System

Abstract

A system is provided.SOLUTION: A system comprising: means for automatically collecting answer data published by each local government; means for converting the collected data into a text format and performing data cleaning; means for generating an embedding from the cleaned text data using a generated AI model; means for registering the generated embedding in a database and updating an index; means for providing an interface for answer data search to a user; means for receiving a search query of the user, converting the query into an embedding, and searching the database; and means for displaying a search result to the user.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

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

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

[0004] Many local governments across the country spend a great deal of time and effort preparing answers to questions in assembly meetings. Local governments hold assembly meetings multiple times a year, and each time, they must conduct numerous case studies and refer to previous examples in order to respond appropriately to questions from assembly members and citizens. However, this information is scattered across multiple locations, making it difficult to search and reference in a centralized location. Therefore, there is a need for a system that can reduce the workload of local government employees and enable efficient case study and answer preparation. [Means for solving the problem]

[0005] In order to solve the above problems, the present invention provides the following means. First, a means is provided for automatically collecting response data made public by each local government. A means is implemented for converting this collected data into text format and performing data cleaning. Next, a means is provided for generating embeddings from the cleaned text data using a generative AI model. A means is also included for registering the generated embeddings in a database and updating the index. Furthermore, a means is provided for providing users with an interface for searching response data. A means is provided for receiving a user's search query, converting the query into an embedding, and searching the database. Finally, a means is provided for displaying search results to the user. In this way, a system is realized that centrally manages response data from dispersed local governments and allows for efficient search and reference.

[0006] "Local public entities" refer to local governments established in each region of the country, and specifically include institutions such as cities, towns, villages, and prefectures that provide administrative services to local residents.

[0007] "Response data" refers to records of answers and explanations officially provided in response to questions from assembly members and citizens in local government assemblies and committees.

[0008] "Collection means" refers to programs or mechanisms for automatically obtaining data that is publicly available on the Internet, and specifically includes the function of collecting data using web crawling technology.

[0009] "Text format" refers to a format in which digital data is expressed as text, and includes, for example, text information extracted from PDF or HTML.

[0010] "Data cleaning" refers to the process of removing unnecessary information and errors from collected text data and preparing it in an analyzable format.

[0011] A "generative AI model" refers to a machine learning model that uses artificial intelligence techniques to process text data and make predictions or transformations tailored to a specific purpose (e.g., generating text embeddings).

[0012] "Embedding" refers to a technique that converts text data into fixed-length numerical vectors, allowing for numerical comparison of the similarity between different texts.

[0013] "Database" refers to a digital storage system that efficiently stores and manages collected and generated data and embeddings, and allows for quick search and reference later.

[0014] An "index" refers to structured information that indicates the location and content of data in a database in order to enable efficient searches of the data.

[0015] An "interface" is something that provides a contact point for users to operate a system, and specifically includes the screens and operation methods of web applications and desktop applications.

[0016] A "query" refers to a question or keyword that a user enters into a system to search for specific information.

[0017] "Search engine" refers to a program or function that receives a query input from a user and provides optimal search results for that query.

[0018] The term "search result display means" refers to a program or function for visually displaying the search results generated by the search means to the user. [Brief explanation of the drawings]

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

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

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

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

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

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

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

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

[0027] [First embodiment]

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

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

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

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

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

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

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

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

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

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

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

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

[0040] This invention provides a system that collects parliamentary response data made public by local governments, generates embeddings using a generative AI model, and allows users to easily search and reference the data.

[0041] Data collection implementation

[0042] The server accesses the official websites of each local government and automatically crawls the publicly available parliamentary response data. During this process, the web crawler analyzes the HTML structure and extracts only the necessary information from the specified elements. For example, it automatically downloads the PDF link for the response data on each local government website.

[0043] A specific example is a process that collects PDF files related to responses from the websites of multiple local governments and saves them in a specific folder.

[0044] Data Preprocessing Embodiments

[0045] The server converts the collected PDF files into text data. This involves extracting text from the PDF using a dedicated library. Data cleaning is then performed. This cleaning process removes unnecessary line breaks and special characters, resulting in pure text data.

[0046] A specific example is deleting unnecessary header and footer information from text data extracted from a PDF file and formatting it to leave only text.

[0047] An embodiment of embedding generation

[0048] The server uses a pre-trained generative AI model to generate embeddings from the cleaned text data, which convert the text data into numerical vectors for easier future search and comparison.

[0049] Specifically, the text of parliamentary responses is input into a generative AI model, which outputs a fixed-length vector that is later stored in a database and used as a search index.

[0050] Database Update Implementation Example

[0051] The server registers the generated embedding in its database and updates its search index, allowing it to respond quickly to future user search queries.

[0052] A concrete example is the process of adding newly generated embeddings to a database and rebuilding the search index along with the existing data.

[0053] User Interface Embodiments

[0054] The terminal provides the user with an interface for searching the answer data, which has a user-friendly design and includes a search box and a search result display area.

[0055] A specific example is an interface that provides a search box in a web browser, allowing users to simply enter keywords and display related response data.

[0056] Query Processing Embodiments

[0057] A user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data.

[0058] As a concrete example, when a user enters "environmental policy," the process converts the query into an embedding, searches a database, and lists relevant response data.

[0059] Display of search results

[0060] The server returns the search results to the terminal and displays them visually to the user, allowing the user to easily refer to and utilize the response data of other local governments.

[0061] As a specific example, related defense data is displayed in a list format on the Web, and an interface is provided that allows users to click on details to view the full text.

[0062] By combining the above means, the present invention provides a system that efficiently collects response data from local governments and allows users to easily search and refer to it.

[0063] The processing flow will be explained below.

[0064] Step 1:

[0065] The server accesses the list of website URLs collected and crawls response data from the official websites of each local government.

[0066] Specifically, the server analyzes the HTML structure, extracts links to the response data based on specific HTML tags and classes, and follows those links to download PDF files, etc.

[0067] Step 2:

[0068] The server converts the collected PDF files into text format.

[0069] Specifically, it uses a PDF to text conversion library to extract the text from each page and combine it into a single text file.

[0070] Step 3:

[0071] The server performs data cleaning on the text data.

[0072] Specifically, it removes extra line breaks, special characters, and unnecessary header and footer information to ensure data consistency.

[0073] Step 4:

[0074] The server uses a generative AI model to generate embeddings from the cleaned text data.

[0075] Specifically, text data is input into a pre-trained language model and converted into a fixed-length vector.

[0076] Step 5:

[0077] The server registers the generated embedding in the database and updates the index.

[0078] Specifically, it inserts new entries into the database and rebuilds existing indexes to enable efficient searches.

[0079] Step 6:

[0080] The terminal provides a search interface to the user.

[0081] Specifically, it installs a search box and results display area in a web browser or dedicated application, providing an interface that is easy for users to use.

[0082] Step 7:

[0083] The server receives a search query entered by the user, converts the query into an embedding, and searches the database.

[0084] Specifically, the search query is input into the generative AI model, and the generated embedding is compared with embeddings in the database to extract entries with high similarity.

[0085] Step 8:

[0086] The server returns the search results to the terminal and displays them visually to the user.

[0087] Specifically, the search results are sent to the terminal in JSON format, and the terminal formats the data it receives and displays it to the user.

[0088] Example 1

[0089] 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."

[0090] The data on parliamentary responses made public by local governments is scattered across various websites, making it difficult to centrally collect and easily search and reference. Furthermore, the volume of information is enormous, so manually collecting and organizing the data takes time and effort. Furthermore, variations in linguistic expression and clerical errors make data cleaning difficult, reducing the efficiency of information searches. There is a need to resolve these issues and enable users to efficiently access response data.

[0091] 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.

[0092] In this invention, the server includes means for automatically collecting response data published by each local government, means for converting the collected data into text format and performing data cleaning, means for generating embeddings from the cleaned text data using a generative AI model, and means for registering the generated embeddings in a database and updating an index, thereby enabling efficient collection, processing, and search of response data from local governments.

[0093] "Response data" refers to official statements and responses recorded in local government assemblies and committees.

[0094] "Automatic collection means" refers to a system that uses web crawlers or scraping technology to automatically obtain specified information from specific websites.

[0095] "Text conversion method" refers to a technology that extracts text information from collected PDFs, image files, etc., and converts it into a digital text format.

[0096] "Data cleaning means" is a process of formatting collected text data by removing typos, unnecessary special characters, line breaks, etc.

[0097] A "generative AI model" is an artificial intelligence model that is trained in advance with large amounts of data and exhibits high performance for specific tasks.

[0098] "Embedding generation method" is a technology that converts text data into numerical vectors and expresses their characteristics.

[0099] The "database registration method" is the process of saving the generated embedding to a database and updating the search index.

[0100] A "search query" refers to a search term or phrase that a user enters to find desired information.

[0101] A "user interface" is an interface that includes the screens and functions that a user operates when using a system.

[0102] An "index" refers to a structure or mechanism for efficiently searching data in a database.

[0103] The present invention is a system that automatically collects assembly response data published by local governments, generates embeddings using a generative AI model, and allows users to easily search and reference them. Specific embodiments for implementing the present invention are described below.

[0104] The server accesses the official websites of each local government and automatically collects publicly available parliamentary response data. This data collection is performed using web crawlers and scraping techniques. Specifically, it uses HTML analysis libraries such as BeautifulSoup to analyze the HTML structure of each website and obtain links to response data and PDF files.

[0105] Next, the server converts the collected PDF files into text data. To do this, it uses a library called PyMuPDF to extract text from the PDF. Then, it uses a data cleaning method to remove unnecessary line breaks and special characters from the extracted text data and format it. This results in clean text data.

[0106] Embeddings are generated from the cleaned text data using a generative AI model. This generative AI model uses a pre-trained language model such as GPT-3 (registered trademark). Using the text data as input, the model generates a numerical vector called an embedding. This allows the features of the text data to be represented as a numerical vector.

[0107] The generated embeddings are registered in a database by the server, and as part of this registration process, the server also updates a search index, allowing the data in the database to be efficiently searched.

[0108] The user interface is provided using a terminal. The terminal provides an interface with a search box and a result display area so that the user can search for answer data. For example, the interface is implemented using the React.js framework in a web browser.

[0109] A user inputs a search query through this interface. The search query is converted into an embedding by the server. The converted embedding is used to search the database and extract the most relevant response data. For example, if a user inputs the query "environmental policy," the query is converted into an embedding and the relevant response data in the database is listed.

[0110] Finally, the server returns the search results to the terminal and visually displays them to the user, allowing the user to easily browse related answer data and view detailed information.

[0111] A specific example of a prompt sentence could be, "Search for questions about education policy in local councils." By inputting this prompt sentence into a generative AI model, generating an embedding, and searching the database, relevant answer data can be easily obtained.

[0112] The above is a specific embodiment for carrying out the present invention, which realizes a system that can efficiently collect, search, and use assembly response data from local public organizations.

[0113] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0114] Step 1:

[0115] The server accesses the official websites of each local government and automatically collects data on parliamentary responses.

[0116] Specifically, the server uses a web crawler and an HTML parsing library such as BeautifulSoup to analyze the HTML structure of the web page.

[0117] The input is the URL of the official website of each local government, and the output is a link to the response data or a PDF file.

[0118] During this process, the server searches for the specified keywords and file formats and retrieves links containing the response data. For example, the server accesses "https: / / example-localgovsite.jp / kaigi" and retrieves PDF links to the parliamentary response data.

[0119] Step 2:

[0120] The server converts the collected PDF files into text data.

[0121] For this, we use a PDF parsing library called PyMuPDF.

[0122] The input is the PDF files collected in step 1, and the output is text data.

[0123] Specifically, the server opens the PDF file and extracts its contents in text format. For example, it uses PyMuPDF to extract the text from "example.pdf" and saves it as text data.

[0124] Step 3:

[0125] The server performs data cleaning on the text data.

[0126] The input is the text data extracted in step 2, and the output is the cleaned text data.

[0127] Specifically, the server deletes unnecessary line breaks and special characters from the text data and arranges the data. For example, the server deletes line breaks and special characters to obtain text data that has been arranged in a single line.

[0128] Step 4:

[0129] The server uses a generative AI model to generate embeddings from the cleaned text data.

[0130] The input is the text data cleaned in step 3, and the output is the embedding (a numeric vector).

[0131] Specifically, the server inputs the cleaned text data into a generative AI model to obtain embeddings. For example, the server inputs the text data into a GPT-3 model and obtains the generated numeric vectors.

[0132] Step 5:

[0133] The server registers the generated embedding in the database and updates the index.

[0134] The input is the embedding generated in step 4, and the output is the updated database and index.

[0135] Specifically, the server adds the embedding to the database and rebuilds the search index, e.g., registers the embedding in the PostgreSQL database and updates the index.

[0136] Step 6:

[0137] The terminal provides the user with an interface for searching for answer data.

[0138] There is no input, and the output is an interface with a search box and a results display area.

[0139] Specifically, the device displays an interface using React.js on a web browser, allowing users to enter search queries. For example, it provides an interface that displays a search box and a search button.

[0140] Step 7:

[0141] A user enters a search query into the interface, and the server converts the query into an embedding and searches the database.

[0142] The input is a user's search query and the output is a list of relevant answer data.

[0143] Specifically, the server passes the input search query to the generative AI model, generates an embedding, and then searches the database. For example, input the query "environmental policy" and list related response data.

[0144] Step 8:

[0145] The server returns the search results to the terminal and displays them visually to the user.

[0146] The input is the list of search results obtained in step 7, and the output is a visual display that the user can refer to.

[0147] Specifically, the server sends the search results in JSON format to the terminal, and the terminal displays the results in a list format, for example, by displaying a list of related answer data on a web interface so that the user can check the details.

[0148] The above are the specific program processing steps of this system.

[0149] (Application example 1)

[0150] 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."

[0151] In today's brick-and-mortar stores, store employees are required to quickly and accurately provide product information and FAQs in response to customer inquiries. However, with existing systems, this process is often manual and inefficient, which can lead to a decline in the quality of customer service. Furthermore, collecting necessary information from various data sources, integrating it, and making it searchable takes time and effort, increasing the burden on employees. It is necessary to solve these problems and improve customer service and employee work efficiency.

[0152] 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.

[0153] In this invention, the server includes: means for automatically collecting response data published by each local government; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for allowing store employees to quickly search for product information and FAQs in response to customer inquiries; and means for displaying search results to the user. This enables employees to respond to customer inquiries quickly and accurately based on integrated data collected from various information sources. Furthermore, this significantly improves employee work efficiency and contributes to increased customer satisfaction.

[0154] A "local public entity" is a local government that handles administrative affairs as a local government separate from the nation.

[0155] "Response data" refers to data that records the content of questions and answers that were asked in parliaments or conferences.

[0156] "Automatic collection means" refers to a mechanism by which a system or program collects data automatically, without human intervention.

[0157] "Means of converting to text format" refers to the process of converting data such as images and PDFs into a format that can be handled as text information.

[0158] "Data cleaning" is the process of removing unnecessary information and errors from collected data, making it easier to analyze.

[0159] A "generative AI model" is an artificial intelligence system that has been pre-trained using machine learning or deep learning to perform a specific task.

[0160] An "embedding" is a representation of text data converted into a numerical vector that is used when processing it with a machine learning model.

[0161] A "database" is an information management system that systematically stores specific information and allows it to be searched and used when needed.

[0162] "Means of updating the index" refers to the process of updating the reference system to make searches more efficient for newly added data.

[0163] An "interface" is the hardware and software structure through which a user interacts with a computer system.

[0164] A "search query" is a keyword or phrase that a user enters into a search system to retrieve specific information.

[0165] "Customer inquiries" are questions or clarifications made by store patrons regarding products or services.

[0166] "Product information" is a detailed description of the product's specifications, price, usage, etc.

[0167] A "FAQ" is a collection of frequently asked questions and their answers.

[0168] The "means for displaying search results" is a method for visually conveying to the user the data retrieved based on the query entered by the user.

[0169] MODE FOR CARRYING OUT THE INVENTION

[0170] This invention provides a system that collects response data from local governments based on a series of procedures and enables quick and efficient search using a generative AI model. Specific embodiments of this system are described below.

[0171] Data collection implementation

[0172] The server accesses the official websites of each local government and automatically collects publicly available response data. During this process, a web crawler running on the server analyzes the HTML structure and extracts only the necessary information from specified elements. For example, it automatically downloads PDF links to response data on each local government's website. The software used is requests and BeautifulSoup.

[0173] Data Preprocessing Embodiments

[0174] The server converts the collected PDF files into text data. To do this, it uses a dedicated library called pdfminer to extract text from the PDF. Data cleaning is then performed to remove unnecessary line breaks and special characters and format the text data.

[0175] An embodiment of embedding generation

[0176] The server uses a pre-trained generative AI model (e.g., SentenceTransformer) to generate embeddings from the cleaned text data, which convert the text data into numerical vectors for easier later search and comparison.

[0177] Database Update Implementation Example

[0178] The server registers the generated embedding in a database and updates the index, allowing for quick responses when users later enter search queries. The database uses a data management system such as SQLite.

[0179] User Interface Embodiments

[0180] The terminal provides the user with an interface for searching for answer data. This interface has a user-friendly design, including a search box and a search result display area. As a specific example, a search box is installed in a web browser, and the user can simply enter keywords to display related answer data.

[0181] Query Processing Embodiments

[0182] When a user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data. For example, if a user enters "environmental policy," the server converts the query into an embedding and searches the database to list relevant answer data.

[0183] Display of search results

[0184] The server returns the search results to the terminal and visually displays them to the user, allowing the user to easily refer to and utilize related answer data. For example, related answer data is displayed in a list format on the web, and an interface is provided where the user can click on details to view the full text.

[0185] Specific examples

[0186] For example, if a drugstore staff member is asked by a customer about "side effects of allergy medication," they can use this system to quickly search and refer to the information. The staff member simply types "side effects of allergy medication" into the tablet or smartphone interface, and the system instantly displays relevant information.

[0187] Prompt Sentence Examples

[0188] "Please tell me about the side effects of allergy medication."

[0189] In this way, a system is created in which complex information processing is seamlessly performed on the back end, allowing store staff to quickly and accurately answer customer questions.

[0190] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0191] Step 1:

[0192] The server accesses the official websites of each local government and automatically collects publicly available response data. Specifically, a web crawler running on the server analyzes the HTML structure and extracts PDF links to the response data. It receives a list of website URLs as input and downloads and saves the PDF data as output.

[0193] Step 2:

[0194] The server converts the collected PDF files into text data. Specifically, it uses the pdfminer library to extract text from the PDF, and then cleans it by removing unnecessary line breaks and special characters. It takes PDF files as input and generates clean text data as output.

[0195] Step 3:

[0196] The server uses a generative AI model (e.g., SentenceTransformer) to generate embeddings from the cleaned text data. This process involves calculations to convert the text data into a numerical vector. It receives the clean text data as input and generates embeddings (numerical vectors) as output.

