.

[1365] 6. Emotion engine recognizes user emotions

[1366] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[1367] 7. The server adjusts the response

[1368] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[1369] 8. The server sends the response to the device

[1370] The server sends the tailored response to the terminal for transmission back to the user.

[1371] 9. The device displays the response

[1372] The terminal displays the response on the user interface, and the user performs an operation based on the response.

[1373] This system allows users to receive quick and sensitive responses, freeing sales representatives from having to deal with inquiries about how to use the system, allowing them to focus on more important tasks.

[1374] The processing flow will be explained below.

[1375] Step 1:

[1376] The server collects the instruction manuals and FAQ data of the information terminals it handles and stores them in a database. This is done by scraping data from the internet and retrieving data from the company's internal database. The collected data is pre-processed to remove duplicate and irrelevant data.

[1377] Step 2:

[1378] The server trains a generative model based on the preprocessed data, using machine learning algorithms to analyze the data and train the model to generate appropriate responses to queries.

[1379] Step 3:

[1380] The server uses an emotion engine to train the system to recognize user emotions based on the text data contained in the user's inquiry. The emotion engine uses natural language processing technology to build an emotion analysis model.

[1381] Step 4:

[1382] The user inputs an inquiry about operation methods through a dedicated user interface. For example, the user inputs a specific question such as "How do I set up Wi-Fi?"

[1383] Step 5:

[1384] The terminal converts the query entered by the user into an appropriate format, such as JSON, and sends it to the server's API using the HTTPS protocol.

[1385] Step 6:

[1386] The server parses the received query and performs any necessary preprocessing, including tokenizing and encoding the text, before feeding the preprocessed data into the generative model.

[1387] Step 7:

[1388] The server receives the response from the generative model and inputs it into the emotion engine to analyze the user's emotion. The emotion engine recognizes the user's emotion and returns the result to the server.

[1389] Step 8:

[1390] The server adapts the tone and content of the response based on the results of the emotion engine. For example, if the user is angry, the tone of the response will be more polite. It will also escalate the situation to a human operator if necessary.

[1391] Step 9:

[1392] The server sends the tailored response to the terminal for transmission back to the user, using a secure communications protocol.

[1393] Step 10:

[1394] The terminal displays the response received from the server on the user interface, and the user checks the displayed response and performs an operation based on it.

[1395] Step 11:

[1396] Perform actions based on the responses the user provides, for example, stepping through Wi-Fi setup procedures to resolve the issue.

[1397] Step 12:

[1398] The server logs all user queries and generated responses, providing data for later performance evaluation and model improvement.

[1399] Step 13:

[1400] The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine, including response accuracy, emotion recognition precision, and response time.

[1401] Step 14:

[1402] If the server finds that the model's performance falls below a predetermined standard, it collects new data and retrains the model, using the most recent query data and user feedback.

[1403] Step 15:

[1404] The user inputs feedback on the provided responses and sends it to the server via the terminal, which is used to improve the performance of the system.

[1405] As a result, this system can reduce the burden on sales staff and improve customer satisfaction.

[1406] Example 2

[1407] 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."

[1408] In modern information systems, it is difficult for users to obtain prompt and appropriate responses when inquiring about operation methods or specific functions. Furthermore, responses that take emotion into consideration are not provided, resulting in a poor user experience. Furthermore, it is difficult to generate adaptive responses based on the content of the inquiry, and there is a need for improved user satisfaction. Therefore, a system that solves these issues and provides efficient and emotion-sensitive responses is needed.

[1409] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1410] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving inquiries entered by a user and passing them to the trained model, means for analyzing the content of the user's inquiry to generate a prompt, means for analyzing the content of the inquiry and performing necessary preprocessing, means for passing the generated response to an emotion engine to recognize the user's emotion, means for adapting the tone and content of the response based on the recognized emotion, means for escalating the response to a human operator as necessary, means for returning the response generated from the model to the user, and means for recording the inquiries and the generated responses in a log and periodically evaluating the performance of the model. This makes it possible to provide quick and appropriate responses that take the user's emotions into consideration and improve user satisfaction.

[1411] "Information handling terminal" refers to all electronic devices that users can use to operate and configure.

[1412] An "instruction manual" refers to a document that describes how to use and set up the information terminal, as well as precautions to take.

[1413] "FAQ data" refers to data that compiles answers to questions and inquiries frequently asked by users.

[1414] A "database" refers to a system for efficiently storing, searching, and managing large amounts of data.

[1415] A "generative model" refers to an algorithm or machine learning model that automatically generates sentences or responses based on data.

[1416] "User" refers to an individual or organization that uses the system or service.

[1417] "Inquiry" refers to a question or concern a user has about a system or service.

[1418] A "prompt" refers to an input sentence that prompts a generative model to generate an appropriate response.

[1419] An "emotion engine" refers to a system or technology that analyzes emotions from a user's text data and recognizes their emotional state.

[1420] An "operator" refers to a human worker in charge of operating and managing a system or service.

[1421] "Log" refers to data that records system operation history and events.

[1422] "Performance evaluation" refers to methods for evaluating the accuracy of systems and models, response time, and emotion recognition accuracy.

[1423] "Preprocessing" refers to the process of converting data into the required format or removing noise before inputting it into a generative model.

[1424] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. The overall flow of this system will be specifically explained from the perspectives of the server, terminal, and user.

[1425] Data collection and learning phase

[1426] server

[1427] The server first collects the user's manual and FAQ data for the information terminal from the internet or an internal database. Web scraping technology (e.g., BeautifulSoup or Selenium) is used for collection. The collected data is centrally managed in CSV or JSON format. Next, data cleaning is performed, using regular expressions and database queries (e.g., SQL) to remove duplicate data and unnecessary information. The cleaned data is stored in a dedicated database. The server then trains the preprocessed data using a generative model (e.g., GPT-3). Machine learning frameworks such as TensorFlow and PyTorch are used for training. The trained model serves as the basis for responding quickly and accurately to future inquiries.

[1428] Query Receipt and Response Generation Phase

[1429] User

[1430] Users enter a query through a dedicated web interface, for example, "How do I set up Wi-Fi?"

[1431] Terminal

[1432] The user's device uses JavaScript or Ajax to convert the inquiry content into JSON format and sends it to the server's API endpoint.

[1433] server

[1434] The server parses the received JSON-formatted query, performs any necessary preprocessing, and then passes it to the generative model. The generative model generates an appropriate answer based on the query content. For example, the generative model retrieves information about "how to set up Wi-Fi" from a database and generates the appropriate steps.

[1435] Emotion engine recognizes emotions and adapts responses

[1436] server

[1437] The generated response is passed to an emotion engine (e.g., Google Cloud Natural Language API) to analyze the user's emotions. Emotions such as anger, sadness, and joy are detected from the text data. The emotion engine returns the analysis results to the server. The server then adjusts the tone and content of the response based on the analysis results. For example, if the user expresses anger, the tone of the response needs to be changed to be more polite. It also includes a means to escalate the response to a human operator if necessary.

[1438] Sending a Response

[1439] server

[1440] The server then converts the adjusted response back into JSON format and sends it to the device using a secure protocol such as HTTPS.

[1441] Terminal

[1442] The terminal displays the received response on the user interface, and the user can proceed with the operation by checking the displayed response.

[1443] Performance evaluation and improvement

[1444] server

[1445] The server records user queries and generated responses in a log file. This accumulates data for later performance evaluation and model improvement. The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[1446] User

[1447] Users can input their satisfaction with the responses and feedback through a dedicated UI, which is then sent to the server via their device. This feedback is used to improve the system.

[1448] Specific examples

[1449] For example, if a user asks, "How do I set up Wi-Fi?", the specific operation is as follows.

[1450] 1. User enters a query

[1451] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[1452] 2. The device sends a query to the server

[1453] This query is converted to JSON format using JavaScript and sent to the server's API endpoint via Ajax.

[1454] 3. The server receives and analyzes the query

[1455] Analyze the received JSON data using Python and perform the necessary preprocessing.

[1456] 4. The generative model generates a response

[1457] The preprocessed data is fed into a generative model (e.g., GPT-3) to obtain information about "how to set up Wi-Fi" and generate appropriate instructions.

