Information generation system

Abstract

This enables flexible information provision tailored to individual user needs. [Solution] An information generation system 1 for generating information about fluctuating products comprises: an information collection means 10B for collecting raw information about fluctuating products from external information sources 5A, 5B, 6A, 6B or internal information source 77; and an information generation means 40 for generating generated information about fluctuating products based on the raw information collected by the information collection means.

Description

Technical Field

[0001] The present invention relates to an information generation system.

Background Art

[0002] Conventionally, an information providing system for advising individual investors has been known. For example, Patent Document 1 discloses a system that generates advice according to the amount of the user's initial set funds (investment amount) and displays it on the user's terminal.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] It has been difficult for conventional information providing systems to provide appropriate information.

Means for Solving the Problems

[0005] In order to solve the above problems, an information generation system according to an aspect of the present invention is an information generation system that generates information related to variable products, and includes information collection means for collecting original information related to variable products from an external information source or an internal information source, and information generation means for generating generated information related to variable products based on the original information collected by the information collection means.

[0006] According to the above configuration, generated information such as reports, explanatory texts, commentary texts, answers to questions, advice information, etc. can be automatically generated from the original information related to variable products collected from an external information source or an internal information source.

[0007] Furthermore, in an automatic report generation system according to one aspect of the present invention, there may be a distribution means that automatically distributes the generated information generated by the information generation means to a user when predetermined distribution conditions are met.

[0008] According to the above configuration, the generated information produced by the information generation means can be delivered to the user without human intervention.

[0009] Furthermore, an information generation system according to one aspect of the present invention may have a user information storage means for storing user-specific information of a specific user, and the information generation means may generate the generated information based on the user-specific information stored in the user information storage means and the raw information collected by the information collection means.

[0010] According to the above configuration, generated information is automatically generated from raw information about variable products collected from external or internal sources, in a format that is tailored to each user's individual information. Therefore, it is possible to automatically generate generated information such as reports, descriptions, explanatory texts, answers to questions, and advice information that include an appropriate amount of useful information about variable products for each individual user, taking into account the individual circumstances of each user.

[0011] Furthermore, in an information generation system according to one aspect of the present invention, the user-specific information may include user attribute information and request information, and the information generation means may generate the generated information based on both the attribute information and the request information.

[0012] With the above configuration, by using both user attribute information and request information, it becomes possible to create generated information such as reports that take into more detail the individual circumstances of each user.

[0013] Furthermore, in an information generation system according to one aspect of the present invention, the individual user information may include user behavior history information and held stock information, and the information generation means may generate the generated information based on both the behavior history information and the held stock information.

[0014] With the above configuration, it becomes possible to automatically generate information such as individual securities analyst reports for each user, based on each user's past behavioral history and the securities held by each user.

[0015] Furthermore, in an information generation system according to one aspect of the present invention, the user-specific information may include information on the distribution schedule set by the user, and the system may have a distribution means that distributes the generated information generated by the information generation means on a daily, weekly, or monthly basis based on the distribution schedule.

[0016] According to the above configuration, generated information produced by the information generation means is delivered daily, weekly, or monthly according to the delivery schedule set by the user, so that users can receive generated information such as reports at the time they want.

[0017] Furthermore, in an information generation system according to one aspect of the present invention, the user-specific information may include information on the content of the user's question, the source information may include time-series data, and the information generation means may analyze the question content using vector search technology and generate the generated information based on the time-series data or the results of the analysis of the time-series data.

[0018] With the above configuration, by performing a vector search on the user's question content, it becomes possible to perform a similarity search based on keywords that are not included in the question content but have a similar meaning. As a result, it becomes possible to automatically generate information such as reports that contain more appropriate answers to the user's questions.

[0019] Furthermore, in an information generation system according to one aspect of the present invention, the user-specific information may include information on the user's selection or instructions regarding the chapter structure of a report, and the information generation means may generate the generated information such that it includes a report having the chapter structure selected or instructed by the user.

[0020] According to the above configuration, it is possible to automatically generate generated information that includes a report with the chapter structure requested by the user. This makes it possible to create generated information that includes a report in a format that more closely matches the user's requirements.

[0021] Furthermore, in an information generation system according to one aspect of the present invention, the user-specific information may include the user's holdings, investment style, risk tolerance, or areas of interest, and the information generation means may generate the generated information such that it includes a report whose chapter structure is dynamically changed based on the user's holdings, investment style, risk tolerance, or areas of interest.

[0022] According to the above configuration, it is possible to automatically generate generated information that includes reports containing an appropriate amount of useful investment information for the user, depending on their holdings, investment style, risk tolerance, or areas of interest. In addition to holdings, investment style, risk tolerance, and areas of interest, the system may also use information such as the user's age, asset information, investment experience, and bank / securities account information.

[0023] Furthermore, in an information generation system according to one aspect of the present invention, the user-specific information may include the user's attribute information, risk tolerance, or holdings, and the information generation means may generate the generated information such that it includes a report in which the weights of technical indicators and fundamental analysis are adjusted based on the user's attribute information, risk tolerance, or holdings.

[0024] According to the above configuration, it is possible to automatically generate generated information including a report in which the amount of technical indicators and fundamental analysis information is adjusted according to the user's attribute information, risk tolerance, or held stocks. As a result, it is possible to create generated information including a report in which information useful for each individual user who is an investor is included in an appropriate amount without excess or deficiency.

[0025] Also, in the information generation system according to one aspect of the present invention, the original information may include time-series data, and the information generation means may generate the generated information when the time-series data exceeds a preset threshold value.

[0026] According to the above configuration, when the time-series data exceeds a predetermined threshold value, generated information based on the time-series data is generated. According to this, for example, a large change in the situation can be quantitatively determined, so that the timing when such a change occurs can be clearly grasped. Therefore, the user can quickly obtain generated information in which such a change is taken into account, for example, at the timing when a large change in the situation occurs.

[0027] Also, in the information generation system according to one aspect of the present invention, the original information may include technical indicators, and the information generation means may generate the generated information so that the generated information includes a report including text that dynamically changes based on the numerical values of the technical indicators.

[0028] Generally, updating report content in response to changes in time-series data such as technical indicators and stock price data requires manual work. The above configuration allows for the automatic generation of generated information, including reports containing text that dynamically changes based on the values ​​of technical indicators. This enables the automatic generation of more useful information by including investment advice (for example, whether to buy, sell, or hold a particular stock) and other necessary investment information, based on the values ​​of the technical indicators. The content of the text that dynamically changes based on the values ​​of the technical indicators may show individual decision results, include useful investment information, or present an overall assessment of the technical indicators. Preferably, the automatically generated information includes, for example, charts or other diagrams related to technical indicators within the report, and when the pointer (cursor) is placed over, clicked, or touched, dynamically changing text based on the values ​​of the technical indicators is displayed in a pop-up or similar manner. In particular, in the case of charts showing time-series data (for example, charts showing the stock price trends of a specific company), when a past point in time is specified on the chart (when the cursor is placed over it), investment advice text or useful investment information for that past point in time is displayed, making it possible to check, for example, whether the stock of that particular company was a buy or a sell at that past point in time, and to confirm historical investment information.

[0029] Furthermore, in an information generation system according to one aspect of the present invention, the information generation means may start generating the information in response to a question or inquiry from a user.

[0030] According to the above configuration, generated information is automatically generated in response to a question or inquiry from the user. This allows the automatically generated information to be provided to the user, for example, in the form of an answer to the user's question or inquiry.

[0031] Furthermore, in an information generation system according to one aspect of the present invention, the information collection means may also collect other source information in addition to the source information relating to the fluctuating product, and the information generation means may perform a second process in addition to the first process of generating generated information relating to the fluctuating product, which generates other generated information based on the other source information collected by the information collection means.

[0032] According to the above configuration, it is possible to automatically generate not only generated information based on raw information about variable products, but also other types of generated information based on other raw information.

[0033] Furthermore, in an information generation system according to one aspect of the present invention, the information collection means may collect the other raw information from a relevant external or internal information source, depending on the type of other generated information generated by the information generation means performing the second processing.

[0034] According to the above configuration, it is possible to collect source information to be used for the automatic generation of a given type of generated information by linking with information sources that hold source information related to the type of generated information to be generated. For example, for a nursing care report, information on nursing care facilities that is updated over time is collected from a nursing care facility database, and for a travel plan report, information on weather forecasts and transportation that changes over time is collected from a weather forecast database and a transportation database. This makes it possible to collect appropriate source information to be used for the type of generated information that is automatically generated, and to automatically generate generated information with more appropriate content.

[0035] Furthermore, in an information generation system according to one aspect of the present invention, the information generation means may perform the second processing by having a computer execute a program to which a different algorithm is applied for each type of other generated information.

[0036] According to the above configuration, each type of generated information can be automatically generated using algorithms optimized for each type of generated information, thus enabling the automatic generation of each type of generated information with more appropriate content. For example, for caregiving reports, an algorithm that conforms to the latest long-term care insurance system can be used to automatically generate caregiving reports, and for travel planning reports, a schedule change algorithm based on weather forecasts can be used to create travel planning reports.

[0037] Furthermore, in an information generation system according to one aspect of the present invention, the information collection means may collect the raw information from a plurality of information sources, and the information generation means may generate the generated information based on a single raw information generated based on a plurality of raw information obtained from the plurality of information sources.

[0038] According to the above configuration, generated information is automatically generated using a single source information (secondary information) generated from multiple pieces of information (primary information) obtained from multiple information sources (data sources). Therefore, it becomes possible to automatically generate generated information that analyzes multiple source information and makes a comprehensive judgment. It is preferable to utilize so-called AI for the process of generating a single source information based on multiple pieces of information.

[0039] Furthermore, in an information generation system according to one aspect of the present invention, the information generation means may generate the generated information based on a single source information that is generated by assigning different weights to each of the multiple source information obtained from the multiple source information.

[0040] According to the above configuration, generated information is automatically generated using a single source information (secondary information) that is generated from multiple weighted source information (primary information), making it possible to automatically generate more appropriate generated information. For example, in the case of a securities analyst report, multiple source information such as technical indicators, fundamental analysis, time series data, and supplementary information can be weighted for each user, and generated information such as a securities analyst report can be automatically generated based on an overall judgment made by analyzing these multiple source information. It is preferable that the time series data includes the date and time information of when each time series data occurred.