[0197] Step 4:

[0198] The server registers the generated embedding in a database and updates the index. Specifically, it uses a data management system such as SQLite to insert the newly generated embedding and the original text data into the database and update the search index. It receives the generated embedding and the original text data as input and obtains an updated database as output.

[0199] Step 5:

[0200] The terminal provides a user with an interface for searching answer data. The interface displays a search box where the user can enter a search query and a results display area. The terminal receives a search query from the user as input and prepares an output to process the query.

[0201] Step 6:

[0202] The server converts the user's search query into an embedding and searches the database. Specifically, the query is input into a generative AI model to generate an embedding. It then compares it with existing embeddings in the database to extract the most relevant answer data. It receives the user's search query as input and obtains search results as output.

[0203] Step 7:

[0204] The server returns the search results to the terminal and displays them visually to the user. For example, the server may display the relevant answer data in a list format, and provide an interface that allows the user to click on details to view the full text. The server receives the search results as input and generates the data to be displayed to the user as output.

[0205] 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.

[0206] The present invention provides a system that automatically collects parliamentary response data made public by local governments and can efficiently search it using embeddings generated using a generative AI model, as well as a system that combines an emotion engine that recognizes the user's emotions.

[0207] Data collection implementation

[0208] The server accesses the official websites of each local government and crawls the publicly available data on parliamentary responses. The server analyzes the HTML structure and extracts links to the response data based on specific HTML tags and classes. The server then follows the links and automatically downloads PDF files and other data.

[0209] A specific example is a process for collecting PDF files related to parliamentary responses from the websites of multiple local governments in bulk.

[0210] Data Preprocessing Embodiments

[0211] The server converts the collected PDF files into text format. The server uses a PDF-to-text conversion library to extract the text from each page and compile it into a single text file. The server then performs data cleaning to remove unnecessary line breaks, special characters, and unnecessary header and footer information.

[0212] As a concrete example, it removes unnecessary information from text data extracted from a PDF file and generates text data that is formatted to be easy to read.

[0213] An embodiment of embedding generation

[0214] The server uses a pre-trained generative AI model to generate embeddings from the cleaned text data, which convert the text data into numerical vectors that can be used for efficient search and comparison.

[0215] Specifically, the text of parliamentary responses is input into a generative AI model, which outputs a fixed-length vector that is later stored in a database and used as a search index.

[0216] Database Update Implementation Example

[0217] The server registers the generated embedding in its database and updates its search index, allowing it to respond quickly when a user enters a search query.

[0218] The specific operation involves adding the newly generated embeddings to the database and rebuilding the index along with the existing data.

[0219] User Interface Embodiments

[0220] The terminal provides the user with an interface for searching answer data. The user interface includes a search box and a search result display area, and is designed to be easy for users to operate.

[0221] A specific example is an interface that provides a search box in a web browser, and when a user enters a keyword, related answer data is displayed.

[0222] Query Processing Embodiments

[0223] A user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data.

[0224] As a concrete example, when a user inputs "environmental policy," the query is converted into an embedding, a database search is performed, and relevant response data is extracted.

[0225] Embodiment of search result display

[0226] The server returns the search results to the terminal and displays them visually to the user, who can then easily refer to and manipulate the response data of other local governments.

[0227] As a concrete example, related response data is displayed in a list format on the web, and users can click on each item to view details.

[0228] Embodiment of Emotion Engine

[0229] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when entering a search query. If the user is in a positive emotion, this emotion engine prioritizes the display of highly relevant search results, and if the user is in a negative emotion, it displays appropriate results and messages to reduce stress.

[0230] A specific example is an interface in which an emotion engine analyzes a user's facial expressions and vocal tone, and if they are judged to be negative, provides relaxing information.

[0231] By combining the above methods, the present invention provides a system that efficiently collects response data from local governments and allows users to easily search and refer to it. Furthermore, by combining it with an emotion engine, it is possible to provide optimal search results according to the user's emotional state, improving the user experience.

[0232] The processing flow will be explained below.

[0233] Step 1:

[0234] The server accesses the official websites of each local government and crawls the data on parliamentary responses. Based on the URL list, the server analyzes the HTML structure of each website, extracts links to the response data from elements that match specific HTML tags and classes, and follows the links to download PDF files.

[0235] Step 2:

[0236] The server converts the collected PDF files into text format. The server uses a dedicated library for extracting text from PDFs, extracts the text for each page, and saves it as a single text file.

[0237] Step 3:

[0238] The server performs data cleaning on the text data. Specifically, it removes unnecessary line breaks, special characters, and unnecessary header and footer information to generate formatted text data. This prepares the data in a consistent, parseable format.

[0239] Step 4:

[0240] The server uses a generative AI model to generate embeddings from the cleaned text data. The server inputs the text data into a pre-trained language model and outputs a fixed-length numeric vector (embedding). This embedding is a numeric representation of the text's meaning.

[0241] Step 5:

[0242] The server registers the generated embeddings in the database and updates the index. The server adds new embeddings to the database and rebuilds existing data indexes to improve search efficiency.

[0243] Step 6:

[0244] The terminal provides the user with an interface for searching answer data. Specifically, a user-friendly interface is provided with a search box and a search result display area, allowing the user to easily operate it.

[0245] Step 7:

[0246] The server receives the user's search query, converts the query into an embedding, and searches the database. The user-entered search query is fed into a generative AI model, which compares the converted embedding with existing embeddings in the database.

[0247] Step 8:

[0248] The server sends the search results to the device and displays them visually to the user. The search results are sent to the device in JSON format, and the device formats the received data and displays it to the user in a list format.

[0249] Step 9:

[0250] The server is equipped with an emotion engine that recognizes the user's emotions and optimizes search results according to the user's emotional state. When a user enters a search query, the emotion engine analyzes the user's facial expressions and tone of voice, and prioritizes relevant search results if the user's emotions are positive, and displays appropriate results and messages to reduce stress if the user's emotions are negative.

[0251] Step 10:

[0252] The server reflects the results of the emotion engine and adjusts the ranking of search results to provide an optimal user experience. By displaying search results with the display order adjusted based on the emotion engine, users can obtain information efficiently and without stress.

[0253] Example 2

[0254] 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."

[0255] Although data on responses from local governments is made public by each organization, there is no way to efficiently collect it and organize it in a searchable format. Furthermore, there is no system in existence that uses an interface that takes user emotions into consideration to efficiently search for the necessary response data while reducing stress. This forces users to manually search for the information they need from large amounts of data, which is a laborious and time-consuming process.

[0256] 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.

[0257] In this invention, the server includes: means for automatically collecting response data published by each local government; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for displaying search results to the user; and means including an emotion engine for recognizing the user's emotional state and displaying search results and appropriate messages. This allows efficient collection of response data from local governments, allows users to easily search and reference it, and provides optimal search results according to the user's emotional state.

[0258] "Response data" refers to data that records the answers and explanations given by administrative representatives to questions from assembly members in local government assemblies.

[0259] A "text format" is a form of digital data that is recorded as human-readable character information.

[0260] "Data cleaning" is the process of removing unnecessary information and errors from collected data and formatting it.

[0261] A "generative AI model" is an artificial intelligence model that has been pre-trained with large amounts of data and is capable of generating appropriate outputs for specific inputs.

[0262] An "embedding" is a conversion of text data into a numerical vector, which is used for efficient searching and comparison.

[0263] A "database" is an information system that efficiently stores and manages large amounts of data and allows quick access when needed.

[0264] A "search index" is an index used to efficiently search for information in a database.

[0265] An "interface" is a point of contact or means for exchanging information between a user and a system, and includes, for example, a search box or a display screen.

[0266] A "search query" refers to an inquiry that a user enters into a search system, and the system searches for related information based on the query.

[0267] The "emotion engine" is a system component that has the function of recognizing the user's emotional state and providing appropriate information and messages according to that state.

[0268] This invention is a system that automatically collects parliamentary response data published by local governments and efficiently searches it using embeddings generated by a generative AI model. Furthermore, we aim to improve the search experience by combining it with an emotion engine that recognizes the user's emotions.

[0269] Data collection

[0270] The server accesses the official websites of each local government and crawls the publicly available data on assembly responses. Specifically, it analyzes the HTML structure, extracts links to the response data based on specific HTML tags and classes, and automatically downloads PDF files by following those links. For example, the server accesses the Tokyo Metropolitan Assembly website, extracts links to PDF files containing keywords including "response," and downloads them.

[0271] Data Preprocessing

[0272] The server converts the collected PDF files into text format. Using a library for converting PDF to text format (e.g., PyPDF2), the server extracts text from each page and compiles it into a single text file. It then performs data cleaning, removing unnecessary line breaks, special characters, and unnecessary header and footer information. Specifically, the server uses PyPDF2 to extract text from the downloaded PDF files, removes unnecessary information, and creates a unified text document.

[0273] embedding generation

[0274] The server uses a pre-trained generative AI model (e.g., a pre-trained language model) to generate embeddings from the cleaned text data. These embeddings convert the text data into numerical vectors that can be used for efficient search and comparison. For example, the server inputs the cleaned text data into a language model, which outputs a fixed-length vector. This vector is stored in a database and serves as a search index.

[0275] Database Update

[0276] The server registers the generated embedding in the database and updates the search index. Specifically, the server adds the newly generated embedding to the existing database and rebuilds the search index, allowing for quick responses when a user enters a search query.

[0277] User Interface

[0278] The terminal provides the user with an interface for searching for answer data. The interface is equipped with a search box and a search result display area, and is designed to be easy for users to operate. For example, if a search box is provided in a web browser and the user enters a keyword, related answer data will be displayed.

[0279] Query processing and search result display

[0280] A user enters a search query, and the server converts the query into an embedding and searches the database. The server compares the generated embedding with existing embeddings in the database and extracts the most relevant answer data. For example, if a user enters "environmental policy," the server converts the query into an embedding and searches the database to extract relevant answer data. The server returns the results to the device and displays them visually to the user.

[0281] Emotion engine integration

[0282] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when they enter a search query. This emotion engine prioritizes relevant search results if the emotion is positive, and displays appropriate results and messages to reduce stress if the emotion is negative. For example, the emotion engine analyzes the user's facial expressions and tone of voice, and if it determines that the emotion is negative, it provides an interface with information to help users relax.

[0283] This system efficiently collects response data from local governments and provides an environment where users can easily search and refer to it. Furthermore, by combining it with an emotion engine, it is possible to provide optimal search results according to the user's emotional state, improving the user experience.

[0284] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0285] Step 1: Data collection

[0286] The server accesses the official websites of each local government and crawls the publicly available data on parliamentary responses. The input is the URL of the local government's website, and the output is a link to a PDF file of the parliamentary response data. Specifically, the server analyzes the HTML structure of the webpage, extracts links to the response data based on specific HTML tags and classes, and automatically downloads the PDF file by following the link.

[0287] Step 2: Data Preprocessing

[0288] The server converts the collected PDF files into text format. The input is a PDF file, and the output is text data. Specifically, a library for converting PDF to text (e.g., PyPDF2) is used to extract the text from each page and combine it into a single text file. The data is then cleaned to remove unnecessary line breaks, special characters, and unnecessary header and footer information.

[0289] Step 3: Embedding generation

[0290] The server generates embeddings from the cleaned text data. The input is the cleaned text data, and the output is the embedding (a numerical vector). Specifically, it uses a pre-trained generative AI model (e.g., a pre-trained language model) to convert the text data into a numerical vector, which makes the text data in a format that can be efficiently searched and compared.

[0291] Step 4: Update the database

[0292] The server registers the generated embedding in the database and updates the search index. The input is the generated embedding, and the output is the updated search index. The specific operation is to add the newly generated embedding to the existing database and rebuild the search index. This allows for quick response when a user enters a search query.

[0293] Step 5: User Interface

[0294] The terminal provides the user with an interface for searching for answer data. The input is the user's operation, and the output is the interface that displays the search results. Specifically, the operation is to provide a search box and a search result display area in the web browser, allowing the user to enter keywords. Related answer data is displayed based on the entered keywords.

[0295] Step 6: Query Processing

[0296] The server receives a search query entered by a user, converts the query into an embedding, and searches a database. The input is the user's search query (text), and the output is related answer data. Specifically, the server inputs the user's search query into a generative AI model, searches the database using the resulting embedding, and extracts the most relevant answer data.

[0297] Step 7: Displaying search results

[0298] The server returns the search results to the terminal and displays them visually to the user. The input is the answer data extracted by the search, and the output is the displayed search results. Specifically, the server finds the relevant answer data and sends it to the terminal, and the terminal displays the search results in list format on the web browser. The user can click on each item to view details.

[0299] Step 8: Integrating the Emotion Engine

[0300] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when they enter a search query. The input is the user's facial expression and tone of voice, and the output is appropriate search results and messages. Specifically, the emotion engine analyzes the user's emotional state, and if the emotion is positive, it prioritizes relevant search results, and if the emotion is negative, it provides appropriate messages and information to reduce stress.

[0301] (Application example 2)

[0302] 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."

[0303] Factories require efficient management of maintenance information for multiple pieces of equipment and machinery. However, manual collection and analysis is time-consuming and labor-intensive, often resulting in insufficient data management and analysis. Furthermore, while it is known that workers' emotional states affect work efficiency and safety, there is a lack of systems that address this issue. Therefore, there is a need for a system that can automatically collect and efficiently search maintenance information, as well as a system that can provide feedback based on workers' emotions.

[0304] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: means for automatically collecting response data made public by local governments; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for displaying search results to the user; means for processing maintenance information collected from a machine as text data; and means for providing search results according to the user's emotional state by incorporating an emotion engine for recognizing the user's emotions. This not only enables efficient collection and search of maintenance information, but also enables appropriate feedback to be provided according to the worker's emotional state.

[0305] 1. "Response data" refers to data showing the content of responses given in local government assemblies and other such meetings.

[0306] 2. "Text format" means a data format expressed as human-readable characters or sentences.

[0307] 3. "Data cleaning" is the process of removing redundant information and noise from collected text data.

[0308] 4. A "generative AI model" is an artificial intelligence model that is pre-trained using massive amounts of data.

[0309] 5. "Embedding" is a method of representing text data as a numerical vector.

[0310] 6. A "database" is a system for efficiently storing, managing, and retrieving structured data.

[0311] 7. "Interface" refers to the operation screens and input fields that allow users to interact with the system.

[0312] 8. A "query" is a question or keyword entered by a user to search for specific information.

[0313] 9. "Search query" means input to the system to specify the information to be searched.

[0314] 10. "Machinery" refers to production equipment and devices installed in factories and production lines.

[0315] 11. "Maintenance Information" means data relating to the maintenance and repair of facilities and machinery.

[0316] 12. An "emotion engine" is an algorithm or system for determining a user's emotional state.

[0317] 13. "Feedback" refers to the response or information that a system returns in response to a user's input or actions.

[0318] To implement this invention, a server first accesses the official websites of each local government and automatically collects publicly available response data. Specifically, the server uses web crawling technology to extract and download response data links based on specific HTML tags and classes. For example, there is a process for collecting PDF files from the websites of multiple local government organizations in bulk.

[0319] Next, the server converts the collected PDF files into text format. This is done by extracting the text from each page using a PDF to text conversion library (e.g., PyPDF2) and compiling it into a single text file. After that, data cleaning is performed, removing unnecessary line breaks and special characters, and generating formatted text data. For example, there is a process to remove unnecessary information from the text data extracted from the PDF files.

[0320] The server then uses a generative AI model (e.g., OpenAI® GPT-3 / BERT) to generate embeddings from the cleaned text data. These embeddings convert the text data into numerical vectors that can be used for efficient search and comparison. Specifically, the text data is input into the generative AI model, which outputs a fixed-length vector. This vector is later stored in a database and used as a search index.

[0321] Next, the server registers the generated embedding in a database (e.g., ElasticSearch (registered trademark)) and updates the index. This allows for quick response when a user enters a search query. Specific operations include adding the newly generated embedding to the database and rebuilding the index with the existing data.

[0322] The server also provides a user interface to allow users to search for the answer data. This interface is installed on a terminal and includes a search box and a search result display area. For example, there is an interface in which a search box is provided in a web browser, and when a user enters a keyword, related answer data is displayed.

[0323] Next, when a user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data. For example, if a user enters "machine wear diagnosis," there is a process to convert the query into an embedding, perform a database search, and extract relevant answer data.

[0324] The server also returns the search results to the terminal and visually displays them to the user. The user can easily refer to and manipulate the related answer data. For example, the related answer data is displayed in list format on the web, and the user can click on each item to view details.

[0325] Finally, the server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when entering a search query. This emotion engine prioritizes relevant search results when the user is in a positive emotion, and displays appropriate results and messages to reduce stress when the user is in a negative emotion. For example, the emotion engine analyzes the user's facial expressions and tone of voice, and if the engine determines that the user is in a negative emotion, it provides an interface that provides information to help them relax.

[0326] Prompt Sentence Examples

[0327] Parliamentary response data analysis. Generate an embedding from the following text:

[0328] "Answer regarding environmental policy..."

[0329] This invention not only makes it possible to efficiently collect and search maintenance information, but also makes it possible to provide appropriate feedback to the worker according to his or her emotional state.

[0330] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0331] Step 1:

[0332] The server accesses the official websites of local government organizations and collects response data. Specifically, the server crawls multiple specified websites using web crawling technology (e.g., BeautifulSoup, Scrapy), extracts links to response data based on specific HTML tags and classes, and downloads PDF files. The input is the website URL, and the output is the downloaded PDF file.

[0333] Step 2:

[0334] The server converts the collected PDF files into text format. Specifically, it uses the PyPDF2 library to extract text from each page of the PDF file and combines it into a single text file. The input is a PDF file and the output is a text file.

[0335] Step 3:

[0336] The server cleans the text files. Specifically, it uses libraries such as NLTK to remove unnecessary line breaks, special characters, and unnecessary header and footer information, generating readable, formatted text data. The input is the raw text file, and the output is the cleaned text data.

[0337] Step 4:

[0338] The server generates embeddings from the cleaned text data using a generative AI model (e.g., OpenAI GPT-3 / BERT). Specifically, the text data is input to the AI ​​model as a prompt sentence, and a fixed-length vector is output. The input is the cleaned text data, and the output is the generated embedding.

[0339] Step 5:

[0340] The server registers the generated embeddings in a database (e.g., Elasticsearch) and updates the index. Specifically, it adds the newly generated embeddings to the database and rebuilds the index with the existing data. The input is the generated embeddings, and the output is the updated database index.

[0341] Step 6:

[0342] The terminal provides the user with an interface for searching answer data via a user interface. Specifically, the terminal displays a web browser interface including a search box and a search result display area. The input is the user interface design specification, and the output is the interface available to the user.

[0343] Step 7:

[0344] When a user enters a search query, the server converts the query into an embedding and searches the database. Specifically, the search query is input into the AI ​​model, and the generated embedding is compared with existing embeddings in the database. The input is the user's search query, and the output is highly relevant answer data.