[1458] 5. The server passes the query to the emotion engine

[1459] The generated response is passed to an emotion engine to recognize the user's emotion.

[1460] 6. Emotion engine recognizes user emotions

[1461] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[1462] 7. The server adjusts the response

[1463] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[1464] 8. The server sends the response to the device

[1465] The adjusted response is converted back to JSON format and sent to the terminal.

[1466] 9. The device displays the response

[1467] The received response is displayed in the user interface, and the user performs an action based on it.

[1468] Prompt Sentence Examples

[1469] "Please tell me how to set up Wi-Fi. And please provide appropriate emotion-aware responses to my queries."

[1470] This system allows users to receive prompt and sensitive answers, eliminating any concerns or questions they may have about the system.

[1471] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1472] Program processing steps

[1473] Step 1: Collect and store data

[1474] server

[1475] The server collects instruction manuals and FAQ data from the internet and internal databases. Specifically, it uses web scraping technology (e.g., BeautifulSoup or Selenium) to extract the necessary data from HTML pages such as instruction manuals. The collected data is saved in CSV or JSON format. It is then stored in a database so that it can be used for later processing.

[1476] Input: instruction manual, FAQ data

[1477] Output: Clean data stored in a database

[1478] Step 2: Cleaning the data

[1479] server

[1480] The server cleans the collected data, using regular expressions and database queries (e.g., SQL) to remove duplicate data and unnecessary information. The cleaned data is then stored in a dedicated database. This prepares the data for use in training the generative model.

[1481] Input: Collected data in the database

[1482] Output: Data in a clean database

[1483] Step 3: Train the model

[1484] server

[1485] The server trains a generative model (e.g., GPT-3) on clean data using a machine learning framework such as TensorFlow or PyTorch. The trained model then becomes the basis for responding to queries.

[1486] Input: Data in a clean database

[1487] Output: A trained generative model

[1488] Step 4: Receiving a user query

[1489] User

[1490] Users enter queries through a dedicated web interface, for example, "How do I set up Wi-Fi?"

[1491] Terminal

[1492] The terminal uses JavaScript or Ajax to convert the query content into JSON format and send it to the server's API endpoint.

[1493] Input: User's inquiry

[1494] Output: Query sent to the server in JSON format

[1495] Step 5: Parsing and preprocessing the query

[1496] server

[1497] The server parses the received JSON query and converts it into an appropriate format. Specifically, it uses Python to parse the JSON data and performs the necessary preprocessing before passing it to the generative model.

[1498] Input: JSON format query

[1499] Output: Preprocessed query data

[1500] Step 6: Generate a response

[1501] server

[1502] The server then passes the preprocessed query to the generative model, which then generates the appropriate response. For example, the generative model retrieves information about "how to set up Wi-Fi" from a database and generates the appropriate instructions.

[1503] Input: Preprocessed query data

[1504] Output: The generated response

[1505] Step 7: Emotion Recognition

[1506] server

[1507] The generated response is passed to an emotion engine (e.g., Google Cloud Natural Language API) to analyze the user's emotions. Emotions such as anger, sadness, and joy are detected from the text data. The emotion engine returns the analysis results to the server.

[1508] Input: Generated response

[1509] Output: User sentiment analysis results

[1510] Step 8: Adjusting the response

[1511] server

[1512] The server adjusts the tone and content of the response based on the results of the emotion engine. For example, if the user is expressing anger, the tone of the response should be changed to be more polite. It also includes a means to escalate the response to a human operator if necessary.

[1513] Input: User sentiment analysis results

[1514] Output: Adjusted response

[1515] Step 9: Sending a Response

[1516] server

[1517] The server then converts the adjusted response back into JSON format and sends it to the device using a secure protocol such as HTTPS.

[1518] Input: Adjusted response

[1519] Output: JSON formatted response sent to the terminal

[1520] Step 10: View the response

[1521] Terminal

[1522] The terminal displays the received response on the user interface, allowing the user to confirm the displayed response and proceed with the operation.

[1523] Input: The response sent by the server

[1524] Output: The response displayed in the user interface

[1525] Step 11: Logging queries and responses

[1526] server

[1527] The server records user queries and generated responses in a log file, which provides data for later performance evaluation and model improvement.

[1528] Input: The query and generated response

[1529] Output: Data recorded in the log file

[1530] Step 12: Evaluate the model's performance

[1531] server

[1532] The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine, including response accuracy, response time, and emotion recognition accuracy. If necessary, it collects new data and retrains the model.

[1533] Input: Log data

[1534] Output: Performance evaluation results and retrained model

[1535] Step 13: Gather user feedback

[1536] User

[1537] Users input their satisfaction and feedback regarding the provided responses through a dedicated UI and send it to the server via their device.

[1538] server

[1539] The server analyzes the collected feedback and uses it to improve the system.

[1540] Input: User feedback

[1541] Output: Improved system

[1542] Through these steps, users can receive quick and emotionally sensitive answers.

[1543] (Application example 2)

[1544] 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."

[1545] Conventional content distribution services have the problem of reducing user satisfaction because they recommend content without considering the user's emotional state. Furthermore, they often return a uniform response to user inquiries, failing to consider the user's emotions. This results in issues such as a decrease in service usage and an increased likelihood that users will switch to other services.

[1546] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1547] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving inquiries entered by users and passing them to the trained model, means for returning responses generated from the model to the users, means for recording the inquiries and the generated responses and periodically evaluating the performance of the model, emotion recognition means for analyzing emotions from the user's inquiry text and adjusting the response content, and means for recommending appropriate content according to the user's emotions in a content distribution service, thereby enabling responses and content recommendations that take user emotions into consideration.

[1548] "Handling information terminal" refers to a device used by a user to perform operations or make inquiries.

[1549] An "instruction manual" is a document or material that explains how to use an information terminal and its functions.

[1550] "FAQ data" is data that compiles information about questions and doubts that users frequently have.

[1551] A "generative model" is an algorithm or machine learning model that generates responses to queries based on collected data.

[1552] "Training means" refers to the method or process for training a generative model using collected data.

[1553] A "user query" is a question or request made by a user to a system for information or support.

[1554] A "response" is an answer or instruction that a generative model generates in response to a user query.

[1555] A "database" is a system for systematically storing and managing collected instruction manuals and frequently asked questions.

[1556] "Emotion recognition means" refers to the technology or function for analyzing and extracting emotions from the user's inquiry text.

[1557] A "content distribution service" is a service that provides users with content such as videos, music, and articles.

[1558] "Feedback means" refers to a method or function for collecting user evaluations and opinions on responses and using them to improve the system.

[1559] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. This system performs the following processes from the perspectives of the server, terminal, and user.

[1560] 1. Data collection and learning phase

[1561] server

[1562] First, the server collects instruction manuals and frequently asked questions data and stores them in a database. The collected data is pre-cleaned to remove duplicate data and unnecessary information. After data cleaning is complete, the data is stored in the database. Next, the server analyzes the pre-processed data using a generative model to train the model. A specific machine learning algorithm is used for this training. Once the model is trained, it serves as the basis for responding quickly and accurately to future inquiries.

[1563] 2. Query Receipt and Response Generation Phase

[1564] User

[1565] If a user has a question about how to operate the system or a specific function, they can input a query into the system through a dedicated user interface, such as "What movies do you recommend?"

[1566] Terminal

[1567] User inquiries received via the user interface are sent by the terminal to the server. Specifically, the inquiry content is converted into JSON format or similar and sent to the server's API.

[1568] server

[1569] The server analyzes the received query, performs necessary preprocessing, and then passes it to the generative model, which generates an appropriate answer based on the query and returns the response to the server.

[1570] 3. Emotion recognition and response adaptation using an emotion engine

[1571] server

[1572] Before sending the generated response, the server passes the query content to an emotion engine to recognize the user's emotion. The emotion engine uses natural language processing technology to analyze emotions from the user's text data. As a result, it determines whether the user is feeling anger, sadness, joy, anticipation, or other emotions. Based on the emotion recognized by the emotion engine, the server adapts the tone and content of the response. For example, if the user is expressing anger, the tone of the response will be adjusted to be more polite. It also includes a function to escalate to a human operator if necessary.