[0041] Furthermore, in an information generation system according to one aspect of the present invention, the information generation means may generate generated information by integrating the generated information generated by the first process and the generated information generated by the second process.

[0042] With the above configuration, it becomes possible to automatically generate and provide to the user integrated generation information of different types.

[0043] Furthermore, in an information generation system according to one aspect of the present invention, the information collection means may collect the raw information after receiving a question or inquiry from a user.

[0044] A certain processing time is required between receiving a question or inquiry from a user and the information generation means automatically creating the generated information. Therefore, the generated information provided to the user will have a time lag of at least that processing time, and will inevitably be based on older source information. In this case, if the generated information is based on source information that the information gathering means has collected in advance before receiving the question or inquiry from the user, the generated information provided to the user will be based on even older source information. According to the above configuration, generated information is produced based on source information collected after the time the question or inquiry is received. This makes it possible to provide users with generated information based on more recent source information than when generated information is produced based on source information collected before the time the question or inquiry is received.

[0045] Furthermore, in an information generation system according to one aspect of the present invention, the information generation means may generate the generated information in the order of information collection date and time, information generation date and time, and viewing date and time, starting from the user's requested date and time.

[0046] According to the above configuration, the information generation process in the information generation means proceeds in the time series of request date and time → information collection date and time → information generation date and time → viewing date and time. In many cases, the usual order of information provision is information collection date and time → information generation date and time → request date and time → viewing date and time (for example, when an analyst creates and publishes a report and a user accesses and views it), but unlike that, by making the process order start from the user's request, it becomes possible to generate information on demand in response to requests. In other words, since information generation is performed in a time series starting from the user's request date and time, it becomes an on-demand time series flow in which information generation starts according to the user's request date and time, and information can be generated dynamically based on user requests. In contrast to inquiry-triggered information gathering, which begins only after receiving an inquiry or question from a user, this approach features a process sequence that starts from the user's request date and time, executing the information generation timeline (request date and time, information collection date and time, information generation date and time, viewing date and time) in a specific order. This can be described as rearranging the timeline of the information generation flow itself, starting from the user's request. [Effects of the Invention]

[0047] According to the present invention, it becomes possible to provide appropriate information. [Brief explanation of the drawing]

[0048] [Figure 1] This is an explanatory diagram showing the hardware configuration of the system according to the embodiment. [Figure 2] This is a block diagram of the user terminal and server according to the embodiment. [Figure 3] This is a functional block diagram of the system according to the embodiment. [Figure 4] This is a flowchart showing the processing flow in the system according to the embodiment. [Figure 5] This is an explanatory diagram illustrating an example of the structure of a report automatically generated by the system according to the embodiment. [Figure 6] This is an explanatory diagram showing an example of a display screen (partial) of a report automatically generated by the system according to the embodiment. [Figure 7] This is an explanatory diagram showing a continuation of the display screen shown in the example in Figure 6. [Figure 8] This is an explanatory diagram showing a continuation of the display screen shown in the example in Figure 7. [Figure 9] This is an explanatory diagram showing a continuation of the display screen shown in the example in Figure 9. [Figure 10] This is an explanatory diagram showing an example of a display screen in a conventional investment information provision system. [Figure 11] This diagram illustrates an investment support service that uses an information generation system to provide users with expert reports, similar to those provided by experts, in near real-time, based on the latest stock price information. [Figure 12] This is an explanatory diagram showing an example of the display screen when a user clicks the "Technical Analysis" button on the display screen in Figure 11, and the "Technical Analysis" report is displayed. [Figure 13] This is an explanatory diagram showing an example of the display screen when a user clicks the "Trend Analysis" button on the display screen in Figure 11, and the "Trend Analysis" report is displayed. [Figure 14] This is an explanatory diagram showing an example of the display screen when a user clicks the "Compare Last Week's Closing Prices" button on the display screen in Figure 11, and the "Compare Last Week's Closing Prices" report is displayed. [Figure 15]This is an explanatory diagram showing an example of the display screen when a user clicks the "25-Day Moving Average Deviation Analysis" button on the display screen in Figure 11, and the "25-Day Moving Average Deviation Analysis" report is displayed. [Figure 16] This is an explanatory diagram showing an example of the display screen when a user clicks the "Signal Analysis" button on the display screen in Figure 11, and the "Signal Analysis" report is displayed. [Modes for carrying out the invention]

[0049] Embodiments of the present invention are described below. The embodiments relate to a system and method for performing data retrieval and analysis using information-generating artificial intelligence (AI) to provide personalized information and reports to users in various fields such as finance, nursing care, travel, education, and healthcare. In particular, the embodiments relate to a system and method for analyzing structured and unstructured data collected in each field and supporting automated information provision and decision-making based on changes in the data. In the descriptions of each embodiment described below, for the sake of clarity, the same reference numerals will be used for components that have the same function as those described, and the descriptions will not be repeated. The content conveyed in each embodiment is not independent, and can be freely combined.

[0050] In fields such as finance, elderly care, travel, education, and healthcare, vast amounts of data are generated daily, but conventional information provision systems have struggled to flexibly respond to individual user needs. For example, in the financial sector, investors make investment decisions based on a vast amount of market information, while in the medical sector, medical reports are needed based on patients' symptoms and past diagnostic history. Conventional systems have struggled to adapt to the significant changes in needs that occur over time.

[0051] The embodiment is a system that searches and analyzes financial data, information and records related to nursing care, travel data, educational materials, medical data, etc., and uses AI to provide personalized information and reports tailored to each user's needs and circumstances. Furthermore, the embodiment also includes interactive information provision, automatic notification functions based on specific conditions, and even automated trading and response functions, realizing a system that supports user decision-making.

[0052] This embodiment allows users to receive personalized information and reports in a timely manner, supporting them in making appropriate decisions in various fields. Furthermore, the trigger system automatically provides information in response to changes in data, improving user convenience.

[0053] Figure 1 is an explanatory diagram showing the hardware configuration of System 1 according to an embodiment. The system 1 according to this embodiment mainly consists of a user terminal (terminal device) 2, which is a user terminal, and a server 3. The user terminal 2 and the server 3 are connected to each other via a communication network 4.

[0054] User terminal 2 is a computer device, such as a personal computer (PC), tablet, or smartphone, equipped with an operating unit operated by the user and a display unit for presenting information to the user. Server 3 is a computer device that performs an automated creation process, automatically generating information such as reports to be provided to the user using a computer program. Communication network 4 is a network including the Internet.

[0055] Figure 2 is a block diagram of the user terminal 2 and server 3 according to the embodiment. The main components of the user terminal 2 consist of a communication unit 2A, a control unit 2B, a display unit 2C, and an operation reception unit 2D. The communication unit 2A communicates with the server 3 via the communication network 4. The control unit 2B controls the entire user terminal 2 and consists of a computer including, for example, one or more processors and memory devices. The display unit 2C displays information under the control of the control unit 2B and consists of, for example, a liquid crystal display. The operation reception unit 2D receives user input and consists of, for example, a keyboard, mouse, or touch panel.

[0056] The main components of server 3 consist of a communication unit 3A, a control unit 3B, and a storage unit 3C. The communication unit 3A communicates with user terminal 2 via a communication network and with one or more external servers (external information sources). In this embodiment, server 3 can also communicate with various external servers 5A, 5B, ..., 6A, 6B, ... via the communication network. The control unit 3B controls the entire server 3 and consists of, for example, a computer including one or more processors, storage devices, etc. The storage unit 3C stores information or data under the control of the control unit 2B and also functions as an internal information source that stores a database of accumulated data and consists of, for example, a hard disk drive, flash memory, etc.

[0057] Figure 3 is a functional block diagram of System 1 according to an embodiment. In the system 1 of this embodiment, the control unit 3B of the server 3 mainly implements the functions of the information collection module 10, the information processing module (also called the information processing module) 20, the trigger module 30, the information generation module (also called the report generation module) 40, the interface module 50, and the automated trading module 60. It may also implement the functions of the risk management module.

[0058] [Information Gathering Module] Information gathering module 10 is a module that collects data from information sources in various fields (finance, nursing care, travel, education, and healthcare), and mainly consists of user information gathering module 10A and information gathering module 10B. This module collects the latest information using API integration, crawling, and real-time acquisition, and obtains structured data for each target field.

[0059] API integration is a technology that uses APIs (Application Programming Interfaces) to link data and functions between different applications and systems. For example, the Tokyo Stock Exchange, Inc. ("TSE") has been providing a service since 2021 to distribute timely disclosure information and stock prices in Web-API format to further improve the convenience of market data distributed by the TSE. Information collection module 10 connects to the TSE's service that distributes timely disclosure information and stock prices in Web-API format, which is connected to the Internet 4 via the communication unit 3A, using KPI integration technology to collect timely disclosure information and stock price information for any given stock. In addition, the crawler, through a program called crawling by information collection module 10B, visits target websites on the Internet 4, and uses link tags in HTML documents published on the websites as markers to visit web pages one after another, collecting information on linked sites such as HTML files and PHP files. The collected information is stored in a database in the storage unit of the information generation system.

[0060] The user information collection module 10A is a module that collects individual user information. Individual user information is user-specific information that is useful for understanding the individual circumstances of each user when providing information to users.

[0061] User-specific information used to understand the individual circumstances of each user includes, for example, user attribute information. User attribute information is information that indicates the attributes of each user, and is information that categorizes a large number of users according to their nature and characteristics. Specifically, user attribute information includes gender, age / age group, education level, occupation / industry, annual income, place of residence, employment status, marital status, asset status, investment experience information, bank and securities account information, email address information, LINE account information, telephone number, dialogue history information such as questions from the user and answers (including date and time information of the dialogue), user identification information such as user ID, password information, and membership level information such as paid member / free member. However, it is preferable that the attribute information is useful for creating individual expert reports that include text describing information that is useful for each user in an appropriate amount of detail.