[0345] Step 8:

[0346] The server returns the search results to the terminal and displays them visually to the user. Specifically, it sends the relevant answer data in a list format to the terminal so that the user can view the details. The input is the search results, and the output is the displayed search results.

[0347] Step 9:

[0348] The server uses an emotion engine to recognize the user's emotional state. Specifically, it analyzes the user's facial expressions and voice tone using the device's camera and microphone. The input is real-time video and audio data, and the output is the user's emotional state.

[0349] Step 10:

[0350] The server provides appropriate search results and messages based on the user's emotional state. Specifically, if the user's emotional state is positive, it prioritizes relevant results, and if the user's emotional state is negative, it displays relaxing information and messages. The input is the user's emotional state, and the output is the adjusted search results and messages.

[0351] 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.

[0352] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0353] 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.

[0354] [Second embodiment]

[0355] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0356] 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.

[0357] 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).

[0358] 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.

[0359] 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.

[0360] 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).

[0361] 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.

[0362] 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.

[0363] 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.

[0364] 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.

[0365] 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.

[0366] 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."

[0367] This invention provides a system that collects parliamentary response data made public by local governments, generates embeddings using a generative AI model, and allows users to easily search and reference the data.

[0368] Data collection implementation

[0369] The server accesses the official websites of each local government and automatically crawls the publicly available parliamentary response data. During this process, the web crawler analyzes the HTML structure and extracts only the necessary information from the specified elements. For example, it automatically downloads the PDF link for the response data on each local government website.

[0370] A specific example is a process that collects PDF files related to responses from the websites of multiple local governments and saves them in a specific folder.

[0371] Data Preprocessing Embodiments

[0372] The server converts the collected PDF files into text data. This involves extracting text from the PDF using a dedicated library. Data cleaning is then performed. This cleaning process removes unnecessary line breaks and special characters, resulting in pure text data.

[0373] A specific example is deleting unnecessary header and footer information from text data extracted from a PDF file and formatting it to leave only text.

[0374] An embodiment of embedding generation

[0375] The server uses a pre-trained generative AI model to generate embeddings from the cleaned text data, which convert the text data into numerical vectors for easier future search and comparison.

[0376] Specifically, the text of parliamentary responses is input into a generative AI model, which outputs a fixed-length vector that is later stored in a database and used as a search index.

[0377] Database Update Implementation Example

[0378] The server registers the generated embedding in its database and updates its search index, allowing it to respond quickly to future user search queries.

[0379] A concrete example is the process of adding newly generated embeddings to a database and rebuilding the search index along with the existing data.

[0380] User Interface Embodiments

[0381] The terminal provides the user with an interface for searching the answer data, which has a user-friendly design and includes a search box and a search result display area.

[0382] A specific example is an interface that provides a search box in a web browser, allowing users to simply enter keywords and display related response data.

[0383] Query Processing Embodiments

[0384] A user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data.

[0385] As a concrete example, when a user enters "environmental policy," the process converts the query into an embedding, searches a database, and lists relevant response data.

[0386] Display of search results

[0387] The server returns the search results to the terminal and displays them visually to the user, allowing the user to easily refer to and utilize the response data of other local governments.

[0388] As a specific example, related defense data is displayed in a list format on the Web, and an interface is provided that allows users to click on details to view the full text.

[0389] By combining the above means, the present invention provides a system that efficiently collects response data from local governments and allows users to easily search and refer to it.

[0390] The processing flow will be explained below.

[0391] Step 1:

[0392] The server accesses the list of website URLs collected and crawls response data from the official websites of each local government.

[0393] Specifically, the server analyzes the HTML structure, extracts links to the response data based on specific HTML tags and classes, and follows those links to download PDF files, etc.

[0394] Step 2:

[0395] The server converts the collected PDF files into text format.

[0396] Specifically, it uses a PDF to text conversion library to extract the text from each page and combine it into a single text file.

[0397] Step 3:

[0398] The server performs data cleaning on the text data.

[0399] Specifically, it removes extra line breaks, special characters, and unnecessary header and footer information to ensure data consistency.

[0400] Step 4:

[0401] The server uses a generative AI model to generate embeddings from the cleaned text data.

[0402] Specifically, text data is input into a pre-trained language model and converted into a fixed-length vector.

[0403] Step 5:

[0404] The server registers the generated embedding in the database and updates the index.

[0405] Specifically, it inserts new entries into the database and rebuilds existing indexes to enable efficient searches.

[0406] Step 6:

[0407] The terminal provides a search interface to the user.

[0408] Specifically, it installs a search box and results display area in a web browser or dedicated application, providing an interface that is easy for users to use.

[0409] Step 7:

[0410] The server receives a search query entered by the user, converts the query into an embedding, and searches the database.

[0411] Specifically, the search query is input into the generative AI model, and the generated embedding is compared with embeddings in the database to extract entries with high similarity.

[0412] Step 8:

[0413] The server returns the search results to the terminal and displays them visually to the user.

[0414] Specifically, the search results are sent to the terminal in JSON format, and the terminal formats the data it receives and displays it to the user.

[0415] Example 1

[0416] 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."

[0417] The data on parliamentary responses made public by local governments is scattered across various websites, making it difficult to centrally collect and easily search and reference. Furthermore, the volume of information is enormous, so manually collecting and organizing the data takes time and effort. Furthermore, variations in linguistic expression and clerical errors make data cleaning difficult, reducing the efficiency of information searches. There is a need to resolve these issues and enable users to efficiently access response data.

[0418] 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.

[0419] In this invention, the server includes means for automatically collecting response data published by each local government, means for converting the collected data into text format and performing data cleaning, means for generating embeddings from the cleaned text data using a generative AI model, and means for registering the generated embeddings in a database and updating an index, thereby enabling efficient collection, processing, and search of response data from local governments.

[0420] "Response data" refers to official statements and responses recorded in local government assemblies and committees.

[0421] "Automatic collection means" refers to a system that uses web crawlers or scraping technology to automatically obtain specified information from specific websites.

[0422] "Text conversion method" refers to a technology that extracts text information from collected PDFs, image files, etc., and converts it into a digital text format.

[0423] "Data cleaning means" is a process of formatting collected text data by removing typos, unnecessary special characters, line breaks, etc.

[0424] A "generative AI model" is an artificial intelligence model that is trained in advance with large amounts of data and exhibits high performance for specific tasks.

[0425] "Embedding generation method" is a technology that converts text data into numerical vectors and expresses their characteristics.

[0426] The "database registration method" is the process of saving the generated embedding to a database and updating the search index.

[0427] A "search query" refers to a search term or phrase that a user enters to find desired information.

[0428] A "user interface" is an interface that includes the screens and functions that a user operates when using a system.

[0429] An "index" refers to a structure or mechanism for efficiently searching data in a database.

[0430] The present invention is a system that automatically collects assembly response data published by local governments, generates embeddings using a generative AI model, and allows users to easily search and reference them. Specific embodiments for implementing the present invention are described below.

[0431] The server accesses the official websites of each local government and automatically collects publicly available parliamentary response data. This data collection is performed using web crawlers and scraping techniques. Specifically, it uses HTML analysis libraries such as BeautifulSoup to analyze the HTML structure of each website and obtain links to response data and PDF files.

[0432] Next, the server converts the collected PDF files into text data. To do this, it uses a library called PyMuPDF to extract text from the PDF. Then, it uses a data cleaning method to remove unnecessary line breaks and special characters from the extracted text data and format it. This results in clean text data.

[0433] Embeddings are generated from the cleaned text data using a generative AI model. This generative AI model uses a pre-trained language model such as GPT-3. Using the text data as input, the model generates a numerical vector called an embedding. This allows the features of the text data to be represented as a numerical vector.

[0434] The generated embeddings are registered in a database by the server, and as part of this registration process, the server also updates a search index, allowing the data in the database to be efficiently searched.

[0435] The user interface is provided using a terminal. The terminal provides an interface with a search box and a result display area so that the user can search for answer data. For example, the interface is implemented using the React.js framework in a web browser.

[0436] A user inputs a search query through this interface. The search query is converted into an embedding by the server. The converted embedding is used to search the database and extract the most relevant response data. For example, if a user inputs the query "environmental policy," the query is converted into an embedding and the relevant response data in the database is listed.

[0437] Finally, the server returns the search results to the terminal and visually displays them to the user, allowing the user to easily browse related answer data and view detailed information.

[0438] A specific example of a prompt sentence could be, "Search for questions about education policy in local councils." By inputting this prompt sentence into a generative AI model, generating an embedding, and searching the database, relevant answer data can be easily obtained.

[0439] The above is a specific embodiment for carrying out the present invention, which realizes a system that can efficiently collect, search, and use assembly response data from local public organizations.

[0440] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0441] Step 1:

[0442] The server accesses the official websites of each local government and automatically collects data on parliamentary responses.

[0443] Specifically, the server uses a web crawler and an HTML parsing library such as BeautifulSoup to analyze the HTML structure of the web page.

[0444] The input is the URL of the official website of each local government, and the output is a link to the response data or a PDF file.

[0445] During this process, the server searches for the specified keywords and file formats and retrieves links containing the response data. For example, the server accesses "https: / / example-localgovsite.jp / kaigi" and retrieves PDF links to the parliamentary response data.

[0446] Step 2:

[0447] The server converts the collected PDF files into text data.

[0448] For this, we use a PDF parsing library called PyMuPDF.

[0449] The input is the PDF files collected in step 1, and the output is text data.

[0450] Specifically, the server opens the PDF file and extracts its contents in text format. For example, it uses PyMuPDF to extract the text from "example.pdf" and saves it as text data.

[0451] Step 3:

[0452] The server performs data cleaning on the text data.

[0453] The input is the text data extracted in step 2, and the output is the cleaned text data.

[0454] Specifically, the server deletes unnecessary line breaks and special characters from the text data and arranges the data. For example, the server deletes line breaks and special characters to obtain text data that has been arranged in a single line.

[0455] Step 4:

[0456] The server uses a generative AI model to generate embeddings from the cleaned text data.

[0457] The input is the text data cleaned in step 3, and the output is the embedding (a numeric vector).

[0458] Specifically, the server inputs the cleaned text data into a generative AI model to obtain embeddings. For example, the server inputs the text data into a GPT-3 model and obtains the generated numeric vectors.

[0459] Step 5:

[0460] The server registers the generated embedding in the database and updates the index.

[0461] The input is the embedding generated in step 4, and the output is the updated database and index.

[0462] Specifically, the server adds the embedding to the database and rebuilds the search index, e.g., registers the embedding in the PostgreSQL database and updates the index.

[0463] Step 6:

[0464] The terminal provides the user with an interface for searching for answer data.

[0465] There is no input, and the output is an interface with a search box and a results display area.

[0466] Specifically, the device displays an interface using React.js on a web browser, allowing users to enter search queries. For example, it provides an interface that displays a search box and a search button.

[0467] Step 7:

[0468] A user enters a search query into the interface, and the server converts the query into an embedding and searches the database.

[0469] The input is a user's search query and the output is a list of relevant answer data.

[0470] Specifically, the server passes the input search query to the generative AI model, generates an embedding, and then searches the database. For example, input the query "environmental policy" and list related response data.

[0471] Step 8:

[0472] The server returns the search results to the terminal and displays them visually to the user.

[0473] The input is the list of search results obtained in step 7, and the output is a visual display that the user can refer to.

[0474] Specifically, the server sends the search results in JSON format to the terminal, and the terminal displays the results in a list format, for example, by displaying a list of related answer data on a web interface so that the user can check the details.

[0475] The above are the specific program processing steps of this system.

[0476] (Application example 1)

[0477] 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."

[0478] In today's brick-and-mortar stores, store employees are required to quickly and accurately provide product information and FAQs in response to customer inquiries. However, with existing systems, this process is often manual and inefficient, which can lead to a decline in the quality of customer service. Furthermore, collecting necessary information from various data sources, integrating it, and making it searchable takes time and effort, increasing the burden on employees. It is necessary to solve these problems and improve customer service and employee work efficiency.

[0479] 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.

[0480] In this invention, the server includes: means for automatically collecting response data published by each local government; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for allowing store employees to quickly search for product information and FAQs in response to customer inquiries; and means for displaying search results to the user. This enables employees to respond to customer inquiries quickly and accurately based on integrated data collected from various information sources. Furthermore, this significantly improves employee work efficiency and contributes to increased customer satisfaction.

[0481] A "local public entity" is a local government that handles administrative affairs as a local government separate from the nation.

[0482] "Response data" refers to data that records the content of questions and answers that were asked in parliaments or conferences.

[0483] "Automatic collection means" refers to a mechanism by which a system or program collects data automatically, without human intervention.

[0484] "Means of converting to text format" refers to the process of converting data such as images and PDFs into a format that can be handled as text information.

[0485] "Data cleaning" is the process of removing unnecessary information and errors from collected data, making it easier to analyze.

[0486] A "generative AI model" is an artificial intelligence system that has been pre-trained using machine learning or deep learning to perform a specific task.

[0487] An "embedding" is a representation of text data converted into a numerical vector that is used when processing it with a machine learning model.

[0488] A "database" is an information management system that systematically stores specific information and allows it to be searched and used when needed.

[0489] "Means of updating the index" refers to the process of updating the reference system to make searches more efficient for newly added data.

[0490] An "interface" is the hardware and software structure through which a user interacts with a computer system.

[0491] A "search query" is a keyword or phrase that a user enters into a search system to retrieve specific information.

[0492] "Customer inquiries" are questions or clarifications made by store patrons regarding products or services.

[0493] "Product information" is a detailed description of the product's specifications, price, usage, etc.

[0494] A "FAQ" is a collection of frequently asked questions and their answers.

[0495] The "means for displaying search results" is a method for visually conveying to the user the data retrieved based on the query entered by the user.

[0496] MODE FOR CARRYING OUT THE INVENTION

[0497] This invention provides a system that collects response data from local governments based on a series of procedures and enables quick and efficient search using a generative AI model. Specific embodiments of this system are described below.

[0498] Data collection implementation

[0499] The server accesses the official websites of each local government and automatically collects publicly available response data. During this process, a web crawler running on the server analyzes the HTML structure and extracts only the necessary information from specified elements. For example, it automatically downloads PDF links to response data on each local government's website. The software used is requests and BeautifulSoup.

[0500] Data Preprocessing Embodiments

[0501] The server converts the collected PDF files into text data. To do this, it uses a dedicated library called pdfminer to extract text from the PDF. Data cleaning is then performed to remove unnecessary line breaks and special characters and format the text data.

[0502] An embodiment of embedding generation

[0503] The server uses a pre-trained generative AI model (e.g., SentenceTransformer) to generate embeddings from the cleaned text data, which convert the text data into numerical vectors for easier later search and comparison.

[0504] Database Update Implementation Example

[0505] The server registers the generated embedding in a database and updates the index, allowing for quick responses when users later enter search queries. The database uses a data management system such as SQLite.

[0506] User Interface Embodiments

[0507] The terminal provides the user with an interface for searching for answer data. This interface has a user-friendly design, including a search box and a search result display area. As a specific example, a search box is installed in a web browser, and the user can simply enter keywords to display related answer data.

[0508] Query Processing Embodiments

[0509] When a user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data. For example, if a user enters "environmental policy," the server converts the query into an embedding and searches the database to list relevant answer data.

[0510] Display of search results

[0511] The server returns the search results to the terminal and visually displays them to the user, allowing the user to easily refer to and utilize related answer data. For example, related answer data is displayed in a list format on the web, and an interface is provided where the user can click on details to view the full text.

[0512] Specific examples

[0513] For example, if a drugstore staff member is asked by a customer about "side effects of allergy medication," they can use this system to quickly search and refer to the information. The staff member simply types "side effects of allergy medication" into the tablet or smartphone interface, and the system instantly displays relevant information.

[0514] Prompt Sentence Examples

[0515] "Please tell me about the side effects of allergy medication."

[0516] In this way, a system is created in which complex information processing is seamlessly performed on the back end, allowing store staff to quickly and accurately answer customer questions.

[0517] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0518] Step 1:

[0519] The server accesses the official websites of each local government and automatically collects publicly available response data. Specifically, a web crawler running on the server analyzes the HTML structure and extracts PDF links to the response data. It receives a list of website URLs as input and downloads and saves the PDF data as output.

[0520] Step 2:

[0521] The server converts the collected PDF files into text data. Specifically, it uses the pdfminer library to extract text from the PDF, and then cleans it by removing unnecessary line breaks and special characters. It takes PDF files as input and generates clean text data as output.

[0522] Step 3:

[0523] The server uses a generative AI model (e.g., SentenceTransformer) to generate embeddings from the cleaned text data. This process involves calculations to convert the text data into a numerical vector. It receives the clean text data as input and generates embeddings (numerical vectors) as output.

[0524] Step 4:

[0525] The server registers the generated embedding in a database and updates the index. Specifically, it uses a data management system such as SQLite to insert the newly generated embedding and the original text data into the database and update the search index. It receives the generated embedding and the original text data as input and obtains an updated database as output.

[0526] Step 5:

[0527] The terminal provides a user with an interface for searching answer data. The interface displays a search box where the user can enter a search query and a results display area. The terminal receives a search query from the user as input and prepares an output to process the query.

[0528] Step 6:

[0529] The server converts the user's search query into an embedding and searches the database. Specifically, the query is input into a generative AI model to generate an embedding. It then compares it with existing embeddings in the database to extract the most relevant answer data. It receives the user's search query as input and obtains search results as output.

[0530] Step 7:

[0531] The server returns the search results to the terminal and displays them visually to the user. For example, the server may display the relevant answer data in a list format, and provide an interface that allows the user to click on details to view the full text. The server receives the search results as input and generates the data to be displayed to the user as output.

[0532] 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.

[0533] The present invention provides a system that automatically collects parliamentary response data made public by local governments and can efficiently search it using embeddings generated using a generative AI model, as well as a system that combines an emotion engine that recognizes the user's emotions.

[0534] Data collection implementation

[0535] The server accesses the official websites of each local government and crawls the publicly available data on parliamentary responses. The server analyzes the HTML structure and extracts links to the response data based on specific HTML tags and classes. The server then follows the links and automatically downloads PDF files and other data.

[0536] A specific example is a process for collecting PDF files related to parliamentary responses from the websites of multiple local governments in bulk.

[0537] Data Preprocessing Embodiments

[0538] The server converts the collected PDF files into text format. The server uses a PDF-to-text conversion library to extract the text from each page and compile it into a single text file. The server then performs data cleaning to remove unnecessary line breaks, special characters, and unnecessary header and footer information.

[0539] As a concrete example, it removes unnecessary information from text data extracted from a PDF file and generates text data that is formatted to be easy to read.

[0540] An embodiment of embedding generation

[0541] The server uses a pre-trained generative AI model to generate embeddings from the cleaned text data, which convert the text data into numerical vectors that can be used for efficient search and comparison.