[1573] 4. Sending the Response

[1574] server

[1575] Once the tailored response is complete, the server sends it to the terminal for transmission back to the user, using a secure protocol.

[1576] Terminal

[1577] The terminal displays the response received from the server on the user interface, allowing the user to perform operations based on the response.

[1578] 5. Performance evaluation and improvement

[1579] server

[1580] The server logs user queries and generated responses. This accumulates data for later performance evaluation and model improvement. It periodically analyzes the log data to evaluate the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[1581] User

[1582] The user inputs feedback on the provided responses and sends it to the server via the terminal, which helps improve the system.

[1583] Specific examples

[1584] For example, if a user inquires, "What movies do you recommend?", the flow is as follows:

[1585] 1. User enters a query

[1586] The user enters a question.

[1587] 2. The device sends a query to the server

[1588] The user terminal sends this query to the server's API.

[1589] 3. The server receives the query and passes it to the generative model.

[1590] The server analyzes the received query and inputs it into the generative model.

[1591] 4. The generative model generates a response

[1592] The generative model retrieves information about "movie recommendation" from the database and generates appropriate steps.

[1593] 5. The server passes the query to the emotion engine

[1594] The server passes the query content to the emotion engine to recognize the user's emotion.

[1595] 6. Emotion engine recognizes user emotions

[1596] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[1597] 7. The server adjusts the response

[1598] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[1599] 8. The server sends the response to the device

[1600] The server sends the tailored response to the terminal for transmission back to the user.

[1601] 9. The device displays the response

[1602] The terminal displays the response on the user interface, and the user performs an operation based on the response.

[1603] This system allows users to receive prompt and emotionally sensitive answers, improving their satisfaction with the service.

[1604] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1605] Step 1:

[1606] User enters a query

[1607] A user types a question into a smartphone application, for example, "What movies do you recommend?" This query is entered through a user interface, and the entered text is stored on the device as raw text.

[1608] Input: User query text (e.g., "What movies do you recommend?")

[1609] Output: Raw query text

[1610] Step 2:

[1611] The device sends a query to the server

[1612] The terminal converts user inquiries received via the user interface into JSON format and sends it to the server's API, allowing the server to receive the inquiry content.

[1613] Input: Raw query text

[1614] Output: Data converted to JSON format (e.g., {"query": "What movies do you recommend?"})

[1615] Step 3:

[1616] The server receives the query and passes it to the generative model.

[1617] The server parses the received JSON data and extracts the query content, which is then passed to the generative model, which then generates an appropriate response based on this data.

[1618] Input: Query data in JSON format

[1619] Output: Generated response (e.g., "I recommend 'Hero Movie' as an uplifting action movie.")

[1620] Step 4:

[1621] The generative model generates the response

[1622] The generative model generates an appropriate answer based on the query. This process involves retrieving relevant information from a database and performing data processing and calculations to generate the appropriate response.

[1623] Input: The query passed to the generative model

[1624] Output: Correct response

[1625] Step 5:

[1626] The server passes the query to the emotion engine

[1627] The server passes the generated response and the original query text to the emotion engine to recognize the user's emotion. The emotion engine uses natural language processing techniques to analyze the user's emotional state from the text.

[1628] Input: Generated response and original query text

[1629] Output: Emotion recognition result (e.g. positive, negative, neutral)

[1630] Step 6:

[1631] Emotion engine recognizes user emotions

[1632] The emotion engine analyzes the user's emotion from the input text and returns the state to the server. In this step, data processing and calculations are performed to extract the user's emotion using an analysis algorithm and return the result.

[1633] Input: Query text and generated response

[1634] Output: Emotion recognition result (e.g., positive)

[1635] Step 7:

[1636] The server adjusts the response

[1637] Based on the results of the emotion engine, the server adjusts the generated response, ensuring that the response is of the appropriate tone and content depending on the user's emotional state, and escalating to a human operator if necessary.

[1638] Input: Generated response sentences and emotion recognition results

[1639] Output: Tailored response (e.g., "I recommend 'Hero Movie' as an uplifting action movie. I'm glad you liked it!")

[1640] Step 8:

[1641] The server sends a response to the device

[1642] The server sends the tailored response to the terminal for transmission back to the user, using a secure protocol.

[1643] Input: Adjusted response sentence

[1644] Output: Response data sent to the terminal

[1645] Step 9:

[1646] The terminal displays the response

[1647] The terminal displays the response received from the server on the user interface, allowing the user to perform operations and make selections based on the response.

[1648] Input: Response data sent from the server

[1649] Output: The response text displayed in the user interface

[1650] 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.

[1651] 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.

[1652] 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.

[1653] [Fourth embodiment]

[1654] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1655] 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.

[1656] 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).

[1657] 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.

[1658] 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.

[1659] 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).

[1660] 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.

[1661] 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.

[1662] 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.

[1663] 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.

[1664] 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.

[1665] 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.

[1666] 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."

[1667] The present invention is an operation method support system that uses a generative model to provide convenience to both salespeople and customers. The overall flow of this system will be explained from the perspectives of the server, terminals, and users.

[1668] 1. Data collection and learning phase

[1669] server

[1670] The server first collects instruction manuals and FAQ (frequently asked questions) data for all information devices handled from the Internet and internal databases, and then centrally manages them. Since this data is used as learning material for the model, it is cleaned in advance to remove duplicate data and unnecessary information. After data cleaning is complete, the data is organized and saved.

[1671] The server then trains the pre-processed data with a generative model using machine learning algorithms, which then serves as the basis for quickly and accurately responding to future queries.

[1672] 2. Inquiry Receipt and Response Phase

[1673] User

[1674] When a user has a question about an operation method or a specific function, they input their query into the system through a dedicated user interface, which may be provided in the form of a web browser, a mobile application, or a help desk.

[1675] Terminal

[1676] User inquiries received via the user interface are sent from the terminal to the server. The terminal converts the inquiry content into a format such as JSON and sends it to the server's data reception API.

[1677] server

[1678] The server analyzes the received query and passes it to a trained generative model. The generative model generates an appropriate answer based on the query, constructs a response, and returns it to the server. The server then sends this response back to the user, allowing the user to receive a quick and accurate answer.

[1679] 3. Performance evaluation and improvement phase

[1680] server

[1681] The server records all queries and responses in log files. Periodically, it analyzes these logs to evaluate the model's performance. This evaluation involves scoring the generated responses to ensure their accuracy. Based on the evaluation results, if the model's accuracy falls below a predetermined standard, it may automatically initiate a process to collect additional data and retrain.

[1682] User

[1683] Additionally, users can enter feedback on the answers they provide, which is sent to the server and used as data for future model improvements.

[1684] Specific examples

[1685] For example, suppose a user asks, "How do I set up Wi-Fi?" The flow is shown below.

[1686] 1. User enters a query

[1687] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[1688] 2. The device sends a query to the server

[1689] The browser on the user's device sends this query to the server's API.

[1690] 3. The server receives the query and passes it to the generative model.

[1691] The server analyzes the received query and inputs it into the generative model.

[1692] 4. The generative model generates a response

[1693] The generative model retrieves information about "how to set up Wi-Fi" from a database and generates appropriate instructions.

[1694] 5. The server sends the response back to the user

[1695] The server generates a response and sends it back to the user, who displays it in their browser.

[1696] 6. The user receives the response and performs the action

[1697] The user follows the displayed steps to set up Wi-Fi.

[1698] This allows users to receive prompt and appropriate responses, and frees sales representatives from having to deal with inquiries about operation methods, allowing them to focus on more important tasks.

[1699] The processing flow will be explained below.

[1700] Step 1:

[1701] The server collects instruction manuals and FAQ data from the information terminals it handles and stores them in a database, including scraping data from the Internet and retrieving it from its internal database.

[1702] Step 2:

[1703] The server preprocesses the collected data, specifically by performing text cleaning and data normalization to remove duplicate data and unnecessary information.

[1704] Step 3:

[1705] The server trains a generative model based on the preprocessed data, using machine learning algorithms to extract the necessary knowledge from the data and train the model.

[1706] Step 4:

[1707] The server stores the trained generative model and prepares it for generating responses to queries.