[0062] Furthermore, individual user information can change over time. For example, email address information and password information may change. User asset status, such as the names of stocks owned and the number of shares held, also changes. Therefore, it is desirable that the user information stored in the user information storage unit 71 of the information generation system be updated as much as possible to reflect actual user information.

[0063] Therefore, the user information collection module 10A checks the accuracy of the user information stored in the user information storage unit 71 for each user, either annually or semi-annually. For example, it can send an email to each user via their email address to check if there have been any changes to their user information, or send a chat to their LINE account requesting them to check if there have been any changes to their user information. The user information collection module 10A can also send all or part of the user information stored in the user information storage unit 71 to each user terminal 2 and display it on the display unit 2C to check if there have been any changes to each user's user information. If a user finds information in the user information stored in the user information storage unit 71 that needs correction, it is desirable for them to correct that information by operating the operation reception unit 2D of the user terminal 2.

[0064] Furthermore, it is desirable that users be able to modify their user information in the information generation system if changes occur in their user information. For example, it is desirable that users be able to access all or part of the user information of the relevant user stored in the user information storage unit 71 of the information generation system from the user terminal 2 to check or modify the necessary user information.

[0065] Furthermore, individual user information includes, for example, user behavior history information. User behavior history information refers to information that shows the past actions taken by each user, and may include actions taken within or outside of this system. Preferably, user behavior history information is information about past actions that are useful for generating useful information (such as reports) for each user. For example, when automatically generating a securities analyst report, this could include past investment history (history of buying and selling financial products), financial assets currently held as a result of past investment actions (such as the names of stocks held), and the content of questions or inquiries the user has made in the past.

[0066] Furthermore, user-specific information includes, for example, user request information. User request information is information that indicates the needs of each user. Preferably, user request information is information that helps generate useful information (such as reports) for each user. For example, when automatically generating securities analyst reports, this could include the user's desired risk tolerance, the type and amount of information the user requests, etc.

[0067] The information gathering module 10B is a module that collects information to be provided to the user, and it collects not only the information itself that is provided to the user, but also a wide range of information used in generating that information. For example, when automatically generating a securities analyst report, the information gathering module 10B collects stock price data, corporate performance data, technical indicators, earnings presentation materials, and IR materials such as securities reports obtained from the market using API integration and crawling functions.

[0068] [Information Processing Module] The information processing module 20 is a module that extracts specific features from collected data and analyzes them using at least one of the following methods: clustering, vector search, JSON search, database search, API search, file search, or internet search. For example, when automatically generating securities analyst reports, this module can also provide personalized search results by weighting them according to individual user information such as the user's investment style.

[0069] Clustering involves calculating features from, for example, stock price movements, corporate performance trends, and technical analysis results. Based on these calculated features, clusters are formed for each individual stock, each price movement, and each group of stocks, using the characteristics of their price movements and performance (e.g., clusters of stocks with increasing revenue and profits, high dividends, and low P / E ratios), as well as technical indicators such as RSI and MACD, to group together those with similar technical indicator values. For example, when using price movement characteristics as features, trend indicators such as moving average deviation and Bollinger Bands can be used as features, and by adding K-Mart analysis and optimizing the number of clusters, the data can be divided into clusters. This has the special effect of allowing users to see at a glance which cluster their current holdings belong to. For example, a stock might belong to a cluster that has been in an upward trend for two months and shows no signs of a decline at the moment.

[0070] Vector search simplifies searching by converting metadata such as titles, summaries, and keywords from text materials like financial results presentations into vectors, enabling similarity searches and cross-searching of content with similar meanings.

[0071] JSON search enables searching by saving metadata such as title, summary, keywords, and article title in JSON format.

[0072] Database searching involves creating databases using PostgreSQL or similar tools to store articles, texts, securities reports, and other documents, allowing users to search for summaries of sales figures, company performance, and other relevant information.

[0073] These searches are not performed in isolation; rather, their interconnectedness is crucial, as they involve searching and displaying results within a dialogue, or searching and displaying results based on the content of the dialogue. By changing the search method and content according to the question, the theme of each chapter, and the trigger, it becomes possible to meet a variety of needs. For example, this includes searching based on the user's holdings, searching based on triggers, and automatically generating search results and reports. Based on the data collected by the information gathering module, various processes are performed in the information processing module to generate search results, clustering results, and so on.

[0074] [Trigger Module] The trigger module 30 provides a system that automatically resumes interaction with the user when specific conditions occur. For example, the trigger is activated and the user is notified when the stock price exceeds a certain threshold or when a specific technical indicator changes. For example, if the user's holdings (user-specific information) rise for three consecutive days, the trigger module 30 activates and a notification is sent to the user. Subsequently, interaction regarding technical indicators and company overview is initiated according to the user's investment style and other user-specific information. Based on this interaction, the user decides on the timing of buying and selling. The method of interaction with the user will be described later. After this trigger module is activated, various information is passed to the information processing module and processing proceeds, and information (search results, etc.) resulting from the various processing is generated.

[0075] [Information generation module] The information generation module 40 is an interactive artificial intelligence (AI) that possesses large-scale language model (LLM) technology capable of predicting the next word and constructing sentences. Based on an LLM that has learned from a large amount of data on the Internet 4, the information generation module 40 can skillfully respond to user questions like a human. In other words, when a user inputs a question or instruction in natural language to the information generation system, the information generation system gathers information from the Web (World Wide Web) on the Internet 4 and generates information that it deems optimal. Furthermore, the information generation system, which has an information generation module, can also converse with the user via voice, just like humans.

[0076] The information generation module 40 is a module that generates information such as reports tailored to various fields, including financial market reports, care plans, travel plans, educational performance reports, and medical diagnosis reports. In particular, it is preferable that the information generation module 40 generates reports including summaries, titles, and body text using AI or the like, based on previously processed information (search results generated by the information processing module), and generates information such as reports according to the trigger content. This makes it possible to provide information optimized for the user's situation and user questions. For example, when automatically generating a securities analyst report, the information generation module 40 generates a stock report that includes a company overview, financial summary, recent news, technical indicator analysis, trend analysis, and overall evaluation, based on the analysis results from the information processing module 20.

[0077] According to System 1 of the embodiment, when a user expresses interest in a particular stock, a report containing the latest financial results and technical indicators for that stock is automatically generated. This report is customized to take into account the user's individual information, such as the user's investment style and past interaction history, and is provided to the user terminal 2 via the interface module 50 described later, either in an interactive format, by email, or through a personal page.

[0078] [Interface Module] The interface module 50 is a module for providing users with generated information, such as reports, generated by the information generation module 40. It provides information in an interactive or non-interactive format, identifies individuals using a user ID, and provides personalized information by referring to user-specific information such as past conversation history. It also integrates with platforms such as LINE®, a two-way communication app that allows real-time chat, and My Page, to provide users with an environment that makes it easy to ask questions and to provide answers in an easy-to-understand manner.

[0079] [Automated trading module] System 1 of the embodiment may optionally include an automated trading module 60. The automated trading module 60 is a module that automatically performs transactions such as buying and selling of investment products such as stocks based on user settings and market conditions. This module, for example, automatically generates and provides securities analyst reports to the user, and works in conjunction with technical indicators and risk management tools to streamline the user's trading. For example, based on the user's settings, when the technical indicators of a specific stock's price meet certain conditions, a buy or sell order is automatically executed to buy or sell shares of that specific stock. In this case, a risk management tool may operate to evaluate the overall portfolio risk and control transactions to ensure appropriate trading is performed. Another example of this module's application is in the medical field, where it may perform medical responses (transactions of medical services) in response to changes in specific vital signs.

[0080] [Risk Management Module] The system 1 of the embodiment may further include a risk management module as needed. This risk management module performs risk assessments and implements management to minimize risks to the user's portfolio, investment strategy, and health status.

[0081] In the system 1 of the embodiment, the storage unit 3C of the server 3 mainly consists of a user information storage unit 71, a collected information storage unit 72, a prompt information storage unit 73, a response information storage unit 74, an interaction information storage unit 75, a program storage unit 76, a database storage unit 77, and the like.

[0082] The user information storage unit 71 primarily stores and accumulates user information (including user attribute information) from the user information collection module 10A of the information collection module 10, associating it with user identification information and collection date and time information for each user.

[0083] The collected information storage unit 72 primarily stores and accumulates information collected by the information collection module 10B of the information collection module 10 (for example, investment information collected from external or internal information sources), associating it with the information source and collection date and time information.

[0084] The prompt information storage unit 73 stores and accumulates prompt information generated by the information collection module 10B or the information generation module 40, along with the date and time information of when the prompt information was generated, for each user. For example, the prompt information storage unit 73 stores and accumulates prompt information from the user-specific information collected by the user information collection module 10A, such as how to respond to questions or inquiries from the user and from what position to generate the response results. This prompt information can be changed for each chapter, each user, each question, and each field of expertise, and can be given the characteristic of being dynamically changeable.

[0085] The response information storage unit 74 stores and accumulates response information for each user, associating it with user identification information and the date and time information of the question or inquiry. The format is not limited, but it may be saved in a report format such as PDF or HTML so that users can quickly check past response results later.

[0086] The dialogue information storage unit 75 stores and accumulates dialogue information between the interface module 50 and the user, associating it with user identification information and the date and time information of the dialogue for each user. Supplementary information such as the date and time of generation, the time involved in generation, user information, and interface information can also be included by recording them.

[0087] The program storage unit 76 stores programs (computer programs) that are executed by the computer constituting the control unit 3B of the server 3. When the computer of the control unit 3B executes the programs stored in the program storage unit 76, functions such as the information collection module 10, information processing module 20, trigger module 30, information generation module 40, interface module 50, and automated trading module 60 are realized.

[0088] The database storage unit 77 stores the databases referenced by the control unit 3B. The databases stored in the database storage unit 77 may include an information database (internal information source) containing information (for example, investment information) used by the information generation module 40 when automatically generating reports.

[0089] Furthermore, System 1 may utilize external services provided by external servers connected to Server 3 via the Internet. Figure 3 illustrates, for example, external servers 7A providing language information generation AI services, external servers 7B providing image information generation AI services, and external servers 7C providing audio and video information generation AI services. External server 7C may be separate servers, such as external server 7D providing audio information generation AI services and server 7E providing video information generation AI services.