[0542] Specifically, the text of parliamentary responses is input into a generative AI model, which outputs a fixed-length vector that is later stored in a database and used as a search index.

[0543] Database Update Implementation Example

[0544] The server registers the generated embedding in its database and updates its search index, allowing it to respond quickly when a user enters a search query.

[0545] The specific operation involves adding the newly generated embeddings to the database and rebuilding the index along with the existing data.

[0546] User Interface Embodiments

[0547] The terminal provides the user with an interface for searching answer data. The user interface includes a search box and a search result display area, and is designed to be easy for users to operate.

[0548] A specific example is an interface that provides a search box in a web browser, and when a user enters a keyword, related answer data is displayed.

[0549] Query Processing Embodiments

[0550] A user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data.

[0551] As a concrete example, when a user inputs "environmental policy," the query is converted into an embedding, a database search is performed, and relevant response data is extracted.

[0552] Embodiment of search result display

[0553] The server returns the search results to the terminal and displays them visually to the user, who can then easily refer to and manipulate the response data of other local governments.

[0554] As a concrete example, related response data is displayed in a list format on the web, and users can click on each item to view details.

[0555] Embodiment of Emotion Engine

[0556] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when entering a search query. If the user is in a positive emotion, this emotion engine prioritizes the display of highly relevant search results, and if the user is in a negative emotion, it displays appropriate results and messages to reduce stress.

[0557] A specific example is an interface in which an emotion engine analyzes a user's facial expressions and vocal tone, and if they are judged to be negative, provides relaxing information.

[0558] By combining the above methods, the present invention provides a system that efficiently collects response data from local governments and allows users to easily search and refer to it. Furthermore, by combining it with an emotion engine, it is possible to provide optimal search results according to the user's emotional state, improving the user experience.

[0559] The processing flow will be explained below.

[0560] Step 1:

[0561] The server accesses the official websites of each local government and crawls the data on parliamentary responses. Based on the URL list, the server analyzes the HTML structure of each website, extracts links to the response data from elements that match specific HTML tags and classes, and follows the links to download PDF files.

[0562] Step 2:

[0563] The server converts the collected PDF files into text format. The server uses a dedicated library for extracting text from PDFs, extracts the text for each page, and saves it as a single text file.

[0564] Step 3:

[0565] The server performs data cleaning on the text data. Specifically, it removes unnecessary line breaks, special characters, and unnecessary header and footer information to generate formatted text data. This prepares the data in a consistent, parseable format.

[0566] Step 4:

[0567] The server uses a generative AI model to generate embeddings from the cleaned text data. The server inputs the text data into a pre-trained language model and outputs a fixed-length numeric vector (embedding). This embedding is a numeric representation of the text's meaning.

[0568] Step 5:

[0569] The server registers the generated embeddings in the database and updates the index. The server adds new embeddings to the database and rebuilds existing data indexes to improve search efficiency.

[0570] Step 6:

[0571] The terminal provides the user with an interface for searching answer data. Specifically, a user-friendly interface is provided with a search box and a search result display area, allowing the user to easily operate it.

[0572] Step 7:

[0573] The server receives the user's search query, converts the query into an embedding, and searches the database. The user-entered search query is fed into a generative AI model, which compares the converted embedding with existing embeddings in the database.

[0574] Step 8:

[0575] The server sends the search results to the device and displays them visually to the user. The search results are sent to the device in JSON format, and the device formats the received data and displays it to the user in a list format.

[0576] Step 9:

[0577] The server is equipped with an emotion engine that recognizes the user's emotions and optimizes search results according to the user's emotional state. When a user enters a search query, the emotion engine analyzes the user's facial expressions and tone of voice, and prioritizes relevant search results if the user's emotions are positive, and displays appropriate results and messages to reduce stress if the user's emotions are negative.

[0578] Step 10:

[0579] The server reflects the results of the emotion engine and adjusts the ranking of search results to provide an optimal user experience. By displaying search results with the display order adjusted based on the emotion engine, users can obtain information efficiently and without stress.

[0580] Example 2

[0581] 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."

[0582] Although data on responses from local governments is made public by each organization, there is no way to efficiently collect it and organize it in a searchable format. Furthermore, there is no system in existence that uses an interface that takes user emotions into consideration to efficiently search for the necessary response data while reducing stress. This forces users to manually search for the information they need from large amounts of data, which is a laborious and time-consuming process.

[0583] 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.

[0584] In this invention, the server includes: means for automatically collecting response data published by each local government; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for displaying search results to the user; and means including an emotion engine for recognizing the user's emotional state and displaying search results and appropriate messages. This allows efficient collection of response data from local governments, allows users to easily search and reference it, and provides optimal search results according to the user's emotional state.

[0585] "Response data" refers to data that records the answers and explanations given by administrative representatives to questions from assembly members in local government assemblies.

[0586] A "text format" is a form of digital data that is recorded as human-readable character information.

[0587] "Data cleaning" is the process of removing unnecessary information and errors from collected data and formatting it.

[0588] A "generative AI model" is an artificial intelligence model that has been pre-trained with large amounts of data and is capable of generating appropriate outputs for specific inputs.

[0589] An "embedding" is a conversion of text data into a numerical vector, which is used for efficient searching and comparison.

[0590] A "database" is an information system that efficiently stores and manages large amounts of data and allows quick access when needed.

[0591] A "search index" is an index used to efficiently search for information in a database.

[0592] An "interface" is a point of contact or means for exchanging information between a user and a system, and includes, for example, a search box or a display screen.

[0593] A "search query" refers to an inquiry that a user enters into a search system, and the system searches for related information based on the query.

[0594] The "emotion engine" is a system component that has the function of recognizing the user's emotional state and providing appropriate information and messages according to that state.

[0595] This invention is a system that automatically collects parliamentary response data published by local governments and efficiently searches it using embeddings generated by a generative AI model. Furthermore, we aim to improve the search experience by combining it with an emotion engine that recognizes the user's emotions.

[0596] Data collection

[0597] The server accesses the official websites of each local government and crawls the publicly available data on assembly responses. Specifically, it analyzes the HTML structure, extracts links to the response data based on specific HTML tags and classes, and automatically downloads PDF files by following those links. For example, the server accesses the Tokyo Metropolitan Assembly website, extracts links to PDF files containing keywords including "response," and downloads them.

[0598] Data Preprocessing

[0599] The server converts the collected PDF files into text format. Using a library for converting PDF to text format (e.g., PyPDF2), the server extracts text from each page and compiles it into a single text file. It then performs data cleaning, removing unnecessary line breaks, special characters, and unnecessary header and footer information. Specifically, the server uses PyPDF2 to extract text from the downloaded PDF files, removes unnecessary information, and creates a unified text document.

[0600] embedding generation

[0601] The server uses a pre-trained generative AI model (e.g., a pre-trained language model) to generate embeddings from the cleaned text data. These embeddings convert the text data into numerical vectors that can be used for efficient search and comparison. For example, the server inputs the cleaned text data into a language model, which outputs a fixed-length vector. This vector is stored in a database and serves as a search index.

[0602] Database Update

[0603] The server registers the generated embedding in the database and updates the search index. Specifically, the server adds the newly generated embedding to the existing database and rebuilds the search index, allowing for quick responses when a user enters a search query.

[0604] User Interface

[0605] The terminal provides the user with an interface for searching for answer data. The interface is equipped with a search box and a search result display area, and is designed to be easy for users to operate. For example, if a search box is provided in a web browser and the user enters a keyword, related answer data will be displayed.

[0606] Query processing and search result display

[0607] A user enters a search query, and the server converts the query into an embedding and searches the database. The server compares the generated embedding with existing embeddings in the database and extracts the most relevant answer data. For example, if a user enters "environmental policy," the server converts the query into an embedding and searches the database to extract relevant answer data. The server returns the results to the device and displays them visually to the user.

[0608] Emotion engine integration

[0609] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when they enter a search query. This emotion engine prioritizes relevant search results if the emotion is positive, and displays appropriate results and messages to reduce stress if the emotion is negative. For example, the emotion engine analyzes the user's facial expressions and tone of voice, and if it determines that the emotion is negative, it provides an interface with information to help users relax.

[0610] This system efficiently collects response data from local governments and provides an environment where users can easily search and refer to it. Furthermore, by combining it with an emotion engine, it is possible to provide optimal search results according to the user's emotional state, improving the user experience.

[0611] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0612] Step 1: Data collection

[0613] The server accesses the official websites of each local government and crawls the publicly available data on parliamentary responses. The input is the URL of the local government's website, and the output is a link to a PDF file of the parliamentary response data. Specifically, the server analyzes the HTML structure of the webpage, extracts links to the response data based on specific HTML tags and classes, and automatically downloads the PDF file by following the link.

[0614] Step 2: Data Preprocessing

[0615] The server converts the collected PDF files into text format. The input is a PDF file, and the output is text data. Specifically, a library for converting PDF to text (e.g., PyPDF2) is used to extract the text from each page and combine it into a single text file. The data is then cleaned to remove unnecessary line breaks, special characters, and unnecessary header and footer information.

[0616] Step 3: Embedding generation

[0617] The server generates embeddings from the cleaned text data. The input is the cleaned text data, and the output is the embedding (a numerical vector). Specifically, it uses a pre-trained generative AI model (e.g., a pre-trained language model) to convert the text data into a numerical vector, which makes the text data in a format that can be efficiently searched and compared.

[0618] Step 4: Update the database

[0619] The server registers the generated embedding in the database and updates the search index. The input is the generated embedding, and the output is the updated search index. The specific operation is to add the newly generated embedding to the existing database and rebuild the search index. This allows for quick response when a user enters a search query.

[0620] Step 5: User Interface

[0621] The terminal provides the user with an interface for searching for answer data. The input is the user's operation, and the output is the interface that displays the search results. Specifically, the operation is to provide a search box and a search result display area in the web browser, allowing the user to enter keywords. Related answer data is displayed based on the entered keywords.

[0622] Step 6: Query Processing

[0623] The server receives a search query entered by a user, converts the query into an embedding, and searches a database. The input is the user's search query (text), and the output is related answer data. Specifically, the server inputs the user's search query into a generative AI model, searches the database using the resulting embedding, and extracts the most relevant answer data.

[0624] Step 7: Displaying search results

[0625] The server returns the search results to the terminal and displays them visually to the user. The input is the answer data extracted by the search, and the output is the displayed search results. Specifically, the server finds the relevant answer data and sends it to the terminal, and the terminal displays the search results in list format on the web browser. The user can click on each item to view details.

[0626] Step 8: Integrating the Emotion Engine

[0627] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when they enter a search query. The input is the user's facial expression and tone of voice, and the output is appropriate search results and messages. Specifically, the emotion engine analyzes the user's emotional state, and if the emotion is positive, it prioritizes relevant search results, and if the emotion is negative, it provides appropriate messages and information to reduce stress.

[0628] (Application example 2)

[0629] 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."

[0630] Factories require efficient management of maintenance information for multiple pieces of equipment and machinery. However, manual collection and analysis is time-consuming and labor-intensive, often resulting in insufficient data management and analysis. Furthermore, while it is known that workers' emotional states affect work efficiency and safety, there is a lack of systems that address this issue. Therefore, there is a need for a system that can automatically collect and efficiently search maintenance information, as well as a system that can provide feedback based on workers' emotions.

[0631] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: means for automatically collecting response data made public by local governments; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for displaying search results to the user; means for processing maintenance information collected from a machine as text data; and means for providing search results according to the user's emotional state by incorporating an emotion engine for recognizing the user's emotions. This not only enables efficient collection and search of maintenance information, but also enables appropriate feedback to be provided according to the worker's emotional state.

[0632] 1. "Response data" refers to data showing the content of responses given in local government assemblies and other such meetings.

[0633] 2. "Text format" means a data format expressed as human-readable characters or sentences.

[0634] 3. "Data cleaning" is the process of removing redundant information and noise from collected text data.

[0635] 4. A "generative AI model" is an artificial intelligence model that is pre-trained using massive amounts of data.

[0636] 5. "Embedding" is a method of representing text data as a numerical vector.

[0637] 6. A "database" is a system for efficiently storing, managing, and retrieving structured data.

[0638] 7. "Interface" refers to the operation screens and input fields that allow users to interact with the system.

[0639] 8. A "query" is a question or keyword entered by a user to search for specific information.

[0640] 9. "Search query" means input to the system to specify the information to be searched.

[0641] 10. "Machinery" refers to production equipment and devices installed in factories and production lines.

[0642] 11. "Maintenance Information" means data relating to the maintenance and repair of facilities and machinery.

[0643] 12. An "emotion engine" is an algorithm or system for determining a user's emotional state.

[0644] 13. "Feedback" refers to the response or information that a system returns in response to a user's input or actions.

[0645] To implement this invention, a server first accesses the official websites of each local government and automatically collects publicly available response data. Specifically, the server uses web crawling technology to extract and download response data links based on specific HTML tags and classes. For example, there is a process for collecting PDF files from the websites of multiple local government organizations in bulk.

[0646] Next, the server converts the collected PDF files into text format. This is done by extracting the text from each page using a PDF to text conversion library (e.g., PyPDF2) and compiling it into a single text file. After that, data cleaning is performed, removing unnecessary line breaks and special characters, and generating formatted text data. For example, there is a process to remove unnecessary information from the text data extracted from the PDF files.

[0647] The server then uses a generative AI model (e.g., OpenAI GPT-3 / BERT) to generate embeddings from the cleaned text data. These embeddings convert the text data into numerical vectors that can be used for efficient search and comparison. Specifically, the text data is input into the generative AI model, which outputs a fixed-length vector. This vector is later stored in a database and used as a search index.

[0648] The server then registers the generated embeddings in a database (e.g., Elasticsearch) and updates the index, allowing for faster response when a user enters a search query. Specifically, the process involves adding the newly generated embeddings to the database and rebuilding the index with the existing data.

[0649] The server also provides a user interface to allow users to search for the answer data. This interface is installed on a terminal and includes a search box and a search result display area. For example, there is an interface in which a search box is provided in a web browser, and when a user enters a keyword, related answer data is displayed.

[0650] Next, when a user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data. For example, if a user enters "machine wear diagnosis," there is a process to convert the query into an embedding, perform a database search, and extract relevant answer data.

[0651] The server also returns the search results to the terminal and visually displays them to the user. The user can easily refer to and manipulate the related answer data. For example, the related answer data is displayed in list format on the web, and the user can click on each item to view details.

[0652] Finally, the server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when entering a search query. This emotion engine prioritizes relevant search results when the user is in a positive emotion, and displays appropriate results and messages to reduce stress when the user is in a negative emotion. For example, the emotion engine analyzes the user's facial expressions and tone of voice, and if the engine determines that the user is in a negative emotion, it provides an interface that provides information to help them relax.

[0653] Prompt Sentence Examples

[0654] Parliamentary response data analysis. Generate an embedding from the following text:

[0655] "Answer regarding environmental policy..."

[0656] This invention not only makes it possible to efficiently collect and search maintenance information, but also makes it possible to provide appropriate feedback to the worker according to his or her emotional state.

[0657] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0658] Step 1:

[0659] The server accesses the official websites of local government organizations and collects response data. Specifically, the server crawls multiple specified websites using web crawling technology (e.g., BeautifulSoup, Scrapy), extracts links to response data based on specific HTML tags and classes, and downloads PDF files. The input is the website URL, and the output is the downloaded PDF file.

[0660] Step 2:

[0661] The server converts the collected PDF files into text format. Specifically, it uses the PyPDF2 library to extract text from each page of the PDF file and combines it into a single text file. The input is a PDF file and the output is a text file.

[0662] Step 3:

[0663] The server cleans the text files. Specifically, it uses libraries such as NLTK to remove unnecessary line breaks, special characters, and unnecessary header and footer information, generating readable, formatted text data. The input is the raw text file, and the output is the cleaned text data.

[0664] Step 4:

[0665] The server generates embeddings from the cleaned text data using a generative AI model (e.g., OpenAI GPT-3 / BERT). Specifically, the text data is input to the AI ​​model as a prompt sentence, and a fixed-length vector is output. The input is the cleaned text data, and the output is the generated embedding.

[0666] Step 5:

[0667] The server registers the generated embeddings in a database (e.g., Elasticsearch) and updates the index. Specifically, it adds the newly generated embeddings to the database and rebuilds the index with the existing data. The input is the generated embeddings, and the output is the updated database index.

[0668] Step 6:

[0669] The terminal provides the user with an interface for searching answer data via a user interface. Specifically, the terminal displays a web browser interface including a search box and a search result display area. The input is the user interface design specification, and the output is the interface available to the user.

[0670] Step 7:

[0671] When a user enters a search query, the server converts the query into an embedding and searches the database. Specifically, the search query is input into the AI ​​model, and the generated embedding is compared with existing embeddings in the database. The input is the user's search query, and the output is highly relevant answer data.

[0672] Step 8:

[0673] The server returns the search results to the terminal and displays them visually to the user. Specifically, it sends the relevant answer data in a list format to the terminal so that the user can view the details. The input is the search results, and the output is the displayed search results.

[0674] Step 9:

[0675] The server uses an emotion engine to recognize the user's emotional state. Specifically, it analyzes the user's facial expressions and voice tone using the device's camera and microphone. The input is real-time video and audio data, and the output is the user's emotional state.

[0676] Step 10:

[0677] The server provides appropriate search results and messages based on the user's emotional state. Specifically, if the user's emotional state is positive, it prioritizes relevant results, and if the user's emotional state is negative, it displays relaxing information and messages. The input is the user's emotional state, and the output is the adjusted search results and messages.

[0678] 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.

[0679] 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.

[0680] 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.

[0681] [Third embodiment]

[0682] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0683] 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.

[0684] 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).

[0685] 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.

[0686] 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.

[0687] 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).

[0688] 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.

[0689] 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.

[0690] 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.

[0691] 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.

[0692] 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.

[0693] 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."

[0694] This invention provides a system that collects parliamentary response data made public by local governments, generates embeddings using a generative AI model, and allows users to easily search and reference the data.

[0695] Data collection implementation

[0696] The server accesses the official websites of each local government and automatically crawls the publicly available parliamentary response data. During this process, the web crawler analyzes the HTML structure and extracts only the necessary information from the specified elements. For example, it automatically downloads the PDF link for the response data on each local government website.

[0697] A specific example is a process that collects PDF files related to responses from the websites of multiple local governments and saves them in a specific folder.

[0698] Data Preprocessing Embodiments

[0699] The server converts the collected PDF files into text data. This involves extracting text from the PDF using a dedicated library. Data cleaning is then performed. This cleaning process removes unnecessary line breaks and special characters, resulting in pure text data.

[0700] A specific example is deleting unnecessary header and footer information from text data extracted from a PDF file and formatting it to leave only text.

[0701] An embodiment of embedding generation

[0702] The server uses a pre-trained generative AI model to generate embeddings from the cleaned text data, which convert the text data into numerical vectors for easier future search and comparison.