[1708] Step 5:

[1709] The user inputs an inquiry about operation methods through a dedicated user interface, for example, "Please tell me how to set up Wi-Fi."

[1710] Step 6:

[1711] The terminal converts the user's inquiry into JSON format or similar and sends it to the server's API. This transmission generally uses the HTTPS protocol.

[1712] Step 7:

[1713] The server parses the received query, performs any necessary preprocessing, and then passes it to the generative model. Parsing includes tokenizing and encoding the text.

[1714] Step 8:

[1715] The server receives the response from the generative model and transforms it into a format that is easy for the user to understand, for example, organizing the steps as a numbered list.

[1716] Step 9:

[1717] The server sends the converted response back to the user via the API to the device, again using a secure protocol.

[1718] Step 10:

[1719] The device displays the response received from the server and provides it to the user, either on a web page or within an application.

[1720] Step 11:

[1721] Performs actions based on the answers the user provides, for example, walking through Wi-Fi setup steps to resolve the issue.

[1722] Step 12:

[1723] The server logs user queries and the responses it generates, providing data for later performance evaluation and model improvement.

[1724] Step 13:

[1725] The server periodically analyzes the log data and evaluates the performance of the generative model using metrics such as accuracy and response time.

[1726] Step 14:

[1727] If the accuracy of the model falls below a predetermined standard, the server collects new data and retrains the model, ensuring that it always provides highly accurate responses.

[1728] Step 15:

[1729] The user inputs feedback on the provided responses and sends it to the server via the terminal, which is used to improve the system.

[1730] As a result, this system can reduce the burden on sales staff and improve customer satisfaction.

[1731] Example 1

[1732] 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."

[1733] Many modern information terminals have complex operation methods and multifunctional settings, which often lead to users becoming confused and not receiving appropriate support. Sales representatives are also overwhelmed by responding to a variety of inquiries, preventing them from concentrating on their core duties. Furthermore, traditional FAQs and instruction manuals only provide static information, making it difficult to provide dynamic and prompt support.

[1734] 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.

[1735] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for cleaning and preprocessing the collected data, means for training the preprocessed data using a generative model, means for receiving inquiries entered by users, converting the inquiry content into JSON format, and sending the JSON format to the server, means for inputting the user's inquiry into the trained generative model and obtaining a response generated from the model, means for returning the obtained response to the user interface, and means for recording the inquiries and the generated responses and periodically evaluating the performance of the model. This allows users to receive prompt and appropriate support, and frees sales representatives from responding to inquiries so that they can focus on their core business.

[1736] An "information terminal" is a device or system that provides information about an electronic device or system operated by a user.

[1737] An "instruction manual" is a document that describes the basic operation methods, setting procedures, precautions, etc. of an information terminal.

[1738] "Frequently asked questions data" is data that compiles questions frequently asked by users and the answers to those questions.

[1739] A "database" is a system for efficiently storing, managing, and searching large amounts of data.

[1740] "Data cleanup" is the process of removing redundancies and noise from collected data and preparing it in a format suitable for analysis.

[1741] "Preprocessing" is a series of steps to convert data into a format suitable for a generative model.

[1742] A "generative model" is a model that uses machine learning algorithms to generate optimal answers from given data.

[1743] "Training" is the process of optimizing the parameters of a generative model using collected data to train the model.

[1744] A "user interface" is an interactive screen or application that allows a user to access and query a system.

[1745] "JSON format" is a data format that structures data in text format and makes it easily exchangeable.

[1746] A "server" is a computer system that processes data and provides services over a network.

[1747] An "inquiry" is a question or request made by a user to the system.

[1748] A "response" is an answer that a system provides to a user's query.

[1749] "Recording" means saving the contents of inquiries and responses in a log file or the like.

[1750] "Performance evaluation" is the process of evaluating the accuracy and appropriateness of the responses of a generative model.

[1751] This invention is an operation method support system that uses a generative AI model to provide convenience to both sales representatives and customers. In this system, the server, terminal, and user each play specific roles, and provide quick and accurate answers to user inquiries.

[1752] Data collection and preprocessing

[1753] server

[1754] The server collects the user manuals and FAQ data for the information terminals from the internet and an internal database. It also cleans the data, removing duplicate data and noise. Specifically, it uses Python scripts and BeautifulSoup to scrape the web, and the Pandas library to process the data. The cleaned data is then stored using a large-scale data processing framework (e.g., Hadoop).

[1755] Training generative AI models

[1756] server

[1757] The server trains a generative model using the collected and preprocessed data. TensorFlow or PyTorch is used for this training. The training data is split using a tokenizer and input to a neural network. A deep learning framework is used to optimize the model parameters. In particular, the Transformers architecture is commonly used.

[1758] Receiving inquiries via a user interface

[1759] User

[1760] Users input questions about operations and specific functions through a user interface, which is provided by a web browser or a mobile application, such as Google Chrome or Safari.

[1761] Specific examples

[1762] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[1763] Analysis of inquiries

[1764] Terminal

[1765] The device converts the user's inquiry into JSON format and sends it to the server's receiving API. For example, it generates JSON data using a JavaScript or Python library and sends it to the server via an HTTP request. Browsers used include Firefox and Google Chrome.

[1766] Answer generation using generative AI models

[1767] server

[1768] The server inputs the received query into a generative model. This process involves encoding the query and converting it into a format suitable for the generative model. The generative AI model (e.g., GPT-3) generates an appropriate answer based on the query. The generated answer is then formatted.

[1769] Specific examples

[1770] The server uses the generative model to generate instructions on how to set up Wi-Fi.

[1771] Returning answers to users

[1772] server

[1773] The server generates an answer and sends it back to the user. The response data is organized in JSON format and sent to the device in an HTTP response.

[1774] Terminal

[1775] The device analyzes the returned data and displays it in a user interface, using HTML and JavaScript to display the answers in the browser.

[1776] Performance evaluation and feedback collection

[1777] server

[1778] The server logs all queries and responses and periodically analyzes them. The log data is evaluated using Python and R to measure the accuracy of the responses. Users can enter feedback on the answers provided, and this data is also collected.

[1779] This process allows users to receive prompt and appropriate answers, and frees up sales representatives from the burden of handling inquiries. The system continuously collects data to improve the model's performance, providing a high-quality user experience.

[1780] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1781] Step 1:

[1782] Data collection and preprocessing

[1783] The server collects the user manuals and frequently asked questions (FAQ data) for the information terminals from the internet and an internal database. The input uses the URL of the online user manual and the access information for the internal database. Specifically, it performs web scraping using a Python script and BeautifulSoup to extract the data. After the data is collected, it is formatted using Pandas to remove duplicate data and noise. The clean data is generated as output and stored in Hadoop or another storage system.

[1784] Step 2:

[1785] Training generative AI models

[1786] The server trains a generative model using the collected and preprocessed data. It uses clean training data as input. Specifically, it tokenizes the training data using the TensorFlow or PyTorch framework and feeds it to a neural network. The selected deep learning model (e.g., Transformers architecture) learns patterns in the data and finds optimal parameters. The output is a trained generative model.

[1787] Step 3:

[1788] Receiving inquiries via a user interface

[1789] The user inputs questions about operation methods or specific functions through the user interface. The user's question is used as input. Specifically, the user enters a question into an inquiry form in a web browser such as Google Chrome or Safari and clicks the submit button. The user's question data is sent to the device as output.

[1790] Step 4:

[1791] Analysis of inquiries

[1792] The terminal converts the user's inquiry into JSON format and sends it to the server's receiving API. The user's question received from the web browser is used as input. Specifically, it generates JSON data using JavaScript or Python libraries and sends it to the server via an HTTP request. The JSON-formatted inquiry data is sent to the server as output.

[1793] Step 5:

[1794] Answer generation using generative AI models

[1795] The server inputs the received query content into the generative model. Parsed JSON data is used as input. Specifically, the server encodes the query content and converts it into a format suitable for the generative model. The generative AI model (e.g., GPT-3) generates an appropriate answer based on the query content. The output is the generated response text.