[0090] Up to this point, the information generation system has been described as System 1 (Server 3) shown in Figure 3, but the information generation system is not necessarily Server 3. For example, by providing the user terminal 2 shown in Figure 3 with a control unit 3B, a storage unit 3C, and a communication unit 3A, the user terminal 2 can be constructed as an information generation system (System 1). In that case, the user, or the operator of terminal 2, can use the terminal 2's operation reception unit (keyboard or touch panel) to input questions and instructions in the form of document information to the information generation AI installed inside terminal 2 or to various information generation AI services published on the Internet 4. Furthermore, questions and instructions in the form of document information can also be input to the information generation AI by voice using the terminal 2's voice input means. In addition, information such as answers, follow-up questions, etc., generated by the information generation AI installed inside terminal 2 or to various information generation AI services published on the Internet 4 in response to questions and instructions input by the user or operator can be output and displayed on the terminal 2's display unit 2C.

[0091] The embodiment can collect and analyze structured and unstructured data used in various fields in a timely manner, and provide users with personalized information and automated decision support using AI. Specifically, the information collection module, information processing module, trigger module, information generation module, automated trading module (automated trading / automated response module), interface module, and risk management module work together to provide information and automated processing that meets the needs of each field.

[0092] The following describes specific embodiments of the system in the financial sector, including support for stock investment, health monitoring and response to abnormalities in the elderly care sector, and personalized travel plan proposals in the travel sector. Each embodiment provides a detailed explanation of how the system according to the present invention operates in its respective field and provides valuable information to the user.

[0093] Furthermore, detailed embodiments that specifically apply the system according to the present invention will be described. Embodiments 1 to 3 are examples of an automated information generation and distribution system based on user information. Embodiment 4 is an example of an automated information generation and distribution system that includes text content that changes dynamically in accordance with changes in time-series data that change over time. Embodiment 5 is an example of an automated information generation and distribution system that is automatically generated starting from various triggers.

[0094] The outline of Embodiment 1 is an example of an automated information generation and distribution system that automatically generates and distributes information desired by the user based on user information. For example, for users with a strong short-term orientation, the system could automatically generate reports that emphasize short-term technical indicators. The outline of Embodiment 2 is an example of an automated information generation and distribution system that automatically generates and delivers information desired by the user based on conditions specified by the user. For example, one example is a user specifying that a report on the trends of the New York market be delivered every morning at 8:00. The outline of Embodiment 3 is an example of an automated information generation and distribution system that answers user questions in real time and automatically generates and distributes the results of the answers to the user's questions. For example, one example is asking about recent trends regarding stock A and receiving an answer in report format. The outline of Embodiment 4 is an example of an automated information generation and distribution system that automatically generates and distributes information including time-series data, that is, information that changes over time. For example, one example is the automatic generation and distribution of information that changes over time, such as technical indicators and trend indicators based on stock price data that changes moment by moment. The generated information includes, for example, text information that changes dynamically in response to those changes. The outline of Embodiment 5 is an example of an automated information generation and distribution system in which a trigger is activated in response to changes in time-series data, and information is automatically generated and transmitted based on that trigger. For example, a question is one trigger, but changes in time-series data can also be one trigger. One example is that information is automatically generated and distributed based on these triggers.

[0095] [Embodiment 1] Embodiment 1 of the present invention will be described in detail below. This first embodiment is an example of an automated report generation system (information generation system) that automatically creates securities analyst reports as reports provided to users who are investors. However, the embodiments of the present invention are not limited to securities analyst reports, but can also be applied to other types of reports. For example, they can be applied to other types of reports such as economic reports, nursing care reports, travel planning reports, advertising reports, and customer reports.

[0096] Variable products refer to goods and services whose prices fluctuate according to demand and circumstances, and are also known as dynamic pricing or variable pricing systems. For example, airfares are variable products because they are more expensive during periods of high demand (such as Golden Week, Obon, and New Year's) and cheaper during periods of low demand. Similarly, hotel accommodations are variable products because they are more expensive during periods of high demand and cheaper during periods of low demand. Retail goods are also variable products because stores may discount unsold items as closing time approaches. Furthermore, seat prices at sporting events and concerts are variable products because seats closer to the stage are more expensive and seats further away are cheaper. Dynamic pricing originated in the hotel and airline industries, but in recent years it has gained attention in other sectors such as retail.

[0097] In particular, regarding investment information, volatile products refer to products whose future prices and profit margins may fluctuate due to changes in prices, interest rates, exchange rates, etc. Examples of volatile products related to investment information include stocks (a typical volatile product, with stock prices changing daily), bonds (whose price fluctuates depending on the market value if sold before maturity, and is greatly affected by interest rate fluctuations), and investment trusts (whose prices fluctuate because they are affected by price changes of the stocks and bonds they hold). FX, foreign exchange, and cryptocurrencies fall into the same category.

[0098] Information generation systems that generate information needed by users come in many varieties depending on the subject and field of information generation (for example, medical information generation systems that generate medical information, educational information generation systems that generate educational information, travel information generation systems, etc.). For example, in the field of investment, such as stocks, which are volatile commodities, information generation systems include stock reports that provide information on individual stock issues, economic reports that provide information on the state and trends of the economy, industry-specific reports that provide information on trends in a particular industry, and technical analysis reports that predict future price movements from the shape of charts that graph the trends in stock and foreign exchange prices and trading volume. These are useful to investors and are widely used to decide on investment targets. However, the attributes of investor users vary greatly, and if each one were to be addressed individually, a tremendous amount of information and reports would be required. Therefore, there is a challenge in being able to meet the diverse attributes and needs of investor users.

[0099] Even when considering only information generation systems that produce useful information for users regarding volatile products such as stocks and the financial / investment sectors, there are significant differences among users in terms of the stocks they hold, the stocks they are considering buying or selling (high-dividend stocks, stocks with shareholder benefits, high-growth stocks, etc.), and whether they want information from a micro or macro perspective.

[0100] This information generation system acquires user attributes and requests from survey results, records of dialogue between the information generation system and the user, self-reported information from the user, the results of the information generation system's analysis of individual user information, and the results of the information generation system's trend analysis of user judgments. It then automatically generates information and reports that can respond to the acquired individual user information (user attribute information and request information).

[0101] To explain with a specific example, it can be determined from the conversation records and survey results of user A, who holds stock A, that A places importance on technical and trend indicators, which are stored in the user information storage unit 71 within the storage unit 3C of server 3. Specifically, as user-specific information for user A, various user-specific information (such as user attribute information and request information) is stored in the user information storage unit 71 within the storage unit 3C, which is a user-specific information storage means provided within server 3, associated with user identification information (user ID information) that can uniquely identify user A. In addition, the user information storage unit 71 within the storage unit 3C, which is a user-specific information storage means, also stores the date and time information of when each user's user-specific information was stored or updated.

[0102] Based on the information stored in the user information storage unit 71 within the storage unit 3C, which serves as a means of storing individual user information, the information generation module 40 automatically generates a technical indicator analysis and trend analysis report for stock A at 4 PM every day for a specific user, Mr. A. The interface module 50, which serves as a means of distribution, then distributes this report to Mr. A's email address, LINE account, or My Page. Each user's information is uniquely identified by the storage unit 3C (user information storage means) of the server 3. The storage unit 3C records individual user information such as Mr. A's holdings, risk tolerance, investment style, and past investment history, and updates it daily. The storage unit 3C also records distribution destination information such as email addresses, LINE accounts, and My Pages, so that reports (information) are automatically distributed to specific destinations at the appropriate time.

[0103] Person A's email address, LINE account, and My Page information (such as My Page identification information) are stored in the user information storage unit 71 within the server that constructs this system 1. The system also provides an information generation system that generates appropriate information according to the user's attributes, such as an information generation system that reports 15 technical indicators (RSI, Bollinger Bands, Moving Average Deviation, etc.), which is more than the usual number.

[0104] For more details, if user A's individual user information specifies that they wish to have a report containing 15 technical indicators (more than the usual number), the information generation system will generate a report consisting of natural language information, image information, audio information, video information, and numerical information related to at least 15 indicators. Technical indicators are used to predict future movements by statistically and psychologically analyzing market information such as stock prices and trading volume. They provide information that allows one to judge market trends and whether the market is overbought or oversold based on the shape and patterns of charts that graph price and volume trends. Specific technical indicators include trend indicators such as moving averages, Ichimoku Kinko Hyo, and Bollinger Bands, and oscillator indicators such as RSI, Stochastic, and Psychological Line.

[0105] Furthermore, information generation systems can be applied not only to the financial sector but also to a wide range of fields such as healthcare, education, and travel. For example, in the healthcare sector, it is possible to automatically generate regular health management reports based on the patient's age and medical history, and in the education sector, it is possible to provide learning plans tailored to students' academic performance and learning history.

[0106] Furthermore, the information generation system may collect information from users through questionnaires or dialogue records regarding whether they prioritize fundamental analysis or technical analysis, and the system may set corresponding weights accordingly. For example, if user A is set to prioritize technical analysis (9 out of 10) and fundamental analysis (1 out of 10), then a report focusing on technical indicators will be generated for user A. Conversely, for user B, fundamental analysis will be prioritized, and the report will focus on recent earnings forecasts and changes in dividend yield.

[0107] Fundamental analysis is a method of analyzing market trends using data on fundamental economic conditions. Fundamental economic conditions refer to indicators that show the economic situation of a country or company. Fundamental analysis refers to analysis based on a wide range of information, including micro-indicators such as corporate performance, corporate forecasts, and various P / E ratios, as well as macroeconomic indicators such as economic growth rates, consumer price indices, and employment statistics, and monetary policy by central banks, politics, and geopolitics. Fundamental information refers to indicators, data, and information that show the economic situation of a country or company. In the case of a country or region, this includes economic growth rates, inflation rates, unemployment rates, fiscal balance, and balance of payments, while in the case of a company, it includes sales, profits, assets, liabilities, price-to-earnings ratio (PER), price-to-book ratio (PBR), and return on equity (ROE). Technical analysis, on the other hand, refers to analysis using technical indicators (RSI, Bollinger Bands, moving average deviation, etc.).