[0703] Specifically, the text of parliamentary responses is input into a generative AI model, which outputs a fixed-length vector that is later stored in a database and used as a search index.

[0704] Database Update Implementation Example

[0705] The server registers the generated embedding in its database and updates its search index, allowing it to respond quickly to future user search queries.

[0706] A concrete example is the process of adding newly generated embeddings to a database and rebuilding the search index along with the existing data.

[0707] User Interface Embodiments

[0708] The terminal provides the user with an interface for searching the answer data, which has a user-friendly design and includes a search box and a search result display area.

[0709] A specific example is an interface that provides a search box in a web browser, allowing users to simply enter keywords and display related response data.

[0710] Query Processing Embodiments

[0711] A user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data.

[0712] As a concrete example, when a user enters "environmental policy," the process converts the query into an embedding, searches a database, and lists relevant response data.

[0713] Display of search results

[0714] The server returns the search results to the terminal and displays them visually to the user, allowing the user to easily refer to and utilize the response data of other local governments.

[0715] As a specific example, related defense data is displayed in a list format on the Web, and an interface is provided that allows users to click on details to view the full text.

[0716] By combining the above means, the present invention provides a system that efficiently collects response data from local governments and allows users to easily search and refer to it.

[0717] The processing flow will be explained below.

[0718] Step 1:

[0719] The server accesses the list of website URLs collected and crawls response data from the official websites of each local government.

[0720] Specifically, the server analyzes the HTML structure, extracts links to the response data based on specific HTML tags and classes, and follows those links to download PDF files, etc.

[0721] Step 2:

[0722] The server converts the collected PDF files into text format.

[0723] Specifically, it uses a PDF to text conversion library to extract the text from each page and combine it into a single text file.

[0724] Step 3:

[0725] The server performs data cleaning on the text data.

[0726] Specifically, it removes extra line breaks, special characters, and unnecessary header and footer information to ensure data consistency.

[0727] Step 4:

[0728] The server uses a generative AI model to generate embeddings from the cleaned text data.

[0729] Specifically, text data is input into a pre-trained language model and converted into a fixed-length vector.

[0730] Step 5:

[0731] The server registers the generated embedding in the database and updates the index.

[0732] Specifically, it inserts new entries into the database and rebuilds existing indexes to enable efficient searches.

[0733] Step 6:

[0734] The terminal provides a search interface to the user.

[0735] Specifically, it installs a search box and results display area in a web browser or dedicated application, providing an interface that is easy for users to use.

[0736] Step 7:

[0737] The server receives a search query entered by the user, converts the query into an embedding, and searches the database.

[0738] Specifically, the search query is input into the generative AI model, and the generated embedding is compared with embeddings in the database to extract entries with high similarity.

[0739] Step 8:

[0740] The server returns the search results to the terminal and displays them visually to the user.

[0741] Specifically, the search results are sent to the terminal in JSON format, and the terminal formats the data it receives and displays it to the user.

[0742] Example 1

[0743] 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."

[0744] The data on parliamentary responses made public by local governments is scattered across various websites, making it difficult to centrally collect and easily search and reference. Furthermore, the volume of information is enormous, so manually collecting and organizing the data takes time and effort. Furthermore, variations in linguistic expression and clerical errors make data cleaning difficult, reducing the efficiency of information searches. There is a need to resolve these issues and enable users to efficiently access response data.

[0745] 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.

[0746] In this invention, the server includes means for automatically collecting response data published by each local government, means for converting the collected data into text format and performing data cleaning, means for generating embeddings from the cleaned text data using a generative AI model, and means for registering the generated embeddings in a database and updating an index, thereby enabling efficient collection, processing, and search of response data from local governments.

[0747] "Response data" refers to official statements and responses recorded in local government assemblies and committees.

[0748] "Automatic collection means" refers to a system that uses web crawlers or scraping technology to automatically obtain specified information from specific websites.

[0749] "Text conversion method" refers to a technology that extracts text information from collected PDFs, image files, etc., and converts it into a digital text format.

[0750] "Data cleaning means" is a process of formatting collected text data by removing typos, unnecessary special characters, line breaks, etc.

[0751] A "generative AI model" is an artificial intelligence model that is trained in advance with large amounts of data and exhibits high performance for specific tasks.

[0752] "Embedding generation method" is a technology that converts text data into numerical vectors and expresses their characteristics.

[0753] The "database registration method" is the process of saving the generated embedding to a database and updating the search index.

[0754] A "search query" refers to a search term or phrase that a user enters to find desired information.

[0755] A "user interface" is an interface that includes the screens and functions that a user operates when using a system.

[0756] An "index" refers to a structure or mechanism for efficiently searching data in a database.

[0757] The present invention is a system that automatically collects assembly response data published by local governments, generates embeddings using a generative AI model, and allows users to easily search and reference them. Specific embodiments for implementing the present invention are described below.

[0758] The server accesses the official websites of each local government and automatically collects publicly available parliamentary response data. This data collection is performed using web crawlers and scraping techniques. Specifically, it uses HTML analysis libraries such as BeautifulSoup to analyze the HTML structure of each website and obtain links to response data and PDF files.

[0759] Next, the server converts the collected PDF files into text data. To do this, it uses a library called PyMuPDF to extract text from the PDF. Then, it uses a data cleaning method to remove unnecessary line breaks and special characters from the extracted text data and format it. This results in clean text data.

[0760] Embeddings are generated from the cleaned text data using a generative AI model. This generative AI model uses a pre-trained language model such as GPT-3. Using the text data as input, the model generates a numerical vector called an embedding. This allows the features of the text data to be represented as a numerical vector.

[0761] The generated embeddings are registered in a database by the server, and as part of this registration process, the server also updates a search index, allowing the data in the database to be efficiently searched.

[0762] The user interface is provided using a terminal. The terminal provides an interface with a search box and a result display area so that the user can search for answer data. For example, the interface is implemented using the React.js framework in a web browser.

[0763] A user inputs a search query through this interface. The search query is converted into an embedding by the server. The converted embedding is used to search the database and extract the most relevant response data. For example, if a user inputs the query "environmental policy," the query is converted into an embedding and the relevant response data in the database is listed.

[0764] Finally, the server returns the search results to the terminal and visually displays them to the user, allowing the user to easily browse related answer data and view detailed information.

[0765] A specific example of a prompt sentence could be, "Search for questions about education policy in local councils." By inputting this prompt sentence into a generative AI model, generating an embedding, and searching the database, relevant answer data can be easily obtained.

[0766] The above is a specific embodiment for carrying out the present invention, which realizes a system that can efficiently collect, search, and use assembly response data from local public organizations.

[0767] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0768] Step 1:

[0769] The server accesses the official websites of each local government and automatically collects data on parliamentary responses.

[0770] Specifically, the server uses a web crawler and an HTML parsing library such as BeautifulSoup to analyze the HTML structure of the web page.

[0771] The input is the URL of the official website of each local government, and the output is a link to the response data or a PDF file.

[0772] During this process, the server searches for the specified keywords and file formats and retrieves links containing the response data. For example, the server accesses "https: / / example-localgovsite.jp / kaigi" and retrieves PDF links to the parliamentary response data.

[0773] Step 2:

[0774] The server converts the collected PDF files into text data.

[0775] For this, we use a PDF parsing library called PyMuPDF.

[0776] The input is the PDF files collected in step 1, and the output is text data.

[0777] Specifically, the server opens the PDF file and extracts its contents in text format. For example, it uses PyMuPDF to extract the text from "example.pdf" and saves it as text data.

[0778] Step 3:

[0779] The server performs data cleaning on the text data.

[0780] The input is the text data extracted in step 2, and the output is the cleaned text data.

[0781] Specifically, the server deletes unnecessary line breaks and special characters from the text data and arranges the data. For example, the server deletes line breaks and special characters to obtain text data that has been arranged in a single line.

[0782] Step 4:

[0783] The server uses a generative AI model to generate embeddings from the cleaned text data.

[0784] The input is the text data cleaned in step 3, and the output is the embedding (a numeric vector).

[0785] Specifically, the server inputs the cleaned text data into a generative AI model to obtain embeddings. For example, the server inputs the text data into a GPT-3 model and obtains the generated numeric vectors.

[0786] Step 5:

[0787] The server registers the generated embedding in the database and updates the index.

[0788] The input is the embedding generated in step 4, and the output is the updated database and index.

[0789] Specifically, the server adds the embedding to the database and rebuilds the search index, e.g., registers the embedding in the PostgreSQL database and updates the index.

[0790] Step 6:

[0791] The terminal provides the user with an interface for searching for answer data.

[0792] There is no input, and the output is an interface with a search box and a results display area.

[0793] Specifically, the device displays an interface using React.js on a web browser, allowing users to enter search queries. For example, it provides an interface that displays a search box and a search button.

[0794] Step 7:

[0795] A user enters a search query into the interface, and the server converts the query into an embedding and searches the database.

[0796] The input is a user's search query and the output is a list of relevant answer data.

[0797] Specifically, the server passes the input search query to the generative AI model, generates an embedding, and then searches the database. For example, input the query "environmental policy" and list related response data.

[0798] Step 8:

[0799] The server returns the search results to the terminal and displays them visually to the user.

[0800] The input is the list of search results obtained in step 7, and the output is a visual display that the user can refer to.

[0801] Specifically, the server sends the search results in JSON format to the terminal, and the terminal displays the results in a list format, for example, by displaying a list of related answer data on a web interface so that the user can check the details.

[0802] The above are the specific program processing steps of this system.

[0803] (Application example 1)

[0804] 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."

[0805] In today's brick-and-mortar stores, store employees are required to quickly and accurately provide product information and FAQs in response to customer inquiries. However, with existing systems, this process is often manual and inefficient, which can lead to a decline in the quality of customer service. Furthermore, collecting necessary information from various data sources, integrating it, and making it searchable takes time and effort, increasing the burden on employees. It is necessary to solve these problems and improve customer service and employee work efficiency.

[0806] 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.

[0807] In this invention, the server includes: means for automatically collecting response data published by each local government; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for allowing store employees to quickly search for product information and FAQs in response to customer inquiries; and means for displaying search results to the user. This enables employees to respond to customer inquiries quickly and accurately based on integrated data collected from various information sources. Furthermore, this significantly improves employee work efficiency and contributes to increased customer satisfaction.

[0808] A "local public entity" is a local government that handles administrative affairs as a local government separate from the nation.

[0809] "Response data" refers to data that records the content of questions and answers that were asked in parliaments or conferences.

[0810] "Automatic collection means" refers to a mechanism by which a system or program collects data automatically, without human intervention.

[0811] "Means of converting to text format" refers to the process of converting data such as images and PDFs into a format that can be handled as text information.

[0812] "Data cleaning" is the process of removing unnecessary information and errors from collected data, making it easier to analyze.

[0813] A "generative AI model" is an artificial intelligence system that has been pre-trained using machine learning or deep learning to perform a specific task.

[0814] An "embedding" is a representation of text data converted into a numerical vector that is used when processing it with a machine learning model.

[0815] A "database" is an information management system that systematically stores specific information and allows it to be searched and used when needed.

[0816] "Means of updating the index" refers to the process of updating the reference system to make searches more efficient for newly added data.

[0817] An "interface" is the hardware and software structure through which a user interacts with a computer system.

[0818] A "search query" is a keyword or phrase that a user enters into a search system to retrieve specific information.

[0819] "Customer inquiries" are questions or clarifications made by store patrons regarding products or services.

[0820] "Product information" is a detailed description of the product's specifications, price, usage, etc.

[0821] A "FAQ" is a collection of frequently asked questions and their answers.

[0822] The "means for displaying search results" is a method for visually conveying to the user the data retrieved based on the query entered by the user.

[0823] MODE FOR CARRYING OUT THE INVENTION

[0824] This invention provides a system that collects response data from local governments based on a series of procedures and enables quick and efficient search using a generative AI model. Specific embodiments of this system are described below.

[0825] Data collection implementation

[0826] The server accesses the official websites of each local government and automatically collects publicly available response data. During this process, a web crawler running on the server analyzes the HTML structure and extracts only the necessary information from specified elements. For example, it automatically downloads PDF links to response data on each local government's website. The software used is requests and BeautifulSoup.

[0827] Data Preprocessing Embodiments

[0828] The server converts the collected PDF files into text data. To do this, it uses a dedicated library called pdfminer to extract text from the PDF. Data cleaning is then performed to remove unnecessary line breaks and special characters and format the text data.

[0829] An embodiment of embedding generation

[0830] The server uses a pre-trained generative AI model (e.g., SentenceTransformer) to generate embeddings from the cleaned text data, which convert the text data into numerical vectors for easier later search and comparison.

[0831] Database Update Implementation Example

[0832] The server registers the generated embedding in a database and updates the index, allowing for quick responses when users later enter search queries. The database uses a data management system such as SQLite.

[0833] User Interface Embodiments

[0834] The terminal provides the user with an interface for searching for answer data. This interface has a user-friendly design, including a search box and a search result display area. As a specific example, a search box is installed in a web browser, and the user can simply enter keywords to display related answer data.

[0835] Query Processing Embodiments

[0836] When a user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data. For example, if a user enters "environmental policy," the server converts the query into an embedding and searches the database to list relevant answer data.

[0837] Display of search results

[0838] The server returns the search results to the terminal and visually displays them to the user, allowing the user to easily refer to and utilize related answer data. For example, related answer data is displayed in a list format on the web, and an interface is provided where the user can click on details to view the full text.

[0839] Specific examples

[0840] For example, if a drugstore staff member is asked by a customer about "side effects of allergy medication," they can use this system to quickly search and refer to the information. The staff member simply types "side effects of allergy medication" into the tablet or smartphone interface, and the system instantly displays relevant information.

[0841] Prompt Sentence Examples

[0842] "Please tell me about the side effects of allergy medication."

[0843] In this way, a system is created in which complex information processing is seamlessly performed on the back end, allowing store staff to quickly and accurately answer customer questions.

[0844] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0845] Step 1:

[0846] The server accesses the official websites of each local government and automatically collects publicly available response data. Specifically, a web crawler running on the server analyzes the HTML structure and extracts PDF links to the response data. It receives a list of website URLs as input and downloads and saves the PDF data as output.

[0847] Step 2:

[0848] The server converts the collected PDF files into text data. Specifically, it uses the pdfminer library to extract text from the PDF, and then cleans it by removing unnecessary line breaks and special characters. It takes PDF files as input and generates clean text data as output.

[0849] Step 3:

[0850] The server uses a generative AI model (e.g., SentenceTransformer) to generate embeddings from the cleaned text data. This process involves calculations to convert the text data into a numerical vector. It receives the clean text data as input and generates embeddings (numerical vectors) as output.

[0851] Step 4:

[0852] The server registers the generated embedding in a database and updates the index. Specifically, it uses a data management system such as SQLite to insert the newly generated embedding and the original text data into the database and update the search index. It receives the generated embedding and the original text data as input and obtains an updated database as output.

[0853] Step 5:

[0854] The terminal provides a user with an interface for searching answer data. The interface displays a search box where the user can enter a search query and a results display area. The terminal receives a search query from the user as input and prepares an output to process the query.

[0855] Step 6:

[0856] The server converts the user's search query into an embedding and searches the database. Specifically, the query is input into a generative AI model to generate an embedding. It then compares it with existing embeddings in the database to extract the most relevant answer data. It receives the user's search query as input and obtains search results as output.

[0857] Step 7:

[0858] The server returns the search results to the terminal and displays them visually to the user. For example, the server may display the relevant answer data in a list format, and provide an interface that allows the user to click on details to view the full text. The server receives the search results as input and generates the data to be displayed to the user as output.

[0859] 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.

[0860] The present invention provides a system that automatically collects parliamentary response data made public by local governments and can efficiently search it using embeddings generated using a generative AI model, as well as a system that combines an emotion engine that recognizes the user's emotions.

[0861] Data collection implementation

[0862] The server accesses the official websites of each local government and crawls the publicly available data on parliamentary responses. The server analyzes the HTML structure and extracts links to the response data based on specific HTML tags and classes. The server then follows the links and automatically downloads PDF files and other data.

[0863] A specific example is a process for collecting PDF files related to parliamentary responses from the websites of multiple local governments in bulk.

[0864] Data Preprocessing Embodiments

[0865] The server converts the collected PDF files into text format. The server uses a PDF-to-text conversion library to extract the text from each page and compile it into a single text file. The server then performs data cleaning to remove unnecessary line breaks, special characters, and unnecessary header and footer information.

[0866] As a concrete example, it removes unnecessary information from text data extracted from a PDF file and generates text data that is formatted to be easy to read.

[0867] An embodiment of embedding generation

[0868] The server uses a pre-trained generative AI model to generate embeddings from the cleaned text data, which convert the text data into numerical vectors that can be used for efficient search and comparison.

[0869] Specifically, the text of parliamentary responses is input into a generative AI model, which outputs a fixed-length vector that is later stored in a database and used as a search index.

[0870] Database Update Implementation Example

[0871] The server registers the generated embedding in its database and updates its search index, allowing it to respond quickly when a user enters a search query.

[0872] The specific operation involves adding the newly generated embeddings to the database and rebuilding the index along with the existing data.

[0873] User Interface Embodiments

[0874] The terminal provides the user with an interface for searching answer data. The user interface includes a search box and a search result display area, and is designed to be easy for users to operate.

[0875] A specific example is an interface that provides a search box in a web browser, and when a user enters a keyword, related answer data is displayed.

[0876] Query Processing Embodiments

[0877] A user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data.

[0878] As a concrete example, when a user inputs "environmental policy," the query is converted into an embedding, a database search is performed, and relevant response data is extracted.

[0879] Embodiment of search result display

[0880] The server returns the search results to the terminal and displays them visually to the user, who can then easily refer to and manipulate the response data of other local governments.

[0881] As a concrete example, related response data is displayed in a list format on the web, and users can click on each item to view details.

[0882] Embodiment of Emotion Engine

[0883] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when entering a search query. If the user is in a positive emotion, this emotion engine prioritizes the display of highly relevant search results, and if the user is in a negative emotion, it displays appropriate results and messages to reduce stress.

[0884] A specific example is an interface in which an emotion engine analyzes a user's facial expressions and vocal tone, and if they are judged to be negative, provides relaxing information.

[0885] By combining the above methods, the present invention provides a system that efficiently collects response data from local governments and allows users to easily search and refer to it. Furthermore, by combining it with an emotion engine, it is possible to provide optimal search results according to the user's emotional state, improving the user experience.

[0886] The processing flow will be explained below.

[0887] Step 1:

[0888] The server accesses the official websites of each local government and crawls the data on parliamentary responses. Based on the URL list, the server analyzes the HTML structure of each website, extracts links to the response data from elements that match specific HTML tags and classes, and follows the links to download PDF files.

[0889] Step 2:

[0890] The server converts the collected PDF files into text format. The server uses a dedicated library for extracting text from PDFs, extracts the text for each page, and saves it as a single text file.