[1796] Step 6:

[1797] Returning answers to users

[1798] The server returns the generated answer to the user. The generated response text is used as input. Specifically, the server formats the response data in JSON format and sends it to the terminal as an HTTP response. The terminal receives the returned data and displays it in the user interface. As output, the final response is displayed on the user's browser.

[1799] Step 7:

[1800] Performance evaluation and feedback collection

[1801] The server logs all queries and responses and periodically analyzes them. The query and response log data is used as input. Specific operations include evaluating the log data using Python or R to measure the model's performance. Users can provide feedback on the answers provided, and this data is also collected. The output is the model's performance evaluation results and user feedback.

[1802] This processing step allows the system to respond quickly and accurately to user inquiries and to continually collect data to improve performance.

[1803] (Application example 1)

[1804] 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."

[1805] Conventional systems have struggled to respond quickly and accurately to user inquiries about content search and playback settings. Content distribution services, in particular, require appropriate recommendations based on users' viewing history and recent popular content, but there has been a lack of efficient ways to provide this information. Furthermore, systems lack the ability to automatically update data and retrain when system performance deteriorates.

[1806] 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.

[1807] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving queries entered by users and passing them to the trained model, means for returning responses generated from the model to the users, means for recording queries and the generated responses and periodically evaluating the performance of the model, and means for generating content recommendations based on the user's viewing history and recent popular content in content distribution, thereby enabling users to receive appropriate recommendations and quick responses regarding content they want to view.

[1808] The term "information terminals" refers to various electronic devices and content playback terminals used by users.

[1809] "Instruction manual" refers to a document that contains detailed information about the operation and functions of the information terminal.

[1810] "Frequently asked questions data" refers to documents or data that systematically compile answers to questions and problems that users frequently have.

[1811] A "database" refers to a system for efficiently storing, searching, and managing collected information and data.

[1812] "Generative model" refers to a system trained using machine learning algorithms to generate appropriate responses to user queries.

[1813] "Training" refers to the process of using collected data to train a generative model and improve its accuracy.

[1814] "Content distribution" refers to the service of providing digital media such as movies, TV shows, and music to users via the Internet.

[1815] "Viewing history" refers to a record of content that a user has viewed in the past.

[1816] "Popular content" refers to digital media that is viewed by many users and has received high ratings.

[1817] "Content recommendation" refers to suggesting the next piece of media to watch based on the user's viewing history and popular content.

[1818] This invention applies a content distribution service to an operation method support system that uses a generative model to provide convenience to both sales representatives and customers. The details of this system will be explained in three phases: data collection and learning phase, inquiry reception and response phase, and performance evaluation and improvement phase.

[1819] 1. Data collection and learning phase

[1820] server

[1821] The server first collects instruction manuals and FAQs (frequently asked questions) for the information terminals (various electronic devices and content playback devices used by users) from the Internet and an internal database, and then centrally manages them. Since this data is used as learning material for the model, it is cleaned in advance to remove duplicate data and unnecessary information. Python-based scripts and data processing libraries (Pandas, NumPy, etc.) are used for data cleaning.

[1822] The server then trains the preprocessed data using a generative model (e.g., GPT-4), which lays the foundation for the model to respond quickly and accurately to user queries.

[1823] 2. Inquiry Receipt and Response Phase

[1824] User

[1825] Users access the "Content Help AI" app on their smartphone or head-mounted display (HMD) and input a query by voice or text, such as "What movie would you recommend I watch next?"

[1826] Terminal

[1827] The device (smartphone or HMD) converts the input voice data into text (using voice recognition software) and converts it into JSON format, which is then sent to the server.

[1828] server

[1829] The server analyzes the received query and passes it to the generative model, which generates an appropriate response from the database based on the prompt "Please provide information about recommended movies that the user should watch." For example, the prompt might look like this:

[1830] "What movie should users watch next? Create a recommendation list that takes into account recent popular movies and the user's viewing history."

[1831] After the generative model generates a response, the server sends the response back to the user device and displays it on the user's smartphone or HMD.

[1832] 3. Performance evaluation and improvement phase

[1833] server

[1834] The server records all queries and the generated responses in a log file, allowing the system's performance to be evaluated periodically. If performance falls below a predetermined standard, the server collects additional data and retrains the model.

[1835] User

[1836] Additionally, users can provide feedback on the answers they provide, which is also sent to the server and used as data for future model improvements.

[1837] As a concrete example, when a user asks, "What movie would you recommend I watch next?" the system follows these steps:

[1838] 1. The user uses the app to enter a question by voice.

[1839] 2. The device converts the voice into text and sends it to the server.

[1840] 3. The server inputs a prompt sentence into the generative model.

[1841] 4. The generative model generates a list of appropriate movie recommendations.

[1842] 5. The server generates a response and sends it back to the user, who displays it in the app.

[1843] This process allows users to get relevant and fast recommendations for content they might want to watch.

[1844] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1845] Step 1:

[1846] User enters a query

[1847] Users use the "Content Help AI" app on their smartphones or head-mounted displays (HMDs) to input questions about the content they want to watch. For example, they might say, "What movie would you recommend I watch next?" This input data is in audio format.

[1848] Step 2:

[1849] The device converts the voice data into text

[1850] The device uses speech recognition software to convert the voice data into text data. The input to this process is the user's voice data, and the output is the question in text form.

[1851] Step 3:

[1852] The device converts the text data into JSON format and sends it to the server.

[1853] The terminal converts text data into JSON format. In this conversion process, the text data is converted into a structured JSON object. The converted JSON data is sent to the server. The input of this process is text data, and the output is JSON format data.

[1854] Step 4:

[1855] The server receives and parses the JSON data.

[1856] The server receives and analyzes the JSON data sent from the terminal. In this analysis process, the JSON object is analyzed to extract the user's query. The input of this process is the JSON data, and the output is the extracted data of the query.

[1857] Step 5:

[1858] The server inputs a prompt to the generative model.

[1859] The server inputs a prompt sentence into the generative model based on the extracted query content. An example of a prompt sentence is, "Please tell me what movie the user should watch next. Please create a recommendation list taking into account recent popular movies and the user's viewing history." The inputs to this process are the extracted data and the prompt sentence, and the output is a response generated by the generative model.

[1860] Step 6:

[1861] The generative model generates the response

[1862] A generative model (e.g., GPT-4) generates a response based on the prompt sentence. This response generation process creates an appropriate recommendation list based on the viewing history and popular content stored in the database. The input of this process is the prompt sentence, and the output is the generated response data.

[1863] Step 7:

[1864] The server generates a response and sends it back to the user.

[1865] The server receives the response generated from the generative model and sends it back to the user terminal. In this process, the generated response data is converted into an appropriate format and sent to the terminal. The input of this process is the generated response data, and the output is the data sent to the user terminal.

[1866] Step 8:

[1867] The terminal displays the response data to the user

[1868] The device analyzes the response data received from the server and displays it to the user. This display process converts the generated response into a format that is easy for the user to understand and displays it within the app. The input of this process is the response data received from the server, and the output is the information displayed on the user's screen.

[1869] In this way, users can get appropriate and quick recommendations and operating procedures for the content they want to watch.

[1870] 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.

[1871] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. The overall flow of this system will be specifically explained from the perspectives of the server, terminal, and user.

[1872] 1. Data collection and learning phase

[1873] server

[1874] The server first collects instruction manuals and FAQ (frequently asked questions) data for all information devices it handles from the Internet and an internal database, and then centrally manages them. This data is used as learning material for the generative model. The collected data is pre-cleaned to remove duplicate data and unnecessary information. After data cleaning is complete, the data is stored in the database.

[1875] The server then trains the pre-processed data with a generative model using machine learning algorithms, which then serves as the basis for quickly and accurately responding to future queries.

[1876] 2. Query Receipt and Response Generation Phase

[1877] User

[1878] If a user has a question about how to operate the system or a specific function, they can enter a query into the system through a dedicated user interface, such as "How do I set up Wi-Fi?"

[1879] Terminal

[1880] User inquiries received via the user interface are sent by the terminal to the server. Specifically, the inquiry content is converted into JSON format or similar and sent to the server's API.

[1881] server

[1882] The server analyzes the received query, performs necessary preprocessing, and then passes it to the generative model, which generates an appropriate answer based on the query and returns the response to the server.