[0108] In the medical field, blood test items and information to be managed (such as blood pressure, weight, and body fat percentage) can be changed according to the user, and in the travel field, the weight given to whether shopping or dining is prioritized can be adjusted.

[0109] Furthermore, the information provision system may utilize user trend analysis and self-reported data to determine what kind of information users are seeking. It may also incorporate an AI-powered automatic learning function that utilizes the user's past browsing history, search history, and dialogue history. In this case, for example, user A's individual information, such as number of holdings = 5 stocks, frequently asked questions about a high-interest stock = stock D, preferences = high-dividend orientation, and trading tendency = several times a year, is accumulated through dialogue and questions. This data is stored in the user information storage unit 71 within the storage unit 3C of server 3, allowing for the retrieval of individual user information at any time and enabling its use in report generation. The content can also be updated at any time.

[0110] Furthermore, users can customize and configure the report content in detail. For example, user A can set the company performance section to 400 characters and the segment information section to 1000 characters, and can also specify which sections to omit. They can also configure the report delivery frequency and recipients (email, LINE, My Page, etc.).

[0111] User attribute information and requests vary widely, and each user's holdings, trading methods, and level of interest differ greatly. An information generation system that automatically generates reports tailored to a specific user's individual information and delivers them to that user has a special effect. In particular, this information generation system is highly effective because it can automatically generate reports that meet the diverse attributes and requests of users and deliver them at the appropriate time. By delivering customized reports based on a user's holdings and areas of interest, user satisfaction can be increased.

[0112] Furthermore, this information generation system can integrate with external APIs to acquire data in real time from external information sources such as external servers 5A, 5B, 6A, and 6B, and generate reports based on that data. For example, it can acquire information from multiple data sources (external information sources) simultaneously and generate separate reports for each in parallel. The information generation system also functions as an information provision system.

[0113] To illustrate with a specific example, the user information storage unit 71 in the storage unit 3C of server 3 stores individual user information about a particular user, Mr. B, such as: Mr. B owns 6,000 shares of stock B for the purpose of medium- to long-term investment; he is interested in quarterly financial results and revised earnings forecasts of company B, as well as news highly relevant to investing in stock B; he is currently in his 60s and is considering investing his retirement savings from a long-term perspective.

[0114] In this case, the information generation system will periodically generate reports for a specific user, Mr. B, focusing on fundamental analysis of stock B, of which Mr. B owns 6,000 shares. These reports will be sent to Mr. B's email address or LINE account.

[0115] Information about person B, such as their email address and LINE account, is stored in the storage unit 3C of server 3 (user information storage unit 71 or database storage unit 77) in association with user identification information such as a user ID that can uniquely identify person B. In addition, individual user information such as the fact that person B owns 6,000 shares of stock B, the purchase cost and date of purchase of stock B, the fact that the purchase funds were retirement funds, person B's investment objective is to achieve an annual return of approximately 5% through long-term investment, and their investment style prioritizes safety over taking risks, is also stored in the user information storage unit 71 within the storage unit 3C of server 3 in association with unique information that identifies person B.

[0116] The information generation system of Embodiment 1 will be described in more detail below. In the information generation system of Embodiment 1, the user information collection module 10A in the server 3 collects user-specific information such as user attribute information and request information from the user terminal 2 connected to the Internet, via the interface module 50 and communication unit 3A in the server 3.

[0117] Specifically, the user information collection module 10A collects user-specific information such as, for example, questionnaire results information received from the user terminal 2 via the interface module 50 (stored in the user information storage unit 71 of the storage unit 3C), dialogue record information between the information generation system (interface module 50) and the user terminal (stored in the dialogue information storage unit 75 of the storage unit 3C), information on the results of the analysis of user-specific information by the information processing module 20 (stored in the dialogue information storage unit 75 of the storage unit 3C), and information on the results of the trend analysis of user judgment by the information processing module 20 (stored in the user information storage unit 71 of the storage unit 3C).

[0118] The user information collected by the user information collection module 10A is stored in the user information storage unit 71, which is part of the storage unit 3C provided in the server 3, in association with user identification information that identifies the user and the date and time information collected.

[0119] The information generation module 40 in the information generation system is an information generation artificial intelligence (AI) system equipped with natural language dialogue (chat) functionality, such as ChatGPT, and can receive arbitrary inquiries and commands in text form and obtain answers to those inquiries and commands. The information generation module 40 can be installed in the control unit 3B of the server 3, or it can be installed outside the server 3 and utilize an information generation AI service (external servers 7A-7E) that is publicly available on the internet.

[0120] When using an information generation AI service such as ChatGPT that is publicly available on the internet, the information processing module 20 located in the control unit 3B of server 3 generates prompt information consisting of textual information in response to arbitrary inquiries or commands, and inputs this prompt information via the communication unit 3A to the external servers 7A-7E of the information generation AI service that are publicly available on the internet. Subsequently, the information processing module 20 of server 3 receives the responses to inquiries and commands created by the external servers 7A-7E of the information generation AI service that are publicly available on the internet, via the internet and the communication unit 3A.

[0121] The information generation module 40 generates an information collection program for a specific user based on the user-specific information of the specific user collected by the user information collection module 10A. The generated information collection program for the specific user is associated with identification information that identifies the specific user and stored in the program storage unit 76 within the storage unit 3C.

[0122] For example, the information gathering module 10B generates an information gathering program based on the user-specific information stored in the user information storage unit 71. This program is designed to collect technical analysis report information and trend analysis report information for stock A from publicly available information on the internet via the communication unit 3A at 4 PM every day for a specific user, Mr. A, and stores the generated information gathering program in the program storage unit 76.

[0123] In this case, the information processing module 20 starts an information gathering program for a specific user, Mr. A, before 4 p.m. (for example, at 3:30 p.m.) and causes the information gathering module 10B to perform information gathering. The information gathering module 10B collects the latest technical analysis report information and trend analysis report information for stock A from various information sources via the internet, consisting of information that includes one or more of the following: natural language information, image information, video information, audio information, and numerical information. The collected information is stored in the collected information storage unit 72, associating it with identification information that identifies Mr. A, information that identifies the information source from which the information was collected, and information on the date the information was collected.

[0124] The information processing module 20 inputs both the latest technical analysis report and trend analysis report information for stock A collected by the information gathering module 10B, and a prompt requesting the generation of the latest technical analysis report and trend analysis report information for stock A that takes into account that person A owns 6,000 shares of stock A. As a result, the information generation module 40 automatically generates a technical analysis report and trend analysis report information for stock A for a specific user, person A, in the form of information (report) that includes one or more of the following: natural language information, image information, video information, audio information, or numerical information.

[0125] The generated report (securities analyst report) containing technical analysis and trend analysis information for stock A, specifically for user A, is automatically delivered at 4:00 PM via interface module 50 and communication unit 3A to user A's email address, LINE account, or My Page. The delivered report containing technical analysis and trend analysis information for stock A, along with delivery date and time information and identification information to identify user A, is stored in, for example, the dialogue information storage unit 75 within the storage unit 3C of server 3.

[0126] The information generation module 40 can be installed inside the server 3, but it can also utilize publicly available information generation AI services (external servers 7A-7C) on the internet. For example, it can be substituted with natural language information generation AI services such as ChatGPT, OpenAI GPT, PerplexityAsk, and BingAI, which are available on the internet. Alternatively, it can be substituted with image information generation AI services such as Midjourney and Stable Diffusion, video information generation AI services such as Runway, and voice information generation AI services such as AITalk. The information generation module 40 installed inside the server 3 may be a multi-information generation AI system that combines not only natural language information information generation AI functions, but also image information generation AI functions, voice information generation AI functions, video information generation AI functions, etc.

[0127] When using various information generation AI services (external servers 7A-7C) available on the internet, the communication unit 3A sends the prompt information from server 3, which consists of arbitrary inquiries and commands, and the publicly available information collected by the information collection module 10B, which consists of text, images, video information, audio information, and numerical information, to the external servers 7A-7C of the various information generation AI services located on the internet.

[0128] As a result, various AI information generation services on the internet automatically generate response information that includes one or more of the information / reports, natural language information, image information, video information, audio information, or numerical information that a specific user, A, needs, based on the input prompt information, and send it back to server 3 via the internet as response information to the prompt information. Server 3, which receives the response information in communication unit 3A, processes the received response information by summarizing it, removing redundant parts, combining information, or personalizing it using the information processing module 20 within the server or an external language information generation AI service, and automatically distributes the processed information to the specific user A's email address, or A's LINE account or My Page via interface module 50 and communication unit 3A. The information that server 3 has provided to A is stored in the dialogue information storage unit 75 within storage unit 3C.

[0129] In this embodiment 1, the information generation system realizes an information generation system that autonomously generates personalized investment field information suitable for each user, according to each user's information and attributes.

[0130] Furthermore, after the information generation system generates information for person A, it is also possible to check the individual user information of users other than person A stored in the user information storage unit 71, extract other specific users X and Y who are similar to person A, and distribute the information intended for person A to X and Y either as is or customized for X and Y.

[0131] Furthermore, while Embodiment 1 is an automated information generation and distribution system that automatically generates and distributes information desired by the user from user information, there is also a method in which the user explicitly declares the information they desire. User preference declaration information, as individual user information, is information that each user declares to the information generation system regarding the information generation conditions and information generation rules for the information that the information generation system generates and provides to each user.

[0132] Specifically, user preference information includes, for example, the subject of the information the information generation system should generate for the user (e.g., US economic reports, fundamental information on a specific company R), the form of the information the information generation system should generate for the user (natural language information, image information, video information, audio information, numerical information, or any combination thereof), the information sources and names of the information to be collected (e.g., collecting securities reports from EDINET, operated by the Financial Services Agency and connected to the internet, or stock reports for individual stock issues published on the internet by each securities company, etc.), the frequency of information provision to each user (e.g., twice a year, monthly, weekly, twice a week, daily), and the timing of information provision to each user (e.g., beginning of the year, end of the year, beginning of the month, end of the month, beginning of the week, weekend, every Sunday, every Monday, every Friday, every morning, every evening, every night, weekly, twice a week, daily, specific times such as 8 a.m. or 4 p.m., etc.).