[0891] Step 3:

[0892] The server performs data cleaning on the text data. Specifically, it removes unnecessary line breaks, special characters, and unnecessary header and footer information to generate formatted text data. This prepares the data in a consistent, parseable format.

[0893] Step 4:

[0894] The server uses a generative AI model to generate embeddings from the cleaned text data. The server inputs the text data into a pre-trained language model and outputs a fixed-length numeric vector (embedding). This embedding is a numeric representation of the text's meaning.

[0895] Step 5:

[0896] The server registers the generated embeddings in the database and updates the index. The server adds new embeddings to the database and rebuilds existing data indexes to improve search efficiency.

[0897] Step 6:

[0898] The terminal provides the user with an interface for searching answer data. Specifically, a user-friendly interface is provided with a search box and a search result display area, allowing the user to easily operate it.

[0899] Step 7:

[0900] The server receives the user's search query, converts the query into an embedding, and searches the database. The user-entered search query is fed into a generative AI model, which compares the converted embedding with existing embeddings in the database.

[0901] Step 8:

[0902] The server sends the search results to the device and displays them visually to the user. The search results are sent to the device in JSON format, and the device formats the received data and displays it to the user in a list format.

[0903] Step 9:

[0904] The server is equipped with an emotion engine that recognizes the user's emotions and optimizes search results according to the user's emotional state. When a user enters a search query, the emotion engine analyzes the user's facial expressions and tone of voice, and prioritizes relevant search results if the user's emotions are positive, and displays appropriate results and messages to reduce stress if the user's emotions are negative.

[0905] Step 10:

[0906] The server reflects the results of the emotion engine and adjusts the ranking of search results to provide an optimal user experience. By displaying search results with the display order adjusted based on the emotion engine, users can obtain information efficiently and without stress.

[0907] Example 2

[0908] 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."

[0909] Although data on responses from local governments is made public by each organization, there is no way to efficiently collect it and organize it in a searchable format. Furthermore, there is no system in existence that uses an interface that takes user emotions into consideration to efficiently search for the necessary response data while reducing stress. This forces users to manually search for the information they need from large amounts of data, which is a laborious and time-consuming process.

[0910] 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.

[0911] In this invention, the server includes: means for automatically collecting response data published by each local government; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for displaying search results to the user; and means including an emotion engine for recognizing the user's emotional state and displaying search results and appropriate messages. This allows efficient collection of response data from local governments, allows users to easily search and reference it, and provides optimal search results according to the user's emotional state.

[0912] "Response data" refers to data that records the answers and explanations given by administrative representatives to questions from assembly members in local government assemblies.

[0913] A "text format" is a form of digital data that is recorded as human-readable character information.

[0914] "Data cleaning" is the process of removing unnecessary information and errors from collected data and formatting it.

[0915] A "generative AI model" is an artificial intelligence model that has been pre-trained with large amounts of data and is capable of generating appropriate outputs for specific inputs.

[0916] An "embedding" is a conversion of text data into a numerical vector, which is used for efficient searching and comparison.

[0917] A "database" is an information system that efficiently stores and manages large amounts of data and allows quick access when needed.

[0918] A "search index" is an index used to efficiently search for information in a database.

[0919] An "interface" is a point of contact or means for exchanging information between a user and a system, and includes, for example, a search box or a display screen.

[0920] A "search query" refers to an inquiry that a user enters into a search system, and the system searches for related information based on the query.

[0921] The "emotion engine" is a system component that has the function of recognizing the user's emotional state and providing appropriate information and messages according to that state.

[0922] This invention is a system that automatically collects parliamentary response data published by local governments and efficiently searches it using embeddings generated by a generative AI model. Furthermore, we aim to improve the search experience by combining it with an emotion engine that recognizes the user's emotions.

[0923] Data collection

[0924] The server accesses the official websites of each local government and crawls the publicly available data on assembly responses. Specifically, it analyzes the HTML structure, extracts links to the response data based on specific HTML tags and classes, and automatically downloads PDF files by following those links. For example, the server accesses the Tokyo Metropolitan Assembly website, extracts links to PDF files containing keywords including "response," and downloads them.

[0925] Data Preprocessing

[0926] The server converts the collected PDF files into text format. Using a library for converting PDF to text format (e.g., PyPDF2), the server extracts text from each page and compiles it into a single text file. It then performs data cleaning, removing unnecessary line breaks, special characters, and unnecessary header and footer information. Specifically, the server uses PyPDF2 to extract text from the downloaded PDF files, removes unnecessary information, and creates a unified text document.

[0927] embedding generation

[0928] The server uses a pre-trained generative AI model (e.g., a pre-trained language model) to generate embeddings from the cleaned text data. These embeddings convert the text data into numerical vectors that can be used for efficient search and comparison. For example, the server inputs the cleaned text data into a language model, which outputs a fixed-length vector. This vector is stored in a database and serves as a search index.

[0929] Database Update

[0930] The server registers the generated embedding in the database and updates the search index. Specifically, the server adds the newly generated embedding to the existing database and rebuilds the search index, allowing for quick responses when a user enters a search query.

[0931] User Interface

[0932] The terminal provides the user with an interface for searching for answer data. The interface is equipped with a search box and a search result display area, and is designed to be easy for users to operate. For example, if a search box is provided in a web browser and the user enters a keyword, related answer data will be displayed.

[0933] Query processing and search result display

[0934] A user enters a search query, and the server converts the query into an embedding and searches the database. The server compares the generated embedding with existing embeddings in the database and extracts the most relevant answer data. For example, if a user enters "environmental policy," the server converts the query into an embedding and searches the database to extract relevant answer data. The server returns the results to the device and displays them visually to the user.

[0935] Emotion engine integration

[0936] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when they enter a search query. This emotion engine prioritizes relevant search results if the emotion is positive, and displays appropriate results and messages to reduce stress if the emotion is negative. For example, the emotion engine analyzes the user's facial expressions and tone of voice, and if it determines that the emotion is negative, it provides an interface with information to help users relax.

[0937] This system efficiently collects response data from local governments and provides an environment where users can easily search and refer to it. Furthermore, by combining it with an emotion engine, it is possible to provide optimal search results according to the user's emotional state, improving the user experience.

[0938] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0939] Step 1: Data collection

[0940] The server accesses the official websites of each local government and crawls the publicly available data on parliamentary responses. The input is the URL of the local government's website, and the output is a link to a PDF file of the parliamentary response data. Specifically, the server analyzes the HTML structure of the webpage, extracts links to the response data based on specific HTML tags and classes, and automatically downloads the PDF file by following the link.

[0941] Step 2: Data Preprocessing

[0942] The server converts the collected PDF files into text format. The input is a PDF file, and the output is text data. Specifically, a library for converting PDF to text (e.g., PyPDF2) is used to extract the text from each page and combine it into a single text file. The data is then cleaned to remove unnecessary line breaks, special characters, and unnecessary header and footer information.

[0943] Step 3: Embedding generation

[0944] The server generates embeddings from the cleaned text data. The input is the cleaned text data, and the output is the embedding (a numerical vector). Specifically, it uses a pre-trained generative AI model (e.g., a pre-trained language model) to convert the text data into a numerical vector, which makes the text data in a format that can be efficiently searched and compared.

[0945] Step 4: Update the database

[0946] The server registers the generated embedding in the database and updates the search index. The input is the generated embedding, and the output is the updated search index. The specific operation is to add the newly generated embedding to the existing database and rebuild the search index. This allows for quick response when a user enters a search query.

[0947] Step 5: User Interface

[0948] The terminal provides the user with an interface for searching for answer data. The input is the user's operation, and the output is the interface that displays the search results. Specifically, the operation is to provide a search box and a search result display area in the web browser, allowing the user to enter keywords. Related answer data is displayed based on the entered keywords.

[0949] Step 6: Query Processing

[0950] The server receives a search query entered by a user, converts the query into an embedding, and searches a database. The input is the user's search query (text), and the output is related answer data. Specifically, the server inputs the user's search query into a generative AI model, searches the database using the resulting embedding, and extracts the most relevant answer data.

[0951] Step 7: Displaying search results

[0952] The server returns the search results to the terminal and displays them visually to the user. The input is the answer data extracted by the search, and the output is the displayed search results. Specifically, the server finds the relevant answer data and sends it to the terminal, and the terminal displays the search results in list format on the web browser. The user can click on each item to view details.

[0953] Step 8: Integrating the Emotion Engine

[0954] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when they enter a search query. The input is the user's facial expression and tone of voice, and the output is appropriate search results and messages. Specifically, the emotion engine analyzes the user's emotional state, and if the emotion is positive, it prioritizes relevant search results, and if the emotion is negative, it provides appropriate messages and information to reduce stress.

[0955] (Application example 2)

[0956] 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."

[0957] Factories require efficient management of maintenance information for multiple pieces of equipment and machinery. However, manual collection and analysis is time-consuming and labor-intensive, often resulting in insufficient data management and analysis. Furthermore, while it is known that workers' emotional states affect work efficiency and safety, there is a lack of systems that address this issue. Therefore, there is a need for a system that can automatically collect and efficiently search maintenance information, as well as a system that can provide feedback based on workers' emotions.

[0958] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: means for automatically collecting response data made public by local governments; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for displaying search results to the user; means for processing maintenance information collected from a machine as text data; and means for providing search results according to the user's emotional state by incorporating an emotion engine for recognizing the user's emotions. This not only enables efficient collection and search of maintenance information, but also enables appropriate feedback to be provided according to the worker's emotional state.

[0959] 1. "Response data" refers to data showing the content of responses given in local government assemblies and other such meetings.

[0960] 2. "Text format" means a data format expressed as human-readable characters or sentences.

[0961] 3. "Data cleaning" is the process of removing redundant information and noise from collected text data.

[0962] 4. A "generative AI model" is an artificial intelligence model that is pre-trained using massive amounts of data.

[0963] 5. "Embedding" is a method of representing text data as a numerical vector.

[0964] 6. A "database" is a system for efficiently storing, managing, and retrieving structured data.

[0965] 7. "Interface" refers to the operation screens and input fields that allow users to interact with the system.

[0966] 8. A "query" is a question or keyword entered by a user to search for specific information.

[0967] 9. "Search query" means input to the system to specify the information to be searched.

[0968] 10. "Machinery" refers to production equipment and devices installed in factories and production lines.

[0969] 11. "Maintenance Information" means data relating to the maintenance and repair of facilities and machinery.

[0970] 12. An "emotion engine" is an algorithm or system for determining a user's emotional state.

[0971] 13. "Feedback" refers to the response or information that a system returns in response to a user's input or actions.

[0972] To implement this invention, a server first accesses the official websites of each local government and automatically collects publicly available response data. Specifically, the server uses web crawling technology to extract and download response data links based on specific HTML tags and classes. For example, there is a process for collecting PDF files from the websites of multiple local government organizations in bulk.

[0973] Next, the server converts the collected PDF files into text format. This is done by extracting the text from each page using a PDF to text conversion library (e.g., PyPDF2) and compiling it into a single text file. After that, data cleaning is performed, removing unnecessary line breaks and special characters, and generating formatted text data. For example, there is a process to remove unnecessary information from the text data extracted from the PDF files.

[0974] The server then uses a generative AI model (e.g., OpenAI GPT-3 / BERT) to generate embeddings from the cleaned text data. These embeddings convert the text data into numerical vectors that can be used for efficient search and comparison. Specifically, the text data is input into the generative AI model, which outputs a fixed-length vector. This vector is later stored in a database and used as a search index.

[0975] The server then registers the generated embeddings in a database (e.g., Elasticsearch) and updates the index, allowing for faster response when a user enters a search query. Specifically, the process involves adding the newly generated embeddings to the database and rebuilding the index with the existing data.

[0976] The server also provides a user interface to allow users to search for the answer data. This interface is installed on a terminal and includes a search box and a search result display area. For example, there is an interface in which a search box is provided in a web browser, and when a user enters a keyword, related answer data is displayed.

[0977] Next, when a user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data. For example, if a user enters "machine wear diagnosis," there is a process to convert the query into an embedding, perform a database search, and extract relevant answer data.

[0978] The server also returns the search results to the terminal and visually displays them to the user. The user can easily refer to and manipulate the related answer data. For example, the related answer data is displayed in list format on the web, and the user can click on each item to view details.

[0979] Finally, the server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when entering a search query. This emotion engine prioritizes relevant search results when the user is in a positive emotion, and displays appropriate results and messages to reduce stress when the user is in a negative emotion. For example, the emotion engine analyzes the user's facial expressions and tone of voice, and if the engine determines that the user is in a negative emotion, it provides an interface that provides information to help them relax.

[0980] Prompt Sentence Examples

[0981] Parliamentary response data analysis. Generate an embedding from the following text:

[0982] "Answer regarding environmental policy..."

[0983] This invention not only makes it possible to efficiently collect and search maintenance information, but also makes it possible to provide appropriate feedback to the worker according to his or her emotional state.

[0984] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0985] Step 1:

[0986] The server accesses the official websites of local government organizations and collects response data. Specifically, the server crawls multiple specified websites using web crawling technology (e.g., BeautifulSoup, Scrapy), extracts links to response data based on specific HTML tags and classes, and downloads PDF files. The input is the website URL, and the output is the downloaded PDF file.

[0987] Step 2:

[0988] The server converts the collected PDF files into text format. Specifically, it uses the PyPDF2 library to extract text from each page of the PDF file and combines it into a single text file. The input is a PDF file and the output is a text file.

[0989] Step 3:

[0990] The server cleans the text files. Specifically, it uses libraries such as NLTK to remove unnecessary line breaks, special characters, and unnecessary header and footer information, generating readable, formatted text data. The input is the raw text file, and the output is the cleaned text data.

[0991] Step 4:

[0992] The server generates embeddings from the cleaned text data using a generative AI model (e.g., OpenAI GPT-3 / BERT). Specifically, the text data is input to the AI ​​model as a prompt sentence, and a fixed-length vector is output. The input is the cleaned text data, and the output is the generated embedding.

[0993] Step 5:

[0994] The server registers the generated embeddings in a database (e.g., Elasticsearch) and updates the index. Specifically, it adds the newly generated embeddings to the database and rebuilds the index with the existing data. The input is the generated embeddings, and the output is the updated database index.

[0995] Step 6:

[0996] The terminal provides the user with an interface for searching answer data via a user interface. Specifically, the terminal displays a web browser interface including a search box and a search result display area. The input is the user interface design specification, and the output is the interface available to the user.

[0997] Step 7:

[0998] When a user enters a search query, the server converts the query into an embedding and searches the database. Specifically, the search query is input into the AI ​​model, and the generated embedding is compared with existing embeddings in the database. The input is the user's search query, and the output is highly relevant answer data.

[0999] Step 8:

[1000] The server returns the search results to the terminal and displays them visually to the user. Specifically, it sends the relevant answer data in a list format to the terminal so that the user can view the details. The input is the search results, and the output is the displayed search results.

[1001] Step 9:

[1002] The server uses an emotion engine to recognize the user's emotional state. Specifically, it analyzes the user's facial expressions and voice tone using the device's camera and microphone. The input is real-time video and audio data, and the output is the user's emotional state.

[1003] Step 10:

[1004] The server provides appropriate search results and messages based on the user's emotional state. Specifically, if the user's emotional state is positive, it prioritizes relevant results, and if the user's emotional state is negative, it displays relaxing information and messages. The input is the user's emotional state, and the output is the adjusted search results and messages.

[1005] 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.

[1006] 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.

[1007] 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.

[1008] [Fourth embodiment]

[1009] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1010] 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.

[1011] 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).

[1012] 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.

[1013] 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.

[1014] 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).

[1015] 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.

[1016] 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.

[1017] 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.

[1018] 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.

[1019] 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.

[1020] 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.

[1021] 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."

[1022] This invention provides a system that collects parliamentary response data made public by local governments, generates embeddings using a generative AI model, and allows users to easily search and reference the data.

[1023] Data collection implementation

[1024] The server accesses the official websites of each local government and automatically crawls the publicly available parliamentary response data. During this process, the web crawler analyzes the HTML structure and extracts only the necessary information from the specified elements. For example, it automatically downloads the PDF link for the response data on each local government website.

[1025] A specific example is a process that collects PDF files related to responses from the websites of multiple local governments and saves them in a specific folder.

[1026] Data Preprocessing Embodiments

[1027] The server converts the collected PDF files into text data. This involves extracting text from the PDF using a dedicated library. Data cleaning is then performed. This cleaning process removes unnecessary line breaks and special characters, resulting in pure text data.

[1028] A specific example is deleting unnecessary header and footer information from text data extracted from a PDF file and formatting it to leave only text.

[1029] An embodiment of embedding generation

[1030] The server uses a pre-trained generative AI model to generate embeddings from the cleaned text data, which convert the text data into numerical vectors for easier future search and comparison.

[1031] Specifically, the text of parliamentary responses is input into a generative AI model, which outputs a fixed-length vector that is later stored in a database and used as a search index.

[1032] Database Update Implementation Example

[1033] The server registers the generated embedding in its database and updates its search index, allowing it to respond quickly to future user search queries.

[1034] A concrete example is the process of adding newly generated embeddings to a database and rebuilding the search index along with the existing data.

[1035] User Interface Embodiments

[1036] The terminal provides the user with an interface for searching the answer data, which has a user-friendly design and includes a search box and a search result display area.

[1037] A specific example is an interface that provides a search box in a web browser, allowing users to simply enter keywords and display related response data.

[1038] Query Processing Embodiments

[1039] A user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data.

[1040] As a concrete example, when a user enters "environmental policy," the process converts the query into an embedding, searches a database, and lists relevant response data.

[1041] Display of search results

[1042] The server returns the search results to the terminal and displays them visually to the user, allowing the user to easily refer to and utilize the response data of other local governments.

[1043] As a specific example, related defense data is displayed in a list format on the Web, and an interface is provided that allows users to click on details to view the full text.

[1044] By combining the above means, the present invention provides a system that efficiently collects response data from local governments and allows users to easily search and refer to it.

[1045] The processing flow will be explained below.

[1046] Step 1:

[1047] The server accesses the list of website URLs collected and crawls response data from the official websites of each local government.

[1048] Specifically, the server analyzes the HTML structure, extracts links to the response data based on specific HTML tags and classes, and follows those links to download PDF files, etc.

[1049] Step 2:

[1050] The server converts the collected PDF files into text format.

[1051] Specifically, it uses a PDF to text conversion library to extract the text from each page and combine it into a single text file.

[1052] Step 3:

[1053] The server performs data cleaning on the text data.

[1054] Specifically, it removes extra line breaks, special characters, and unnecessary header and footer information to ensure data consistency.

[1055] Step 4:

[1056] The server uses a generative AI model to generate embeddings from the cleaned text data.

[1057] Specifically, text data is input into a pre-trained language model and converted into a fixed-length vector.

[1058] Step 5:

[1059] The server registers the generated embedding in the database and updates the index.