[1883] 3. Emotion recognition and response adaptation using an emotion engine

[1884] server

[1885] Before sending the generated response, the server passes the query content to an emotion engine to recognize the user's emotion. The emotion engine uses natural language processing technology to analyze emotions from the user's text data, and determines whether the user is feeling anger, sadness, joy, anticipation, or other emotions.

[1886] Based on the emotions recognized by the emotion engine, the server will adapt the tone and content of the response. For example, if the user is angry, the tone of the response will be more polite. It also includes the ability to escalate to a human operator if necessary.

[1887] 4. Sending the Response

[1888] server

[1889] Once the tailored response is complete, the server sends it to the terminal for transmission back to the user, using a secure protocol.

[1890] Terminal

[1891] The terminal displays the response received from the server on the user interface, allowing the user to perform operations based on the response.

[1892] 5. Performance evaluation and improvement

[1893] server

[1894] The server logs user queries and generated responses. This accumulates data for later performance evaluation and model improvement. It periodically analyzes the log data to evaluate the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[1895] User

[1896] The user inputs feedback on the provided responses and sends it to the server via the terminal, which helps improve the system.

[1897] Specific examples

[1898] For example, if a user asks, "How do I set up Wi-Fi?", the flow is as follows:

[1899] 1. User enters a query

[1900] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[1901] 2. The device sends a query to the server

[1902] The browser on the user's device sends this query to the server's API.

[1903] 3. The server receives the query and passes it to the generative model.

[1904] The server analyzes the received query and inputs it into the generative model.

[1905] 4. The generative model generates a response

[1906] The generative model retrieves information about "how to set up Wi-Fi" from a database and generates appropriate instructions.

[1907] 5. The server passes the query to the emotion engine

[1908] The server passes the query content to the emotion engine to recognize the user's emotion.

[1909] 6. Emotion engine recognizes user emotions

[1910] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[1911] 7. The server adjusts the response

[1912] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[1913] 8. The server sends the response to the device

[1914] The server sends the tailored response to the terminal for transmission back to the user.

[1915] 9. The device displays the response

[1916] The terminal displays the response on the user interface, and the user performs an operation based on the response.

[1917] This system allows users to receive quick and sensitive responses, freeing sales representatives from having to deal with inquiries about how to use the system, allowing them to focus on more important tasks.

[1918] The processing flow will be explained below.

[1919] Step 1:

[1920] The server collects the instruction manuals and FAQ data of the information terminals it handles and stores them in a database. This is done by scraping data from the internet and retrieving data from the company's internal database. The collected data is pre-processed to remove duplicate and irrelevant data.

[1921] Step 2:

[1922] The server trains a generative model based on the preprocessed data, using machine learning algorithms to analyze the data and train the model to generate appropriate responses to queries.

[1923] Step 3:

[1924] The server uses an emotion engine to train the system to recognize user emotions based on the text data contained in the user's inquiry. The emotion engine uses natural language processing technology to build an emotion analysis model.

[1925] Step 4:

[1926] The user inputs an inquiry about operation methods through a dedicated user interface. For example, the user inputs a specific question such as "How do I set up Wi-Fi?"

[1927] Step 5:

[1928] The terminal converts the query entered by the user into an appropriate format, such as JSON, and sends it to the server's API using the HTTPS protocol.

[1929] Step 6:

[1930] The server parses the received query and performs any necessary preprocessing, including tokenizing and encoding the text, before feeding the preprocessed data into the generative model.

[1931] Step 7:

[1932] The server receives the response from the generative model and inputs it into the emotion engine to analyze the user's emotion. The emotion engine recognizes the user's emotion and returns the result to the server.

[1933] Step 8:

[1934] The server adapts the tone and content of the response based on the results of the emotion engine. For example, if the user is angry, the tone of the response will be more polite. It will also escalate the situation to a human operator if necessary.

[1935] Step 9:

[1936] The server sends the tailored response to the terminal for transmission back to the user, using a secure communications protocol.

[1937] Step 10:

[1938] The terminal displays the response received from the server on the user interface, and the user checks the displayed response and performs an operation based on it.

[1939] Step 11:

[1940] Perform actions based on the responses the user provides, for example, stepping through Wi-Fi setup procedures to resolve the issue.

[1941] Step 12:

[1942] The server logs all user queries and generated responses, providing data for later performance evaluation and model improvement.

[1943] Step 13:

[1944] The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine, including response accuracy, emotion recognition precision, and response time.

[1945] Step 14:

[1946] If the server finds that the model's performance falls below a predetermined standard, it collects new data and retrains the model, using the most recent query data and user feedback.

[1947] Step 15:

[1948] The user inputs feedback on the provided responses and sends it to the server via the terminal, which is used to improve the performance of the system.

[1949] As a result, this system can reduce the burden on sales staff and improve customer satisfaction.

[1950] Example 2

[1951] 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."

[1952] In modern information systems, it is difficult for users to obtain prompt and appropriate responses when inquiring about operation methods or specific functions. Furthermore, responses that take emotion into consideration are not provided, resulting in a poor user experience. Furthermore, it is difficult to generate adaptive responses based on the content of the inquiry, and there is a need for improved user satisfaction. Therefore, a system that solves these issues and provides efficient and emotion-sensitive responses is needed.

[1953] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1954] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving inquiries entered by a user and passing them to the trained model, means for analyzing the content of the user's inquiry to generate a prompt, means for analyzing the content of the inquiry and performing necessary preprocessing, means for passing the generated response to an emotion engine to recognize the user's emotion, means for adapting the tone and content of the response based on the recognized emotion, means for escalating the response to a human operator as necessary, means for returning the response generated from the model to the user, and means for recording the inquiries and the generated responses in a log and periodically evaluating the performance of the model. This makes it possible to provide quick and appropriate responses that take the user's emotions into consideration and improve user satisfaction.

[1955] "Information handling terminal" refers to all electronic devices that users can use to operate and configure.

[1956] An "instruction manual" refers to a document that describes how to use and set up the information terminal, as well as precautions to take.

[1957] "FAQ data" refers to data that compiles answers to questions and inquiries frequently asked by users.

[1958] A "database" refers to a system for efficiently storing, searching, and managing large amounts of data.

[1959] A "generative model" refers to an algorithm or machine learning model that automatically generates sentences or responses based on data.

[1960] "User" refers to an individual or organization that uses the system or service.

[1961] "Inquiry" refers to a question or concern a user has about a system or service.

[1962] A "prompt" refers to an input sentence that prompts a generative model to generate an appropriate response.

[1963] An "emotion engine" refers to a system or technology that analyzes emotions from a user's text data and recognizes their emotional state.

[1964] An "operator" refers to a human worker in charge of operating and managing a system or service.

[1965] "Log" refers to data that records system operation history and events.

[1966] "Performance evaluation" refers to methods for evaluating the accuracy of systems and models, response time, and emotion recognition accuracy.

[1967] "Preprocessing" refers to the process of converting data into the required format or removing noise before inputting it into a generative model.

[1968] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. The overall flow of this system will be specifically explained from the perspectives of the server, terminal, and user.

[1969] Data collection and learning phase

[1970] server

[1971] The server first collects the user's manual and FAQ data for the information terminal from the internet or an internal database. Web scraping technology (e.g., BeautifulSoup or Selenium) is used for collection. The collected data is centrally managed in CSV or JSON format. Next, data cleaning is performed, using regular expressions and database queries (e.g., SQL) to remove duplicate data and unnecessary information. The cleaned data is stored in a dedicated database. The server then trains the preprocessed data using a generative model (e.g., GPT-3). Machine learning frameworks such as TensorFlow and PyTorch are used for training. The trained model serves as the basis for responding quickly and accurately to future inquiries.

[1972] Query Receipt and Response Generation Phase

[1973] User

[1974] Users enter a query through a dedicated web interface, for example, "How do I set up Wi-Fi?"

[1975] Terminal

[1976] The user's device uses JavaScript or Ajax to convert the inquiry content into JSON format and sends it to the server's API endpoint.