[0133] User preference information can be obtained by the information generation system by presenting each user with a list of conditions and rules and having each user respond, or by each user voluntarily inputting it into the information generation system.

[0134] One method for inputting user preference information from the user to the information generation system is for the user to create questions, inquiries, or instructions explaining the preference information using natural language on the user terminal 2, and then send the created questions, inquiries, or instructions to the information generation system. The information explaining the preference information, such as questions and inquiries, does not have to be in the form of strings, but can consist of any combination of images, videos, audio, strings, numbers, etc. The information processing module 20 inputs the user preference information to the information generation module 40, causing it to generate prompt information that specifies the processing steps for the information generation module 40 to generate information based on the user preference information. Questions and inquiries are instructions or questions that the user inputs to the information generation AI system to generate the desired information in a dialogue with the information generation AI system or in an interactive system such as a command line interface (CLI).

[0135] For example, suppose a user requests the following information: "Every Monday at 8:00 AM, I would like a report containing a US economic report for that time, information on stocks to buy or sell that week, and reasons for buying or selling, using the stock price charts of those stocks." When this user requests are input into an AI information generation system, the questions or inquiries generated by the AI ​​system might be something like: "Every Monday at 8:00 AM (Eastern Standard Time), please create a report containing the following information: 1. Latest US Economic Information: Briefly summarize key US economic indicators and market trends (e.g., GDP, unemployment rate, consumer confidence index, etc.) based on the latest data. Also, include any news or events that may have an impact on the future economy. 2. Recommended Stocks: Select several stocks that you recommend buying or selling this week, and explain the reasons for your recommendation for each stock. Your reasons should be based on the company's performance, the latest market analysis, trends, etc. 3. Stock Charts and Explanations: Create the latest stock charts for each suggested stock and include technical analysis based on those charts (e.g., support lines, resistance lines, moving averages, etc.). This will make it easier to understand the reasons for recommending a buy or sell. Format: This report should be visually organized in an easy-to-read format, and should include charts and diagrams as needed.

[0136] An example of user request information (an example of a question or inquiry) would be: "Every Monday before the stock market opens, please collect information on the current economic reports for Japan and the United States, the exchange rate of the Japanese yen against the US dollar, and international news, and send me a report that includes information on which stocks to buy or sell that week, along with the reasons for buying or selling, using the stock price charts of those stocks." Server 3, which receives such natural language questions and inquiries via the internet and communication unit 3A, converts the questions and inquiries into user request information using the information generation module 40 within Server 3, or using an information generation AI service (external servers 7A-7C) publicly available on the internet. The information collection module 10B then analyzes and interprets the user request information to determine the frequency and time of information provision for that specific user, the information to be collected and its sources, and the content of the information to be created, and then creates an information collection program for that specific user.

[0137] The information generation module 40 is, for example, an artificial intelligence system such as ChatGPT, which has the function of natural language dialogue (chat) and can receive arbitrary inquiries and commands in natural language and generate answers to those inquiries and commands. The information generation module 40 can receive inquiries and commands and generate answer information not only from natural language information, but also from images, video information, audio information, numerical information, or any combination thereof.

[0138] The system of this embodiment 1 can be applied not only to the financial sector but also to other fields such as healthcare, education, and travel. For example, in the healthcare sector, it is possible to automatically generate regular health management reports based on individual user information such as the patient's age, medical history, past diagnoses, and requests from the patient. In the education sector, it is possible to automatically generate learning plans (reports) based on individual user information such as students' grades and learning history. In the travel sector, it is possible to automatically generate optimal travel plans (reports) based on information collected from external or internal sources such as weather information and opening hours of tourist spots, as well as individual user information such as user preferences.

[0139] Specifically, for Ms. C (in her 70s, with a chronic illness), it will be possible to generate regular health management reports based on her health checkup data (Ms. C's individual user information). For Ms. D, who is planning a family trip, it will be possible to generate a travel plan based on the weather and transportation schedules of her destination.

[0140] Another specific example is a stock investment support system. This example demonstrates the specific operation of a system that supports stock investment in the financial sector. This example describes the timely monitoring of stock price fluctuations, the generation of personalized reports based on the user's investment style, and the execution of automated trades.

[0141] (1) Collection of stock data The information gathering module collects stock price data, news, and corporate financial statements in real time or as needed. This aggregates all data related to the stock market. (2) Operation of the trigger system When the stock price meets conditions set by the user in advance (for example, when the moving average or RSI exceeds a certain value), the trigger system is activated and the user is notified interactively. (3) Processing of stock data By processing collected stock price data, financial statements, and other data in a timely manner, it is possible to convert numerical data into text data and generate search results and processing results by appropriately processing them according to the trigger content. (4) Report generation After the trigger is activated, a data processing step is performed, and based on the processing results, the information generation module generates a report containing the company's financial results, technical analysis, and the latest trends in the stock market, which is then provided to the user. Dynamically generated prompts play a crucial role in this process. (5) Automated trading Based on the automated trading conditions set by the user, the automated trading module is executed, and stocks are bought and sold. If a risk management module is provided, it monitors the system and adjusts the trading content as needed.

[0142] Another specific example is a health monitoring system in the field of elderly care. This example demonstrates the operation of a system that monitors the health data of elderly care recipients in real time or in a timely manner and takes appropriate action when an abnormality occurs.

[0143] (1) Collection of vital data The information gathering module continuously monitors the vital signs (heart rate, blood pressure, body temperature, etc.) of the person requiring care and saves the data to the cloud as needed. (2) Anomaly detection and trigger activation The trigger module immediately notifies care staff and family members if it detects an abnormality in vital signs (for example, an excessively high heart rate or a sudden rise in body temperature). (3) Processing of vital data, etc. This process involves, for example, determining whether the user's heart rate, blood pressure, or other parameters are abnormal compared to their past values. (4) Generate health report The information generation module generates a comprehensive health assessment report based on daily health data and past health history, and provides it to caregivers. (5) Automated response Upon detecting an anomaly, the automated response module acts based on a pre-configured response plan, ensuring appropriate action is taken. This includes contacting medical staff and preparing for emergency response.

[0144] Another specific example is a travel plan suggestion system. This example demonstrates how a system works by suggesting personalized travel plans to travelers and making timely reservations.

[0145] (1) Gathering travel information The information gathering module collects travel-related information (hotel rates, flight information, tourist attractions) and analyzes it based on the user's desired travel conditions. (2) Trigger for price fluctuation The trigger module monitors hotel and flight price fluctuations and automatically notifies the user when the price reaches their desired range. (3) Processing of weather data, etc. This process handles tasks such as deciding how to modify travel plans if, for example, rain is forecast for an extended period. (4) Travel plan generation The information generation module automatically generates personalized travel plans based on the traveler's preferences and provides users with reports containing information on tourist attractions, hotels, and flights. (5) Automatic booking The automated response module (automatic booking module) automatically books the best hotels and flights based on the conditions set by the user, supporting the entire trip.

[0146] These specific examples represent concrete applications in particular industries, but the embodiments are not limited to these and can be similarly applied in a wide range of fields. The combination of each module can be flexibly changed, and it can be expanded to other industries and fields.

[0147] [Embodiment 2] Next, Embodiment 2 of the present invention will be described in detail. This second embodiment is an example of an information generation system (information generation system) that generates reports or answer information based on user instructions or questions.

[0148] In Embodiment 1 described above, a system was described in which user-specific information (such as age, investment style, and holdings) of each specific user is recorded based on dialogue records and survey results between the user and the information provision system, and the interface module 50 automatically distributes useful information to each specific user based on the system's judgment. Embodiment 2 is an example of a system that can also handle cases in which each specific user gives clear instructions or questions to the system.

[0149] More specifically, in the system of this embodiment 2, users can select the type of information or report to be provided, such as whether it is a macroeconomic report or a stock report; select stocks that a particular user is interested in; specify the number of characters and style (such as bullet points) of the information and reports provided by the system; instruct the system on the chapter structure of the information and reports; and each specific user can clearly and specifically instruct the system on what kind of report they want.

[0150] For example, the system could provide the user with instructions regarding their desired conditions (user requests) by presenting multiple examples of user preferences and allowing the user to select their preferred option from these examples.

[0151] In this second embodiment, an information generation system is realized that automatically generates information and reports tailored to the instructions of a specific user by allowing each user to select or specify whether they prefer a macro or micro analytical perspective, the stocks they are interested in, the number of chapters in the information and reports provided to the user, the amount of information and reports provided (such as the number of characters or pages), and the frequency of information and reports being delivered from the system to a specific user.

[0152] For example, in the system of this embodiment 2, the user can specify the chapter structure and character count of the report in detail. For instance, it is possible to specify that the chapter on company performance should be 400 characters long, the segment information 1000 characters long, and the chapter on technical indicators should be omitted.

[0153] Furthermore, the system of this second embodiment can automatically generate reports only when there is a significant change in the user's holdings by using the trigger module 30. For example, it can generate and provide reports to the user when stock prices surge or plummet.

[0154] Each user has different holdings, and their interests in industries and companies also vary. Some users may be interested in US companies, while others may have a strong interest in the US economy and economic indicators. By allowing each user to specify their individual interests to the system, it becomes possible, for example, for the information generation module 40 to automatically generate economic reports (such as GDP reports, employment statistics reports, and financial data reports) that each specific user needs after the weekly release of economic indicators (announcements from the government, national and local authorities, the Bank of Japan, the Federal Reserve, etc.), and for the interface module 50 to distribute them to each specific user. Each specific user can obtain information and reports that are extremely valuable to them. The ability for each user to actively instruct the system on the target stock information, content of the information and reports, and specifications of the information and reports they need, and for the information generation module 40 to automatically generate the information and reports that each specific user needs, yields significant results.

[0155] In Embodiment 1 described above, the system primarily generates information and reports based on each specific user's individual user information, rather than based on user instructions. Instead, the system estimates the information and reports each specific user needs from their individual user information. In Embodiment 2, however, the system primarily generates information and reports each specific user needs according to clear instructions input by the user themselves into the system.