[1060] Specifically, it inserts new entries into the database and rebuilds existing indexes to enable efficient searches.

[1061] Step 6:

[1062] The terminal provides a search interface to the user.

[1063] Specifically, it installs a search box and results display area in a web browser or dedicated application, providing an interface that is easy for users to use.

[1064] Step 7:

[1065] The server receives a search query entered by the user, converts the query into an embedding, and searches the database.

[1066] Specifically, the search query is input into the generative AI model, and the generated embedding is compared with embeddings in the database to extract entries with high similarity.

[1067] Step 8:

[1068] The server returns the search results to the terminal and displays them visually to the user.

[1069] Specifically, the search results are sent to the terminal in JSON format, and the terminal formats the data it receives and displays it to the user.

[1070] Example 1

[1071] 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."

[1072] The data on parliamentary responses made public by local governments is scattered across various websites, making it difficult to centrally collect and easily search and reference. Furthermore, the volume of information is enormous, so manually collecting and organizing the data takes time and effort. Furthermore, variations in linguistic expression and clerical errors make data cleaning difficult, reducing the efficiency of information searches. There is a need to resolve these issues and enable users to efficiently access response data.

[1073] 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.

[1074] In this invention, the server includes means for automatically collecting response data published by each local government, means for converting the collected data into text format and performing data cleaning, means for generating embeddings from the cleaned text data using a generative AI model, and means for registering the generated embeddings in a database and updating an index, thereby enabling efficient collection, processing, and search of response data from local governments.

[1075] "Response data" refers to official statements and responses recorded in local government assemblies and committees.

[1076] "Automatic collection means" refers to a system that uses web crawlers or scraping technology to automatically obtain specified information from specific websites.

[1077] "Text conversion method" refers to a technology that extracts text information from collected PDFs, image files, etc., and converts it into a digital text format.

[1078] "Data cleaning means" is a process of formatting collected text data by removing typos, unnecessary special characters, line breaks, etc.

[1079] A "generative AI model" is an artificial intelligence model that is trained in advance with large amounts of data and exhibits high performance for specific tasks.

[1080] "Embedding generation method" is a technology that converts text data into numerical vectors and expresses their characteristics.

[1081] The "database registration method" is the process of saving the generated embedding to a database and updating the search index.

[1082] A "search query" refers to a search term or phrase that a user enters to find desired information.

[1083] A "user interface" is an interface that includes the screens and functions that a user operates when using a system.

[1084] An "index" refers to a structure or mechanism for efficiently searching data in a database.

[1085] The present invention is a system that automatically collects assembly response data published by local governments, generates embeddings using a generative AI model, and allows users to easily search and reference them. Specific embodiments for implementing the present invention are described below.

[1086] The server accesses the official websites of each local government and automatically collects publicly available parliamentary response data. This data collection is performed using web crawlers and scraping techniques. Specifically, it uses HTML analysis libraries such as BeautifulSoup to analyze the HTML structure of each website and obtain links to response data and PDF files.

[1087] Next, the server converts the collected PDF files into text data. To do this, it uses a library called PyMuPDF to extract text from the PDF. Then, it uses a data cleaning method to remove unnecessary line breaks and special characters from the extracted text data and format it. This results in clean text data.

[1088] Embeddings are generated from the cleaned text data using a generative AI model. This generative AI model uses a pre-trained language model such as GPT-3. Using the text data as input, the model generates a numerical vector called an embedding. This allows the features of the text data to be represented as a numerical vector.

[1089] The generated embeddings are registered in a database by the server, and as part of this registration process, the server also updates a search index, allowing the data in the database to be efficiently searched.

[1090] The user interface is provided using a terminal. The terminal provides an interface with a search box and a result display area so that the user can search for answer data. For example, the interface is implemented using the React.js framework in a web browser.

[1091] A user inputs a search query through this interface. The search query is converted into an embedding by the server. The converted embedding is used to search the database and extract the most relevant response data. For example, if a user inputs the query "environmental policy," the query is converted into an embedding and the relevant response data in the database is listed.

[1092] Finally, the server returns the search results to the terminal and visually displays them to the user, allowing the user to easily browse related answer data and view detailed information.

[1093] A specific example of a prompt sentence could be, "Search for questions about education policy in local councils." By inputting this prompt sentence into a generative AI model, generating an embedding, and searching the database, relevant answer data can be easily obtained.

[1094] The above is a specific embodiment for carrying out the present invention, which realizes a system that can efficiently collect, search, and use assembly response data from local public organizations.

[1095] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1096] Step 1:

[1097] The server accesses the official websites of each local government and automatically collects data on parliamentary responses.

[1098] Specifically, the server uses a web crawler and an HTML parsing library such as BeautifulSoup to analyze the HTML structure of the web page.

[1099] The input is the URL of the official website of each local government, and the output is a link to the response data or a PDF file.

[1100] During this process, the server searches for the specified keywords and file formats and retrieves links containing the response data. For example, the server accesses "https: / / example-localgovsite.jp / kaigi" and retrieves PDF links to the parliamentary response data.

[1101] Step 2:

[1102] The server converts the collected PDF files into text data.

[1103] For this, we use a PDF parsing library called PyMuPDF.

[1104] The input is the PDF files collected in step 1, and the output is text data.

[1105] Specifically, the server opens the PDF file and extracts its contents in text format. For example, it uses PyMuPDF to extract the text from "example.pdf" and saves it as text data.

[1106] Step 3:

[1107] The server performs data cleaning on the text data.

[1108] The input is the text data extracted in step 2, and the output is the cleaned text data.

[1109] Specifically, the server deletes unnecessary line breaks and special characters from the text data and arranges the data. For example, the server deletes line breaks and special characters to obtain text data that has been arranged in a single line.

[1110] Step 4:

[1111] The server uses a generative AI model to generate embeddings from the cleaned text data.

[1112] The input is the text data cleaned in step 3, and the output is the embedding (a numeric vector).

[1113] Specifically, the server inputs the cleaned text data into a generative AI model to obtain embeddings. For example, the server inputs the text data into a GPT-3 model and obtains the generated numeric vectors.

[1114] Step 5:

[1115] The server registers the generated embedding in the database and updates the index.

[1116] The input is the embedding generated in step 4, and the output is the updated database and index.

[1117] Specifically, the server adds the embedding to the database and rebuilds the search index, e.g., registers the embedding in the PostgreSQL database and updates the index.

[1118] Step 6:

[1119] The terminal provides the user with an interface for searching for answer data.

[1120] There is no input, and the output is an interface with a search box and a results display area.

[1121] Specifically, the device displays an interface using React.js on a web browser, allowing users to enter search queries. For example, it provides an interface that displays a search box and a search button.

[1122] Step 7:

[1123] A user enters a search query into the interface, and the server converts the query into an embedding and searches the database.

[1124] The input is a user's search query and the output is a list of relevant answer data.

[1125] Specifically, the server passes the input search query to the generative AI model, generates an embedding, and then searches the database. For example, input the query "environmental policy" and list related response data.

[1126] Step 8:

[1127] The server returns the search results to the terminal and displays them visually to the user.

[1128] The input is the list of search results obtained in step 7, and the output is a visual display that the user can refer to.

[1129] Specifically, the server sends the search results in JSON format to the terminal, and the terminal displays the results in a list format, for example, by displaying a list of related answer data on a web interface so that the user can check the details.

[1130] The above are the specific program processing steps of this system.

[1131] (Application example 1)

[1132] 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."

[1133] In today's brick-and-mortar stores, store employees are required to quickly and accurately provide product information and FAQs in response to customer inquiries. However, with existing systems, this process is often manual and inefficient, which can lead to a decline in the quality of customer service. Furthermore, collecting necessary information from various data sources, integrating it, and making it searchable takes time and effort, increasing the burden on employees. It is necessary to solve these problems and improve customer service and employee work efficiency.

[1134] 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.

[1135] In this invention, the server includes: means for automatically collecting response data published by each local government; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for allowing store employees to quickly search for product information and FAQs in response to customer inquiries; and means for displaying search results to the user. This enables employees to respond to customer inquiries quickly and accurately based on integrated data collected from various information sources. Furthermore, this significantly improves employee work efficiency and contributes to increased customer satisfaction.

[1136] A "local public entity" is a local government that handles administrative affairs as a local government separate from the nation.

[1137] "Response data" refers to data that records the content of questions and answers that were asked in parliaments or conferences.

[1138] "Automatic collection means" refers to a mechanism by which a system or program collects data automatically, without human intervention.

[1139] "Means of converting to text format" refers to the process of converting data such as images and PDFs into a format that can be handled as text information.

[1140] "Data cleaning" is the process of removing unnecessary information and errors from collected data, making it easier to analyze.

[1141] A "generative AI model" is an artificial intelligence system that has been pre-trained using machine learning or deep learning to perform a specific task.

[1142] An "embedding" is a representation of text data converted into a numerical vector that is used when processing it with a machine learning model.

[1143] A "database" is an information management system that systematically stores specific information and allows it to be searched and used when needed.

[1144] "Means of updating the index" refers to the process of updating the reference system to make searches more efficient for newly added data.

[1145] An "interface" is the hardware and software structure through which a user interacts with a computer system.

[1146] A "search query" is a keyword or phrase that a user enters into a search system to retrieve specific information.

[1147] "Customer inquiries" are questions or clarifications made by store patrons regarding products or services.

[1148] "Product information" is a detailed description of the product's specifications, price, usage, etc.

[1149] A "FAQ" is a collection of frequently asked questions and their answers.

[1150] The "means for displaying search results" is a method for visually conveying to the user the data retrieved based on the query entered by the user.

[1151] MODE FOR CARRYING OUT THE INVENTION

[1152] This invention provides a system that collects response data from local governments based on a series of procedures and enables quick and efficient search using a generative AI model. Specific embodiments of this system are described below.

[1153] Data collection implementation

[1154] The server accesses the official websites of each local government and automatically collects publicly available response data. During this process, a web crawler running on the server analyzes the HTML structure and extracts only the necessary information from specified elements. For example, it automatically downloads PDF links to response data on each local government's website. The software used is requests and BeautifulSoup.

[1155] Data Preprocessing Embodiments

[1156] The server converts the collected PDF files into text data. To do this, it uses a dedicated library called pdfminer to extract text from the PDF. Data cleaning is then performed to remove unnecessary line breaks and special characters and format the text data.

[1157] An embodiment of embedding generation

[1158] The server uses a pre-trained generative AI model (e.g., SentenceTransformer) to generate embeddings from the cleaned text data, which convert the text data into numerical vectors for easier later search and comparison.

[1159] Database Update Implementation Example

[1160] The server registers the generated embedding in a database and updates the index, allowing for quick responses when users later enter search queries. The database uses a data management system such as SQLite.

[1161] User Interface Embodiments

[1162] The terminal provides the user with an interface for searching for answer data. This interface has a user-friendly design, including a search box and a search result display area. As a specific example, a search box is installed in a web browser, and the user can simply enter keywords to display related answer data.

[1163] Query Processing Embodiments

[1164] When a user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data. For example, if a user enters "environmental policy," the server converts the query into an embedding and searches the database to list relevant answer data.

[1165] Display of search results

[1166] The server returns the search results to the terminal and visually displays them to the user, allowing the user to easily refer to and utilize related answer data. For example, related answer data is displayed in a list format on the web, and an interface is provided where the user can click on details to view the full text.

[1167] Specific examples

[1168] For example, if a drugstore staff member is asked by a customer about "side effects of allergy medication," they can use this system to quickly search and refer to the information. The staff member simply types "side effects of allergy medication" into the tablet or smartphone interface, and the system instantly displays relevant information.

[1169] Prompt Sentence Examples

[1170] "Please tell me about the side effects of allergy medication."

[1171] In this way, a system is created in which complex information processing is seamlessly performed on the back end, allowing store staff to quickly and accurately answer customer questions.

[1172] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1173] Step 1:

[1174] The server accesses the official websites of each local government and automatically collects publicly available response data. Specifically, a web crawler running on the server analyzes the HTML structure and extracts PDF links to the response data. It receives a list of website URLs as input and downloads and saves the PDF data as output.

[1175] Step 2:

[1176] The server converts the collected PDF files into text data. Specifically, it uses the pdfminer library to extract text from the PDF, and then cleans it by removing unnecessary line breaks and special characters. It takes PDF files as input and generates clean text data as output.

[1177] Step 3:

[1178] The server uses a generative AI model (e.g., SentenceTransformer) to generate embeddings from the cleaned text data. This process involves calculations to convert the text data into a numerical vector. It receives the clean text data as input and generates embeddings (numerical vectors) as output.

[1179] Step 4:

[1180] The server registers the generated embedding in a database and updates the index. Specifically, it uses a data management system such as SQLite to insert the newly generated embedding and the original text data into the database and update the search index. It receives the generated embedding and the original text data as input and obtains an updated database as output.

[1181] Step 5:

[1182] The terminal provides a user with an interface for searching answer data. The interface displays a search box where the user can enter a search query and a results display area. The terminal receives a search query from the user as input and prepares an output to process the query.

[1183] Step 6:

[1184] The server converts the user's search query into an embedding and searches the database. Specifically, the query is input into a generative AI model to generate an embedding. It then compares it with existing embeddings in the database to extract the most relevant answer data. It receives the user's search query as input and obtains search results as output.

[1185] Step 7:

[1186] The server returns the search results to the terminal and displays them visually to the user. For example, the server may display the relevant answer data in a list format, and provide an interface that allows the user to click on details to view the full text. The server receives the search results as input and generates the data to be displayed to the user as output.

[1187] 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.

[1188] The present invention provides a system that automatically collects parliamentary response data made public by local governments and can efficiently search it using embeddings generated using a generative AI model, as well as a system that combines an emotion engine that recognizes the user's emotions.

[1189] Data collection implementation

[1190] The server accesses the official websites of each local government and crawls the publicly available data on parliamentary responses. The server analyzes the HTML structure and extracts links to the response data based on specific HTML tags and classes. The server then follows the links and automatically downloads PDF files and other data.

[1191] A specific example is a process for collecting PDF files related to parliamentary responses from the websites of multiple local governments in bulk.

[1192] Data Preprocessing Embodiments

[1193] The server converts the collected PDF files into text format. The server uses a PDF-to-text conversion library to extract the text from each page and compile it into a single text file. The server then performs data cleaning to remove unnecessary line breaks, special characters, and unnecessary header and footer information.

[1194] As a concrete example, it removes unnecessary information from text data extracted from a PDF file and generates text data that is formatted to be easy to read.

[1195] An embodiment of embedding generation

[1196] The server uses a pre-trained generative AI model to generate embeddings from the cleaned text data, which convert the text data into numerical vectors that can be used for efficient search and comparison.

[1197] Specifically, the text of parliamentary responses is input into a generative AI model, which outputs a fixed-length vector that is later stored in a database and used as a search index.

[1198] Database Update Implementation Example

[1199] The server registers the generated embedding in its database and updates its search index, allowing it to respond quickly when a user enters a search query.

[1200] The specific operation involves adding the newly generated embeddings to the database and rebuilding the index along with the existing data.

[1201] User Interface Embodiments

[1202] The terminal provides the user with an interface for searching answer data. The user interface includes a search box and a search result display area, and is designed to be easy for users to operate.

[1203] A specific example is an interface that provides a search box in a web browser, and when a user enters a keyword, related answer data is displayed.

[1204] Query Processing Embodiments

[1205] A user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data.

[1206] As a concrete example, when a user inputs "environmental policy," the query is converted into an embedding, a database search is performed, and relevant response data is extracted.

[1207] Embodiment of search result display

[1208] The server returns the search results to the terminal and displays them visually to the user, who can then easily refer to and manipulate the response data of other local governments.

[1209] As a concrete example, related response data is displayed in a list format on the web, and users can click on each item to view details.

[1210] Embodiment of Emotion Engine

[1211] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when entering a search query. If the user is in a positive emotion, this emotion engine prioritizes the display of highly relevant search results, and if the user is in a negative emotion, it displays appropriate results and messages to reduce stress.

[1212] A specific example is an interface in which an emotion engine analyzes a user's facial expressions and vocal tone, and if they are judged to be negative, provides relaxing information.

[1213] By combining the above methods, the present invention provides a system that efficiently collects response data from local governments and allows users to easily search and refer to it. Furthermore, by combining it with an emotion engine, it is possible to provide optimal search results according to the user's emotional state, improving the user experience.

[1214] The processing flow will be explained below.

[1215] Step 1:

[1216] The server accesses the official websites of each local government and crawls the data on parliamentary responses. Based on the URL list, the server analyzes the HTML structure of each website, extracts links to the response data from elements that match specific HTML tags and classes, and follows the links to download PDF files.

[1217] Step 2:

[1218] The server converts the collected PDF files into text format. The server uses a dedicated library for extracting text from PDFs, extracts the text for each page, and saves it as a single text file.

[1219] Step 3:

[1220] The server performs data cleaning on the text data. Specifically, it removes unnecessary line breaks, special characters, and unnecessary header and footer information to generate formatted text data. This prepares the data in a consistent, parseable format.

[1221] Step 4:

[1222] The server uses a generative AI model to generate embeddings from the cleaned text data. The server inputs the text data into a pre-trained language model and outputs a fixed-length numeric vector (embedding). This embedding is a numeric representation of the text's meaning.

[1223] Step 5:

[1224] The server registers the generated embeddings in the database and updates the index. The server adds new embeddings to the database and rebuilds existing data indexes to improve search efficiency.

[1225] Step 6:

[1226] The terminal provides the user with an interface for searching answer data. Specifically, a user-friendly interface is provided with a search box and a search result display area, allowing the user to easily operate it.

[1227] Step 7:

[1228] The server receives the user's search query, converts the query into an embedding, and searches the database. The user-entered search query is fed into a generative AI model, which compares the converted embedding with existing embeddings in the database.

[1229] Step 8:

[1230] The server sends the search results to the device and displays them visually to the user. The search results are sent to the device in JSON format, and the device formats the received data and displays it to the user in a list format.

[1231] Step 9:

[1232] The server is equipped with an emotion engine that recognizes the user's emotions and optimizes search results according to the user's emotional state. When a user enters a search query, the emotion engine analyzes the user's facial expressions and tone of voice, and prioritizes relevant search results if the user's emotions are positive, and displays appropriate results and messages to reduce stress if the user's emotions are negative.

[1233] Step 10:

[1234] The server reflects the results of the emotion engine and adjusts the ranking of search results to provide an optimal user experience. By displaying search results with the display order adjusted based on the emotion engine, users can obtain information efficiently and without stress.

[1235] Example 2

[1236] 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."

[1237] Although data on responses from local governments is made public by each organization, there is no way to efficiently collect it and organize it in a searchable format. Furthermore, there is no system in existence that uses an interface that takes user emotions into consideration to efficiently search for the necessary response data while reducing stress. This forces users to manually search for the information they need from large amounts of data, which is a laborious and time-consuming process.

[1238] 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.