[1977] server

[1978] The server parses the received JSON-formatted query, performs any necessary preprocessing, and then passes it to the generative model. The generative model generates an appropriate answer based on the query content. For example, the generative model retrieves information about "how to set up Wi-Fi" from a database and generates the appropriate steps.

[1979] Emotion engine recognizes emotions and adapts responses

[1980] server

[1981] The generated response is passed to an emotion engine (e.g., Google Cloud Natural Language API) to analyze the user's emotions. Emotions such as anger, sadness, and joy are detected from the text data. The emotion engine returns the analysis results to the server. The server then adjusts the tone and content of the response based on the analysis results. For example, if the user expresses anger, the tone of the response needs to be changed to be more polite. It also includes a means to escalate the response to a human operator if necessary.

[1982] Sending a Response

[1983] server

[1984] The server then converts the adjusted response back into JSON format and sends it to the device using a secure protocol such as HTTPS.

[1985] Terminal

[1986] The terminal displays the received response on the user interface, and the user can proceed with the operation by checking the displayed response.

[1987] Performance evaluation and improvement

[1988] server

[1989] The server records user queries and generated responses in a log file. This accumulates data for later performance evaluation and model improvement. The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[1990] User

[1991] Users can input their satisfaction with the responses and feedback through a dedicated UI, which is then sent to the server via their device. This feedback is used to improve the system.

[1992] Specific examples

[1993] For example, if a user asks, "How do I set up Wi-Fi?", the specific operation is as follows.

[1994] 1. User enters a query

[1995] A user types the question "How do I set up Wi-Fi" into a web browser interface.

[1996] 2. The device sends a query to the server

[1997] This query is converted to JSON format using JavaScript and sent to the server's API endpoint via Ajax.

[1998] 3. The server receives and analyzes the query

[1999] Analyze the received JSON data using Python and perform the necessary preprocessing.

[2000] 4. The generative model generates a response

[2001] The preprocessed data is fed into a generative model (e.g., GPT-3) to obtain information about "how to set up Wi-Fi" and generate appropriate instructions.

[2002] 5. The server passes the query to the emotion engine

[2003] The generated response is passed to an emotion engine to recognize the user's emotion.

[2004] 6. Emotion engine recognizes user emotions

[2005] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[2006] 7. The server adjusts the response

[2007] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[2008] 8. The server sends the response to the device

[2009] The adjusted response is converted back to JSON format and sent to the terminal.

[2010] 9. The device displays the response

[2011] The received response is displayed in the user interface, and the user performs an action based on it.

[2012] Prompt Sentence Examples

[2013] "Please tell me how to set up Wi-Fi. And please provide appropriate emotion-aware responses to my queries."

[2014] This system allows users to receive prompt and sensitive answers, eliminating any concerns or questions they may have about the system.

[2015] The flow of the identification process in the second embodiment will be described with reference to FIG.

[2016] Program processing steps

[2017] Step 1: Collect and store data

[2018] server

[2019] The server collects instruction manuals and FAQ data from the internet and internal databases. Specifically, it uses web scraping technology (e.g., BeautifulSoup or Selenium) to extract the necessary data from HTML pages such as instruction manuals. The collected data is saved in CSV or JSON format. It is then stored in a database so that it can be used for later processing.

[2020] Input: instruction manual, FAQ data

[2021] Output: Clean data stored in a database

[2022] Step 2: Cleaning the data

[2023] server

[2024] The server cleans the collected data, using regular expressions and database queries (e.g., SQL) to remove duplicate data and unnecessary information. The cleaned data is then stored in a dedicated database. This prepares the data for use in training the generative model.

[2025] Input: Collected data in the database

[2026] Output: Data in a clean database

[2027] Step 3: Train the model

[2028] server

[2029] The server trains a generative model (e.g., GPT-3) on clean data using a machine learning framework such as TensorFlow or PyTorch. The trained model then becomes the basis for responding to queries.

[2030] Input: Data in a clean database

[2031] Output: A trained generative model

[2032] Step 4: Receiving a user query

[2033] User

[2034] Users enter queries through a dedicated web interface, for example, "How do I set up Wi-Fi?"

[2035] Terminal

[2036] The terminal uses JavaScript or Ajax to convert the query content into JSON format and send it to the server's API endpoint.

[2037] Input: User's inquiry

[2038] Output: Query sent to the server in JSON format

[2039] Step 5: Parsing and preprocessing the query

[2040] server

[2041] The server parses the received JSON query and converts it into an appropriate format. Specifically, it uses Python to parse the JSON data and performs the necessary preprocessing before passing it to the generative model.

[2042] Input: JSON format query

[2043] Output: Preprocessed query data

[2044] Step 6: Generate a response

[2045] server

[2046] The server then passes the preprocessed query to the generative model, which then generates the appropriate response. For example, the generative model retrieves information about "how to set up Wi-Fi" from a database and generates the appropriate instructions.

[2047] Input: Preprocessed query data

[2048] Output: The generated response

[2049] Step 7: Emotion Recognition

[2050] server

[2051] The generated response is passed to an emotion engine (e.g., Google Cloud Natural Language API) to analyze the user's emotions. Emotions such as anger, sadness, and joy are detected from the text data. The emotion engine returns the analysis results to the server.

[2052] Input: Generated response

[2053] Output: User sentiment analysis results

[2054] Step 8: Adjusting the response

[2055] server

[2056] The server adjusts the tone and content of the response based on the results of the emotion engine. For example, if the user is expressing anger, the tone of the response should be changed to be more polite. It also includes a means to escalate the response to a human operator if necessary.

[2057] Input: User sentiment analysis results

[2058] Output: Adjusted response

[2059] Step 9: Sending a Response

[2060] server

[2061] The server then converts the adjusted response back into JSON format and sends it to the device using a secure protocol such as HTTPS.

[2062] Input: Adjusted response

[2063] Output: JSON formatted response sent to the terminal

[2064] Step 10: View the response

[2065] Terminal

[2066] The terminal displays the received response on the user interface, allowing the user to confirm the displayed response and proceed with the operation.

[2067] Input: The response sent by the server

[2068] Output: The response displayed in the user interface

[2069] Step 11: Logging queries and responses

[2070] server

[2071] The server records user queries and generated responses in a log file, which provides data for later performance evaluation and model improvement.

[2072] Input: The query and generated response

[2073] Output: Data recorded in the log file

[2074] Step 12: Evaluate the model's performance

[2075] server

[2076] The server periodically analyzes the log data and evaluates the performance of the generative model and emotion engine, including response accuracy, response time, and emotion recognition accuracy. If necessary, it collects new data and retrains the model.

[2077] Input: Log data

[2078] Output: Performance evaluation results and retrained model

[2079] Step 13: Gather user feedback

[2080] User

[2081] Users input their satisfaction and feedback regarding the provided responses through a dedicated UI and send it to the server via their device.

[2082] server

[2083] The server analyzes the collected feedback and uses it to improve the system.

[2084] Input: User feedback

[2085] Output: Improved system

[2086] Through these steps, users can receive quick and emotionally sensitive answers.

[2087] (Application example 2)

[2088] 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."

[2089] Conventional content distribution services have the problem of reducing user satisfaction because they recommend content without considering the user's emotional state. Furthermore, they often return a uniform response to user inquiries, failing to consider the user's emotions. This results in issues such as a decrease in service usage and an increased likelihood that users will switch to other services.

[2090] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[2091] In this invention, the server includes means for collecting instruction manuals and frequently asked questions data for the information terminal and storing them in a database, means for analyzing the collected data using a generative model and training the model, means for receiving inquiries entered by users and passing them to the trained model, means for returning responses generated from the model to the users, means for recording the inquiries and the generated responses and periodically evaluating the performance of the model, emotion recognition means for analyzing emotions from the user's inquiry text and adjusting the response content, and means for recommending appropriate content according to the user's emotions in a content distribution service, thereby enabling responses and content recommendations that take user emotions into consideration.

[2092] "Handling information terminal" refers to a device used by a user to perform operations or make inquiries.

[2093] An "instruction manual" is a document or material that explains how to use an information terminal and its functions.

[2094] "FAQ data" is data that compiles information about questions and doubts that users frequently have.

[2095] A "generative model" is an algorithm or machine learning model that generates responses to queries based on collected data.