[0156] For example, regular press conferences and reports released by the Bank of Japan and the Federal Reserve, as well as employment statistics, are events that significantly influence the entire stock market, including prior forecasts. For this kind of fundamental investment information that greatly impacts the entire stock market, a specific user, Mr. C, can pre-configure the system with desired chapter structure specifications, such as an overview of the Bank of Japan's regular press conference, the Federal Reserve's future views on policy interest rates, this month's exchange rate trends and outlook, and the impact of policy interest rate trends on the US stock market. Then, at the beginning of each month, the system can automatically generate and deliver information and reports with the specific chapter structure requested by user Mr. C.

[0157] It is also possible to provide a system that combines the function in Embodiment 1 described above, which automatically generates information and reports needed by the user based on individual user information, with the function in Embodiment 2, which automatically generates information and reports needed by the user based on user instructions. Such a system would combine the flexibility of user instructions with automation. Therefore, the user can not only customize the report content, but also obtain the optimal report automatically generated by the system based on the user's attributes and requests. With this, for example, the user can give clear instructions to automatically generate a report, or the system can automatically select the optimal information based on the user's past data and dialogue history to automatically generate a report.

[0158] Furthermore, the selection system simplifies user instructions while allowing the system to generate flexible reports tailored to user needs. For example, users can specify details such as the number of chapters and words in the report, the frequency of delivery, and adjust whether it's macro or micro analysis, the analysis of stocks and industries, and even the delivery timing. Meanwhile, the system learns from the user's past instructions and dialogue history, enabling it to automatically adjust and generate reports in the future. For example, it includes a mechanism to generate timely reports during monthly earnings announcements and deliver them immediately.

[0159] The information generation system of Embodiment 2 will be described in more detail below. Embodiment 2 can also accommodate cases where a specific user can instruct or ask the system for the information they want at any time. In addition to natural language document information, each specific user can also specify the type of information to be generated, such as image information, video information, audio information, and numerical information, and instruct the system at any time what kind of report they want.

[0160] In the system of Embodiment 2, at any time, the user sends the information they need, specifically in the form of natural language information, image information, video information, audio information, numerical information, or any combination thereof, to the system's interface module 50 to provide the information / report description information they desire. In other words, each user creates the information / report description information (user-specific information) they desire by specifically combining the information they need, specifically in the form of natural language information, image information, video information, audio information, numerical information, or any combination thereof, and sends it to the system via the internet from the user terminal 2.

[0161] The explanatory information for the information / report requested by the user is stored in the dialogue information storage unit 75 within the storage unit 3C via the system's communication unit 3A, associated with user identification information that identifies the user.

[0162] By allowing each user to select or specify details such as whether the analytical perspective of the information / report explanations they desire is macro or micro, what stocks they are interested in, the number of chapters in the information / report provided to the user, the amount of information / report provided (e.g., number of characters or pages), the format of the information / report, the frequency and timing of distribution from the system to specific users, and the type of information the system provides to specific users (natural language information, image information, video information, audio information, numerical information, and any combination thereof), it becomes possible to automatically generate information / reports that better suit the specific user's needs, either on an ad-hoc basis or on a regular basis.

[0163] The system's information processing module 20 inputs explanatory information for information and reports requested by a specific user into the information generation module 40. The information generation module 40 creates prompt information for that specific user that reflects the explanatory information for information and reports requested by the specific user, and stores this prompt information in the prompt information storage unit 73 along with user identification information that identifies the specific user and the date and time information when the system received the explanatory information for information and reports requested by the specific user. The information processing module 20 then inputs the prompt information for the specific user into the information generation module 40.

[0164] When the information generation module 40 receives prompt information for a specific user, it instructs the information collection module 10B to collect necessary primary information from publicly available information on the internet, etc., based on the prompt information for the specific user. The primary information collected by the information collection module 10B is stored in the collected information storage unit 72 within the storage unit 3C, along with information identifying the specific user and date and time information.

[0165] For example, the specific steps for generating information in response to a user's instruction or question, such as "Tell me about Company X's automotive battery business," would be as follows:

[0166] (1) First, using the user's question information, "Tell me about X Company's automotive battery business," the information processing module 20 generates the first prompt information to be input to the information generation module 40 as instruction information in order for the information generation module 40 to generate appropriate answer information. Specifically, the information processing module 20 uses the user's question information to input the following first prompt information to the information generation module 40: "I would like the information generation AI to generate the following information: 'Tell me about X Company's automotive battery business.' So please create a prompt to instruct the information generation AI."

[0167] (2) Then the information generation module 40 generates the following second prompt information. Second prompt information: Please provide information regarding Company X's automotive battery business, including the following: 1. Business Overview: Please describe the overview of Company X's automotive battery business. Please also mention specific product lines (e.g., lithium-ion batteries, next-generation battery technology, etc.) and key technologies. 2. Market Position and Competition: Please describe Company X's market share and position in the automotive battery market, as well as its differences and competitiveness compared to its main competitors (e.g., Company Q, Company R, etc.). 3. Partnerships and Key Customers: Please also discuss your partnerships and supply relationships with automobile manufacturers (e.g., Company Y, Company W, etc.), as well as your future development plans. 4. Recent technological developments and innovations: Please describe, with specific examples, the latest technologies and product improvements that Company X is working on, as well as research and development aimed at improving battery performance and reducing costs. 5. Future Outlook: Analyze Company X's future growth strategy for its automotive battery business and its response to future market trends. Format: Please provide the report in a clear, visually organized format, including any relevant diagrams or charts.

[0168] (3) Next, the information processing module 20 inputs the second prompt information to the information generation module 40.

[0169] (4) The information generation module 40 collects necessary information from information published on the Internet 4 via the communication unit 3A and generates primary response information that corresponds to the user's instructions and questions.

[0170] Next, the system's information processing module 20 inputs the generated primary response information and user-specific response specification prompt information that realizes the response specifications for the specific user into the information generation module 40, and automatically generates secondary information, which is information and reports that are based on the specifications desired by the specific user and are specialized for that specific user. The information and reports created by the information generation module 40, which are based on the specifications desired by the specific user and are specialized for that specific user, are stored in the response information storage unit 74 of the storage unit 3C, along with information that identifies the specific user and date and time information. Subsequently, the information processing module 20 distributes the information and reports specialized for the specific user to the specific user via the interface module 50 and the communication unit 3A.

[0171] [Embodiment 3] Next, Embodiment 3 of the present invention will be described in detail. This third embodiment is an example of an information generation system that expands user question information and creates answers to the expanded questions.

[0172] Users have a wide variety of questions regarding finance and stocks. Even questions about individual stocks can range from questions about the price movements of a particular stock, such as "Why is it falling so much?", to questions about the business operations of a company, such as "What is the most profitable product for company D?". The traditional way to resolve these questions is for each user to search for articles and information themselves using keywords, gradually resolving their doubts. Large-scale conversational AI services like ChatGPT are good at answering universal questions, but they have a major weakness: they are very weak at handling time-series data such as stock prices, where the situation changes moment by moment.

[0173] The system incorporates processes for interpreting the meaning of each user's question information (vectorization), breaking down each user's question information into keywords, and converting numerical information into text information. This enables vector searches and full-text searches even for natural language questions from users regarding time-series data where the situation changes moment by moment, such as stock market trends and the management status of individual companies. Vector search utilizes machine learning (ML) to extract meaning and context from unstructured data such as text and images and convert them into numerical representations. Vector search, frequently used in semantic search, uses the Nearest Neighbor (ANN) algorithm to search for similar data. Vector search can generate faster and more relevant results compared to conventional keyword searches.

[0174] For example, in the case of a question from a specific user such as "Tell me about Company X's automotive battery business," Company X is the company name keyword, and automotive battery business is a related keyword, allowing for a search. However, the answer to the aforementioned question from a specific user also needs to include keywords and information about "lithium-ion batteries, the relationship between Company X and Companies Y and W," and "US factories and subsidies." Therefore, it is preferable for the system to have a supplementary information generation function that generates related information such as keywords and similar words that are not present in the natural language question from the specific user mentioned above.

[0175] Vector search is a method that enables the generation of keywords and related information not included in natural language questions from specific users. By vectorizing the natural language question information of a specific user and the information of Company X, it enables similarity searches.

[0176] Similarity search is a general term for searches that absorb variations in spelling and synonyms, or that flexibly interpret similar sentences. Unlike exact match search, it refers to a search method that can find results even if they do not exactly match the search criteria (keywords or a question entered in natural language). For example, if the question is "What are Company S's activities in Europe?", it becomes possible to find related articles and reports about Company S's history in Europe, why they expanded their main business in Europe, their competitiveness, market share, etc.

[0177] In the system of Embodiment 3, for example, when a user asks "Tell me about Company X's automotive battery business," a report including industry trends, market share, strengths, and weaknesses is automatically generated. This provides an answer (report) that not only provides factual information but also takes into account the company's position within the industry and future prospects.

[0178] Even without utilizing vector search, this includes all methods for generating indirect and multifaceted answers to questions. Specifically, it allows for the identification of related, supplementary, and similar information that is not present in the user's original question, such as identifying the names of companies other than Company X that have an automotive battery business (e.g., Company Y or Company W), or the technical term lithium-ion battery, which is a key battery in the automotive battery business. In other words, even for keywords not present in the user's natural language question, the system analyzes the user's question, identifies keywords and information not present in the original question through vector search, and performs information retrieval using supplementary, expanded, and similar question information. As a result, it is possible to retrieve internet paragraphs and page information that are in line with or similar to the meaning of the user's natural language question as search results.