[1239] In this invention, the server includes: means for automatically collecting response data published by each local government; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for displaying search results to the user; and means including an emotion engine for recognizing the user's emotional state and displaying search results and appropriate messages. This allows efficient collection of response data from local governments, allows users to easily search and reference it, and provides optimal search results according to the user's emotional state.

[1240] "Response data" refers to data that records the answers and explanations given by administrative representatives to questions from assembly members in local government assemblies.

[1241] A "text format" is a form of digital data that is recorded as human-readable character information.

[1242] "Data cleaning" is the process of removing unnecessary information and errors from collected data and formatting it.

[1243] A "generative AI model" is an artificial intelligence model that has been pre-trained with large amounts of data and is capable of generating appropriate outputs for specific inputs.

[1244] An "embedding" is a conversion of text data into a numerical vector, which is used for efficient searching and comparison.

[1245] A "database" is an information system that efficiently stores and manages large amounts of data and allows quick access when needed.

[1246] A "search index" is an index used to efficiently search for information in a database.

[1247] An "interface" is a point of contact or means for exchanging information between a user and a system, and includes, for example, a search box or a display screen.

[1248] A "search query" refers to an inquiry that a user enters into a search system, and the system searches for related information based on the query.

[1249] The "emotion engine" is a system component that has the function of recognizing the user's emotional state and providing appropriate information and messages according to that state.

[1250] This invention is a system that automatically collects parliamentary response data published by local governments and efficiently searches it using embeddings generated by a generative AI model. Furthermore, we aim to improve the search experience by combining it with an emotion engine that recognizes the user's emotions.

[1251] Data collection

[1252] The server accesses the official websites of each local government and crawls the publicly available data on assembly responses. Specifically, it analyzes the HTML structure, extracts links to the response data based on specific HTML tags and classes, and automatically downloads PDF files by following those links. For example, the server accesses the Tokyo Metropolitan Assembly website, extracts links to PDF files containing keywords including "response," and downloads them.

[1253] Data Preprocessing

[1254] The server converts the collected PDF files into text format. Using a library for converting PDF to text format (e.g., PyPDF2), the server extracts text from each page and compiles it into a single text file. It then performs data cleaning, removing unnecessary line breaks, special characters, and unnecessary header and footer information. Specifically, the server uses PyPDF2 to extract text from the downloaded PDF files, removes unnecessary information, and creates a unified text document.

[1255] embedding generation

[1256] The server uses a pre-trained generative AI model (e.g., a pre-trained language model) to generate embeddings from the cleaned text data. These embeddings convert the text data into numerical vectors that can be used for efficient search and comparison. For example, the server inputs the cleaned text data into a language model, which outputs a fixed-length vector. This vector is stored in a database and serves as a search index.

[1257] Database Update

[1258] The server registers the generated embedding in the database and updates the search index. Specifically, the server adds the newly generated embedding to the existing database and rebuilds the search index, allowing for quick responses when a user enters a search query.

[1259] User Interface

[1260] The terminal provides the user with an interface for searching for answer data. The interface is equipped with a search box and a search result display area, and is designed to be easy for users to operate. For example, if a search box is provided in a web browser and the user enters a keyword, related answer data will be displayed.

[1261] Query processing and search result display

[1262] A user enters a search query, and the server converts the query into an embedding and searches the database. The server compares the generated embedding with existing embeddings in the database and extracts the most relevant answer data. For example, if a user enters "environmental policy," the server converts the query into an embedding and searches the database to extract relevant answer data. The server returns the results to the device and displays them visually to the user.

[1263] Emotion engine integration

[1264] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when they enter a search query. This emotion engine prioritizes relevant search results if the emotion is positive, and displays appropriate results and messages to reduce stress if the emotion is negative. For example, the emotion engine analyzes the user's facial expressions and tone of voice, and if it determines that the emotion is negative, it provides an interface with information to help users relax.

[1265] This system efficiently collects response data from local governments and provides an environment where users can easily search and refer to it. Furthermore, by combining it with an emotion engine, it is possible to provide optimal search results according to the user's emotional state, improving the user experience.

[1266] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1267] Step 1: Data collection

[1268] The server accesses the official websites of each local government and crawls the publicly available data on parliamentary responses. The input is the URL of the local government's website, and the output is a link to a PDF file of the parliamentary response data. Specifically, the server analyzes the HTML structure of the webpage, extracts links to the response data based on specific HTML tags and classes, and automatically downloads the PDF file by following the link.

[1269] Step 2: Data Preprocessing

[1270] The server converts the collected PDF files into text format. The input is a PDF file, and the output is text data. Specifically, a library for converting PDF to text (e.g., PyPDF2) is used to extract the text from each page and combine it into a single text file. The data is then cleaned to remove unnecessary line breaks, special characters, and unnecessary header and footer information.

[1271] Step 3: Embedding generation

[1272] The server generates embeddings from the cleaned text data. The input is the cleaned text data, and the output is the embedding (a numerical vector). Specifically, it uses a pre-trained generative AI model (e.g., a pre-trained language model) to convert the text data into a numerical vector, which makes the text data in a format that can be efficiently searched and compared.

[1273] Step 4: Update the database

[1274] The server registers the generated embedding in the database and updates the search index. The input is the generated embedding, and the output is the updated search index. The specific operation is to add the newly generated embedding to the existing database and rebuild the search index. This allows for quick response when a user enters a search query.

[1275] Step 5: User Interface

[1276] The terminal provides the user with an interface for searching for answer data. The input is the user's operation, and the output is the interface that displays the search results. Specifically, the operation is to provide a search box and a search result display area in the web browser, allowing the user to enter keywords. Related answer data is displayed based on the entered keywords.

[1277] Step 6: Query Processing

[1278] The server receives a search query entered by a user, converts the query into an embedding, and searches a database. The input is the user's search query (text), and the output is related answer data. Specifically, the server inputs the user's search query into a generative AI model, searches the database using the resulting embedding, and extracts the most relevant answer data.

[1279] Step 7: Displaying search results

[1280] The server returns the search results to the terminal and displays them visually to the user. The input is the answer data extracted by the search, and the output is the displayed search results. Specifically, the server finds the relevant answer data and sends it to the terminal, and the terminal displays the search results in list format on the web browser. The user can click on each item to view details.

[1281] Step 8: Integrating the Emotion Engine

[1282] The server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when they enter a search query. The input is the user's facial expression and tone of voice, and the output is appropriate search results and messages. Specifically, the emotion engine analyzes the user's emotional state, and if the emotion is positive, it prioritizes relevant search results, and if the emotion is negative, it provides appropriate messages and information to reduce stress.

[1283] (Application example 2)

[1284] 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."

[1285] Factories require efficient management of maintenance information for multiple pieces of equipment and machinery. However, manual collection and analysis is time-consuming and labor-intensive, often resulting in insufficient data management and analysis. Furthermore, while it is known that workers' emotional states affect work efficiency and safety, there is a lack of systems that address this issue. Therefore, there is a need for a system that can automatically collect and efficiently search maintenance information, as well as a system that can provide feedback based on workers' emotions.

[1286] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: means for automatically collecting response data made public by local governments; means for converting the collected data into text format and performing data cleaning; means for generating embeddings from the cleaned text data using a generative AI model; means for registering the generated embeddings in a database and updating an index; means for providing a user with an interface for searching response data; means for receiving a user's search query, converting the query into an embedding, and searching the database; means for displaying search results to the user; means for processing maintenance information collected from a machine as text data; and means for providing search results according to the user's emotional state by incorporating an emotion engine for recognizing the user's emotions. This not only enables efficient collection and search of maintenance information, but also enables appropriate feedback to be provided according to the worker's emotional state.

[1287] 1. "Response data" refers to data showing the content of responses given in local government assemblies and other such meetings.

[1288] 2. "Text format" means a data format expressed as human-readable characters or sentences.

[1289] 3. "Data cleaning" is the process of removing redundant information and noise from collected text data.

[1290] 4. A "generative AI model" is an artificial intelligence model that is pre-trained using massive amounts of data.

[1291] 5. "Embedding" is a method of representing text data as a numerical vector.

[1292] 6. A "database" is a system for efficiently storing, managing, and retrieving structured data.

[1293] 7. "Interface" refers to the operation screens and input fields that allow users to interact with the system.

[1294] 8. A "query" is a question or keyword entered by a user to search for specific information.

[1295] 9. "Search query" means input to the system to specify the information to be searched.

[1296] 10. "Machinery" refers to production equipment and devices installed in factories and production lines.

[1297] 11. "Maintenance Information" means data relating to the maintenance and repair of facilities and machinery.

[1298] 12. An "emotion engine" is an algorithm or system for determining a user's emotional state.

[1299] 13. "Feedback" refers to the response or information that a system returns in response to a user's input or actions.

[1300] To implement this invention, a server first accesses the official websites of each local government and automatically collects publicly available response data. Specifically, the server uses web crawling technology to extract and download response data links based on specific HTML tags and classes. For example, there is a process for collecting PDF files from the websites of multiple local government organizations in bulk.

[1301] Next, the server converts the collected PDF files into text format. This is done by extracting the text from each page using a PDF to text conversion library (e.g., PyPDF2) and compiling it into a single text file. After that, data cleaning is performed, removing unnecessary line breaks and special characters, and generating formatted text data. For example, there is a process to remove unnecessary information from the text data extracted from the PDF files.

[1302] The server then uses a generative AI model (e.g., OpenAI GPT-3 / BERT) to generate embeddings from the cleaned text data. These embeddings convert the text data into numerical vectors that can be used for efficient search and comparison. Specifically, the text data is input into the generative AI model, which outputs a fixed-length vector. This vector is later stored in a database and used as a search index.

[1303] The server then registers the generated embeddings in a database (e.g., Elasticsearch) and updates the index, allowing for faster response when a user enters a search query. Specifically, the process involves adding the newly generated embeddings to the database and rebuilding the index with the existing data.

[1304] The server also provides a user interface to allow users to search for the answer data. This interface is installed on a terminal and includes a search box and a search result display area. For example, there is an interface in which a search box is provided in a web browser, and when a user enters a keyword, related answer data is displayed.

[1305] Next, when a user enters a search query, the server converts the query into an embedding and searches the database. The generated embedding is compared with existing embeddings in the database to extract the most relevant answer data. For example, if a user enters "machine wear diagnosis," there is a process to convert the query into an embedding, perform a database search, and extract relevant answer data.

[1306] The server also returns the search results to the terminal and visually displays them to the user. The user can easily refer to and manipulate the related answer data. For example, the related answer data is displayed in list format on the web, and the user can click on each item to view details.

[1307] Finally, the server is equipped with an emotion engine that recognizes the user's emotions and determines the user's emotional state when entering a search query. This emotion engine prioritizes relevant search results when the user is in a positive emotion, and displays appropriate results and messages to reduce stress when the user is in a negative emotion. For example, the emotion engine analyzes the user's facial expressions and tone of voice, and if the engine determines that the user is in a negative emotion, it provides an interface that provides information to help them relax.

[1308] Prompt Sentence Examples

[1309] Parliamentary response data analysis. Generate an embedding from the following text:

[1310] "Answer regarding environmental policy..."

[1311] This invention not only makes it possible to efficiently collect and search maintenance information, but also makes it possible to provide appropriate feedback to the worker according to his or her emotional state.

[1312] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1313] Step 1:

[1314] The server accesses the official websites of local government organizations and collects response data. Specifically, the server crawls multiple specified websites using web crawling technology (e.g., BeautifulSoup, Scrapy), extracts links to response data based on specific HTML tags and classes, and downloads PDF files. The input is the website URL, and the output is the downloaded PDF file.

[1315] Step 2:

[1316] The server converts the collected PDF files into text format. Specifically, it uses the PyPDF2 library to extract text from each page of the PDF file and combines it into a single text file. The input is a PDF file and the output is a text file.

[1317] Step 3:

[1318] The server cleans the text files. Specifically, it uses libraries such as NLTK to remove unnecessary line breaks, special characters, and unnecessary header and footer information, generating readable, formatted text data. The input is the raw text file, and the output is the cleaned text data.

[1319] Step 4:

[1320] The server generates embeddings from the cleaned text data using a generative AI model (e.g., OpenAI GPT-3 / BERT). Specifically, the text data is input to the AI ​​model as a prompt sentence, and a fixed-length vector is output. The input is the cleaned text data, and the output is the generated embedding.

[1321] Step 5:

[1322] The server registers the generated embeddings in a database (e.g., Elasticsearch) and updates the index. Specifically, it adds the newly generated embeddings to the database and rebuilds the index with the existing data. The input is the generated embeddings, and the output is the updated database index.

[1323] Step 6:

[1324] The terminal provides the user with an interface for searching answer data via a user interface. Specifically, the terminal displays a web browser interface including a search box and a search result display area. The input is the user interface design specification, and the output is the interface available to the user.

[1325] Step 7:

[1326] When a user enters a search query, the server converts the query into an embedding and searches the database. Specifically, the search query is input into the AI ​​model, and the generated embedding is compared with existing embeddings in the database. The input is the user's search query, and the output is highly relevant answer data.

[1327] Step 8:

[1328] The server returns the search results to the terminal and displays them visually to the user. Specifically, it sends the relevant answer data in a list format to the terminal so that the user can view the details. The input is the search results, and the output is the displayed search results.

[1329] Step 9:

[1330] The server uses an emotion engine to recognize the user's emotional state. Specifically, it analyzes the user's facial expressions and voice tone using the device's camera and microphone. The input is real-time video and audio data, and the output is the user's emotional state.

[1331] Step 10:

[1332] The server provides appropriate search results and messages based on the user's emotional state. Specifically, if the user's emotional state is positive, it prioritizes relevant results, and if the user's emotional state is negative, it displays relaxing information and messages. The input is the user's emotional state, and the output is the adjusted search results and messages.

[1333] 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.

[1334] 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.

[1335] 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.

[1336] 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.

[1337] FIG. 9 illustrates 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 behaviors 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.

[1338] 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.

[1339] 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).

[1340] 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.

[1341] 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."

[1342] 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.

[1343] 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).

[1344] 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.

[1345] 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.

[1346] 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.

[1347] 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.

[1348] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1349] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1350] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1351] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1352] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1353] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1354] The following is further disclosed regarding the above embodiment.

[1355] (Claim 1)

[1356] A means of automatically collecting response data published by each local government, and

[1357] means for converting the collected data into a text format and performing data cleaning;

[1358] A means of generating embeddings from the cleaned text data using a generative AI model; and

[1359] A means to register the generated embeddings in the database and update the index,

[1360] means for providing a user with an interface for searching answer data;

[1361] A means to receive a user's search query, convert the query into an embedding, and search the database;

[1362] a means for displaying search results to a user;

[1363] A system including:

[1364] (Claim 2)

[1365] 2. The system of claim 1, wherein the data collection means automatically downloads data in a specified file format from a plurality of different websites.

[1366] (Claim 3)

[1367] The system of claim 1, wherein a pre-trained language model is used as the generative AI model.

[1368] "Example 1"

[1369] (Claim 1)

[1370] A means of automatically collecting response data published by each local government, and

[1371] means for converting the collected data into a text format and performing data cleaning;

[1372] A means of generating embeddings from the cleaned text data using a generative AI model; and

[1373] A means to register the generated embeddings in the database and update the index,

[1374] means for providing a user with an interface for searching answer data;

[1375] A means to receive a user's search query, convert the query into an embedding, and search the database;

[1376] a means for displaying search results to a user;

[1377] A system including:

[1378] (Claim 2)

[1379] 2. The system of claim 1, wherein the data collection means automatically downloads data in a specified file format from a plurality of different websites.

[1380] (Claim 3)

[1381] The system of claim 1, wherein a pre-trained language model is used as the generative AI model.

[1382] "Application Example 1"

[1383] (Claim 1)

[1384] A means of automatically collecting response data published by each local government, and

[1385] means for converting the collected data into a text format and performing data cleaning;

[1386] A means of generating embeddings from the cleaned text data using a generative AI model; and

[1387] A means to register the generated embeddings in the database and update the index,

[1388] means for providing a user with an interface for searching answer data;

[1389] A means to receive a user's search query, convert the query into an embedding, and search the database;

[1390] A way for store employees to quickly search for product information and FAQs in response to customer inquiries,

[1391] a means for displaying search results to a user;

[1392] A system including:

[1393] (Claim 2)

[1394] 2. The system of claim 1, wherein the data collection means automatically downloads data in a specified file format from a plurality of different websites.

[1395] (Claim 3)

[1396] The system of claim 1, wherein a pre-trained language model is used as the generative AI model.

[1397] "Example 2: Combining Emotion Engines"

[1398] (Claim 1)

[1399] A means of automatically collecting response data published by each local government, and

[1400] means for converting the collected data into a text format and performing data cleaning;

[1401] A means of generating embeddings from the cleaned text data using a generative AI model; and

[1402] A means to register the generated embeddings in the database and update the index,

[1403] means for providing a user with an interface for searching answer data;

[1404] A means to receive a user's search query, convert the query into an embedding, and search the database;

[1405] a means for displaying search results to a user;

[1406] means for recognizing the emotional state of a user and including an emotion engine for displaying search results and appropriate messages;

[1407] A system including:

[1408] (Claim 2)

[1409] 2. The system of claim 1, wherein the data collection means automatically downloads data in a specified file format from a plurality of different websites.

[1410] (Claim 3)

[1411] 2. The system of claim 1, wherein a pre-trained language model is used as the generative AI model.

[1412] "Application example 2 when combining emotion engines"

[1413] (Claim 1)

[1414] A means of automatically collecting response data published by each local government, and

[1415] means for converting the collected data into a text format and performing data cleaning;

[1416] A means of generating embeddings from the cleaned text data using a generative AI model; and

[1417] A means to register the generated embeddings in the database and update the index,

[1418] means for providing a user with an interface for searching answer data;

[1419] A means to receive a user's search query, convert the query into an embedding, and search the database;

[1420] a means for displaying search results to a user;

[1421] means for processing the maintenance information collected from the machine as text data;

[1422] a means for providing search results according to the emotional state of the user by incorporating an emotion engine for recognizing the user's emotions;

[1423] A system including:

[1424] (Claim 2)

[1425] 2. The system of claim 1, wherein the data collection means automatically downloads data in a specified file format from a plurality of different websites.

[1426] (Claim 3)

[1427] The system of claim 1, wherein a pre-trained language model is used as the generative AI model. [Explanation of symbols]

[1428] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. A means to automatically collect response data published by each local government, and means for converting the collected data into a text format and performing data cleaning; A means of generating embeddings from the cleaned text data using a generative AI model; and A means to register the generated embeddings in the database and update the index, means for providing a user with an interface for searching answer data; A means to receive a user's search query, convert the query into an embedding, and search the database; a means for displaying search results to a user; A system including:

2. 2. The system of claim 1, wherein the data collection means automatically downloads data in a specified file format from a plurality of different websites.

3. The system according to claim 1, wherein a pre-trained language model is used as the generative AI model.

Images