[2096] "Training means" refers to the method or process for training a generative model using collected data.

[2097] A "user query" is a question or request made by a user to a system for information or support.

[2098] A "response" is an answer or instruction that a generative model generates in response to a user query.

[2099] A "database" is a system for systematically storing and managing collected instruction manuals and frequently asked questions.

[2100] "Emotion recognition means" refers to the technology or function for analyzing and extracting emotions from the user's inquiry text.

[2101] A "content distribution service" is a service that provides users with content such as videos, music, and articles.

[2102] "Feedback means" refers to a method or function for collecting user evaluations and opinions on responses and using them to improve the system.

[2103] This invention is a system that provides appropriate and emotionally sensitive responses to user inquiries by combining an emotion engine with an operation method support system that uses a generative model. This system performs the following processes from the perspectives of the server, terminal, and user.

[2104] 1. Data collection and learning phase

[2105] server

[2106] First, the server collects instruction manuals and frequently asked questions data and stores them in a database. The collected data is pre-cleaned to remove duplicate data and unnecessary information. After data cleaning is complete, the data is stored in the database. Next, the server analyzes the pre-processed data using a generative model to train the model. A specific machine learning algorithm is used for this training. Once the model is trained, it serves as the basis for responding quickly and accurately to future inquiries.

[2107] 2. Query Receipt and Response Generation Phase

[2108] User

[2109] If a user has a question about how to operate the system or a specific function, they can input a query into the system through a dedicated user interface, such as "What movies do you recommend?"

[2110] Terminal

[2111] User inquiries received via the user interface are sent by the terminal to the server. Specifically, the inquiry content is converted into JSON format or similar and sent to the server's API.

[2112] server

[2113] The server analyzes the received query, performs necessary preprocessing, and then passes it to the generative model, which generates an appropriate answer based on the query and returns the response to the server.

[2114] 3. Emotion recognition and response adaptation using an emotion engine

[2115] server

[2116] Before sending the generated response, the server passes the query content to an emotion engine to recognize the user's emotion. The emotion engine uses natural language processing technology to analyze emotions from the user's text data. As a result, it determines whether the user is feeling anger, sadness, joy, anticipation, or other emotions. Based on the emotion recognized by the emotion engine, the server adapts the tone and content of the response. For example, if the user is expressing anger, the tone of the response will be adjusted to be more polite. It also includes a function to escalate to a human operator if necessary.

[2117] 4. Sending the Response

[2118] server

[2119] Once the tailored response is complete, the server sends it to the terminal for transmission back to the user, using a secure protocol.

[2120] Terminal

[2121] The terminal displays the response received from the server on the user interface, allowing the user to perform operations based on the response.

[2122] 5. Performance evaluation and improvement

[2123] server

[2124] The server logs user queries and generated responses. This accumulates data for later performance evaluation and model improvement. It periodically analyzes the log data to evaluate the performance of the generative model and emotion engine. This evaluation includes response accuracy, response time, and emotion recognition accuracy. If the model's accuracy falls below a predetermined standard, new data is collected and the model is retrained.

[2125] User

[2126] The user inputs feedback on the provided responses and sends it to the server via the terminal, which helps improve the system.

[2127] Specific examples

[2128] For example, if a user inquires, "What movies do you recommend?", the flow is as follows:

[2129] 1. User enters a query

[2130] The user enters a question.

[2131] 2. The device sends a query to the server

[2132] The user terminal sends this query to the server's API.

[2133] 3. The server receives the query and passes it to the generative model.

[2134] The server analyzes the received query and inputs it into the generative model.

[2135] 4. The generative model generates a response

[2136] The generative model retrieves information about "movie recommendation" from the database and generates appropriate steps.

[2137] 5. The server passes the query to the emotion engine

[2138] The server passes the query content to the emotion engine to recognize the user's emotion.

[2139] 6. Emotion engine recognizes user emotions

[2140] The emotion engine analyzes the user's emotion from the text and returns the emotional state to the server.

[2141] 7. The server adjusts the response

[2142] Based on the results of the emotion engine, the tone and content of the response can be adjusted, and if necessary, escalation to a human operator can be performed.

[2143] 8. The server sends the response to the device

[2144] The server sends the tailored response to the terminal for transmission back to the user.

[2145] 9. The device displays the response

[2146] The terminal displays the response on the user interface, and the user performs an operation based on the response.

[2147] This system allows users to receive prompt and emotionally sensitive answers, improving their satisfaction with the service.

[2148] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[2149] Step 1:

[2150] User enters a query

[2151] A user types a question into a smartphone application, for example, "What movies do you recommend?" This query is entered through a user interface, and the entered text is stored on the device as raw text.

[2152] Input: User query text (e.g., "What movies do you recommend?")

[2153] Output: Raw query text

[2154] Step 2:

[2155] The device sends a query to the server

[2156] The terminal converts user inquiries received via the user interface into JSON format and sends it to the server's API, allowing the server to receive the inquiry content.

[2157] Input: Raw query text

[2158] Output: Data converted to JSON format (e.g., {"query": "What movies do you recommend?"})

[2159] Step 3:

[2160] The server receives the query and passes it to the generative model.

[2161] The server parses the received JSON data and extracts the query content, which is then passed to the generative model, which then generates an appropriate response based on this data.

[2162] Input: Query data in JSON format

[2163] Output: Generated response (e.g., "I recommend 'Hero Movie' as an uplifting action movie.")

[2164] Step 4:

[2165] The generative model generates the response

[2166] The generative model generates an appropriate answer based on the query. This process involves retrieving relevant information from a database and performing data processing and calculations to generate the appropriate response.

[2167] Input: The query passed to the generative model

[2168] Output: Correct response

[2169] Step 5:

[2170] The server passes the query to the emotion engine

[2171] The server passes the generated response and the original query text to the emotion engine to recognize the user's emotion. The emotion engine uses natural language processing techniques to analyze the user's emotional state from the text.

[2172] Input: Generated response and original query text

[2173] Output: Emotion recognition result (e.g. positive, negative, neutral)

[2174] Step 6:

[2175] Emotion engine recognizes user emotions

[2176] The emotion engine analyzes the user's emotion from the input text and returns the state to the server. In this step, data processing and calculations are performed to extract the user's emotion using an analysis algorithm and return the result.

[2177] Input: Query text and generated response

[2178] Output: Emotion recognition result (e.g., positive)

[2179] Step 7:

[2180] The server adjusts the response

[2181] Based on the results of the emotion engine, the server adjusts the generated response, ensuring that the response is of the appropriate tone and content depending on the user's emotional state, and escalating to a human operator if necessary.

[2182] Input: Generated response sentences and emotion recognition results

[2183] Output: Tailored response (e.g., "I recommend 'Hero Movie' as an uplifting action movie. I'm glad you liked it!")

[2184] Step 8:

[2185] The server sends a response to the device

[2186] The server sends the tailored response to the terminal for transmission back to the user, using a secure protocol.

[2187] Input: Adjusted response sentence

[2188] Output: Response data sent to the terminal

[2189] Step 9:

[2190] The terminal displays the response

[2191] The terminal displays the response received from the server on the user interface, allowing the user to perform operations and make selections based on the response.

[2192] Input: Response data sent from the server

[2193] Output: The response text displayed in the user interface

[2194] 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.

[2195] 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.

[2196] 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.

[2197] 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.

[2198] 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.

[2199] 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.

[2200] 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).

[2201] 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.

[2202] 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."

[2203] 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.

[2204] 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).

[2205] 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.

[2206] 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.

[2207] 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.

[2208] 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.

[2209] 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 proces...

Claims

1. A means for collecting instruction manuals and frequently asked questions data for the information terminals handled and storing them in a database; means for analyzing the collected data using a generative model to train the model; a means for receiving a user-entered query and passing it to the trained model; means for returning responses generated from the model to the user; A system that includes a means of recording queries and generated responses and periodically evaluating the performance of the model.

2. The system of claim 1 , further comprising means for collecting additional data and retraining the model if the model's performance falls below a predetermined standard.

3. 2. The system of claim 1, further comprising feedback means for evaluating the appropriateness of responses generated based on said user queries.

Images