[0179] By passing both the search results from the question information created in the specific user's natural language, and the search results obtained by searching a second set of question information that complements, expands, and is similar to the question information created in the specific user's natural language using vector search, along with the first prompt information regarding the user's instructions and questions mentioned above, to a large-scale language conversational AI service such as ChatGPT, it becomes possible to realize an information generation system that has a mechanism to obtain high-quality answer information for questions in the specific user's natural language. Next, a large volume of answer-related information consisting of a specific user-specific answer specification prompt information that realizes the answer information specifications required by the specific user, the first search results from the question information created in the specific user's natural language, and the second search results obtained by searching a second set of question information (enhanced question information) that complements, expands, and is similar to the question information created in the specific user's natural language using vector search, keyword search, JSON search, database search, etc., is input into a large-scale language model generation AI system such as ChatGPT. The system then extracts the important essence from the large volume of answer information and generates natural language answer information for the specific user, eliminating redundant information. The natural language response information provided to this specific user is generated by the information generation system based on predetermined response information specifications, such as a specified number of characters (e.g., 500 characters or 300 characters) and a specified format (e.g., within one page, both document information and image information, text information only, or bulleted list format).

[0180] Furthermore, the system of Embodiment 3 automatically generates related and similar information not included in the user's question using vector search, JSON search, SQL queries, etc. This makes it possible to complement the diverse information that cannot be handled by conventional keyword searches and provide more comprehensive answers.

[0181] Furthermore, the system of Embodiment 3 utilizes NLP technology to deeply understand the user's natural language questions. It uses named entity recognition and topic modeling to supplement relevant information not included in the question and generates the optimal answer based on the meaning of the question.

[0182] Instead of simply searching using natural language question information created by each user, this system also performs vector searches, SQL queries, and JSON searches on the natural language question information of specific users to identify related, supplementary, and similar information that is not present in the specific user's question information. Based on this related, supplementary, and similar information, it performs information searches on the internet, database searches, and file searches to interpret the meaning of the natural language question content, collect a sufficient amount of source data for answers, and then a generative AI system with a large-scale language model converts this source data into a predetermined number of characters and format, including important information while removing redundant information, and provides an answer to the specific user who asked the question. This information generation system, which has question information analysis means and generates answer information to provide to users by searching for information using supplementary, extended, and similar reinforced question information that complements the natural language question information of specific users, can overcome the limitations of keyword searches, where search results do not appear unless they match the keywords.

[0183] Furthermore, the system of Embodiment 3, by using the trigger module 30, can, for example, link with an external data source (external information source) and automatically generate reports in real time triggered by specific events such as stock price fluctuations or earnings announcements. This ensures that users always receive reports based on the latest information.

[0184] The system in Embodiment 3 is not limited to variable products or financial reports. As a concrete example, let's explain how to answer a question from a specific user such as, "What are some good restaurants in Marunouchi?" With a typical keyword search, blogs introducing good restaurants in Marunouchi would be found in the search results, and that information would be provided to the user who asked the question as an answer. The user would then search from there. In contrast, in the system in Embodiment 3, in response to a natural language question from a specific user such as "What are some good restaurants in Marunouchi?", the system converts the words "good restaurants in Marunouchi" into vector numbers, and then performs an information search that includes similar terms such as "Top-ranked restaurants in Marunouchi according to surveys," "Restaurants in Marunouchi with long lines," and "Best Italian restaurants in Marunouchi," thereby generating a report that provides answer information tailored to the questioner's purpose.

[0185] Furthermore, in the system of Embodiment 3, the specifications of the response information to the user, such as the chapter structure, number of characters, format, and data type (text information, image information, audio information), can be freely assembled according to the report, which is the target response information. The specifications of the response information to the user are stored and set in, for example, the storage unit 3C of server 3, but these stored response information specifications can be changed by the user or by the system. For example, the system can store response information specifications for each specific user.

[0186] For example, in a stock information report for a specific stock, which is a volatile investment product, if the chapter structure—industry trends, market share, and strengths and weaknesses in the industry—is designed to reflect the user's response information, then when a specific user asks a question about Company X's automotive battery business in natural language as described above, the system can automatically generate response information such as the trends in Company X's automotive battery business, its market share, and its strengths and weaknesses in the automotive battery business. This response information can be obtained by performing various searches, such as searching for Company X's market share and industry trends, in response to the aforementioned question asked by the specific user about Company X's automotive battery business.

[0187] Generative AIs like ChatGPT have to handle a wide range of general information and an extremely large volume of data, so they cannot keep up with the latest information that changes moment by moment. Embodiment 3 may also have the feature of being a time-series data information generation system that can handle time-series data in which numbers change depending on the date and time.

[0188] The information generation system of Embodiment 3 will be described in more detail below.

[0189] By passing both the first search result, obtained from the first prompt information (question information) created by a specific user, and the second search result, obtained by searching using second prompt information that complements, expands, or is similar to the first prompt information (question information) created by the specific user through vector search, etc., to the information generation module 40 or a large-scale language conversation AI service such as Chat GPT, it becomes possible to realize an information generation system that has a mechanism to obtain high-quality answer information to a specific user's question.

[0190] In other words, a large volume of primary collected information is obtained, consisting of both a first search result based on first prompt information (question information) generated from instruction / question information created by a specific user on user terminal 2, and a second search result obtained by searching using second prompt information (extended question information) which is created by supplementing and extending the first prompt information (question information) created by the specific user through vector search. Then, the obtained large volume of primary collected information is input into a large-scale language model generation AI system such as Chat GPT to extract the important essence from the large volume of answer information and generate answer information including natural language information for the specific user, with redundant information removed. This answer information including natural language information provided to the specific user is generated by the information generation module 40 of server 3 based on prompt information related to predetermined answer information specifications, such as a predetermined number of characters (e.g., within 500 characters or within 300 characters) and a predetermined format (e.g., within 1 page, both document information and image information, text information only, bulleted list format, etc.).

[0191] For example, suppose a specific user sends a question about X Company's business, such as "Please tell me about X Company's automotive battery business," to the information generation system. The information processing module of the information generation system stores the specific user's question in the dialogue information storage unit 75 within the storage unit 3C, associating it with information that identifies the specific user and date and time information. Next, the information processing module 20 inputs the specific user's question into the information generation module 40 and creates a first prompt information for the specific user from the question information. The first prompt information created by the information generation module 40, along with user identification information that identifies the specific user and date and time information, is stored in the prompt information storage unit 73 by the information processing module 20.

[0192] Next, the information processing module 20 inputs the question information of a specific user into the information generation module 40 and performs the steps of breaking down the question information of the specific user into keywords and interpreting the meaning of the question information of the specific user (vectorization). By vectorizing the question information of the specific user and the information of Company X, which is a keyword present in the question information of the specific user, it becomes possible to obtain keywords such as the technical term "lithium-ion battery," the names of Company X's business partners, "Company Y" and "Company W," "State M" where Company X's US factory is located, and "subsidies from State M," which are significant related, complementary, and similar information.

[0193] The information generation module 40 vectorizes the question information of a specific user to obtain significant related, complementary, and similar information, such as the technical term "lithium-ion battery," the names of business partners of Company X, "Company Y" and "Company W," and keywords such as "State M" where Company X's US factory is located, and "subsidies in State M," and adds these keywords to the question information of the specific user to generate extended question information for the specific user. The information processing module 20 inputs the extended question information of the specific user into the information generation module 40 to generate extended second prompt information. The generated second prompt information, along with user identification information that identifies the specific user and date and time information, is stored in the prompt information storage unit 73 by the information processing module 20.

[0194] Next, based on the first prompt information, the information generation module 40 collects information from publicly available information on the internet and other sources using the information collection module 10B, and obtains the first collected information. The first collected information includes information such as that there is a business partnership between Company X and Company T, and that Company X is producing batteries for Company T at a gigafactory in another state N.

[0195] The first collected information gathered by the information gathering module 10B is stored in the collected information storage unit 72. Next, the information generation module 40, based on the second prompt information, causes the information gathering module 10B to perform a second information gathering operation from information publicly available on the internet, thereby obtaining the second collected information. The second collected information includes information not included in the first collected information, such as the names of X Company's business partners, "Y Company" and "W Company," the state of "M" where X Company's lithium-ion battery factory is located, and "subsidies from M State."

[0196] The second set of collected information, gathered by the information gathering module 10B, is stored in the collected information storage unit 72. Subsequently, the information generation module 40 reads both the first and second sets of collected information from the collected information storage unit 72, combines them, and generates the third set of collected information. The third set of collected information is stored in the collected information storage unit 72.

[0197] The third set of collected information is more information-rich than the first and second sets of collected information, but it may be too much information or too redundant. Therefore, the information generation module 40 can create a fourth set of collected information by removing redundant information from the third set of collected information and summarizing it. For example, if a specific user requests information of about 250 characters, the information generation module 40 will generate a fourth set of collected information of about 250 characters.

[0198] An example of the fourth type of information to be collected is as follows: "Company X's automotive battery business has partnered with Company T, leveraging its lithium-ion battery technology. It produces batteries for Company T at its Gigafactory in State N, giving it massive production capacity. Furthermore, Company X's automotive battery business is growing through strong partnerships with Companies Y and W. Utilizing subsidies from State M and the federal government, it is constructing a factory in State M and establishing a system to meet the future needs of the EV market. As a result, Company X is expected to play a crucial role in the advancement of electrification."

[0199] The fourth collected information is stored in the collected information storage unit 72. The information processing module 20 transmits the fourth collected information generated by the information generation module 40 to the user terminal 2 of a specific user as answer information to the specific user's question information via the interface module 50 and the communication unit 3A. The fourth collected information generated by the information generation module 40 is stored in the answer information storage unit 74 within the storage unit 3C, along with user identification information that identifies the specific user and date and time information.

[0200] Furthermore, the information generation module 40 confirms the potential for further utilization of the answer information generated in response to a specific user's question. Specifically, the information generation module 40 analyzes the individual user information (such as user attribute information) of all users stored in the memory unit 3C to find other specific users who need the answer information generated in response to a specific user's question.

[0201] For example, the information generation module 40 finds another specific user from the user information storage unit 71 in the storage unit 3C who owns 5,000 shares of Company W and has a strong interest in electric vehicles. In this case, the information generation module 40 generates information either as is, or by adjusting the response information sent to the specific user to match the other specific user's preferences in terms of information volume and format. The generated information for the other specific user is then sent to the other specific user's user terminal 2 via the interface module 50 and the communication unit 3A.

[0202] Instead of collecting information solely from natural language-based question information created by each user, the system also vectorizes the natural language-based question information of specific users to generate related, supplementary, and similar information that is not present in the question information created by those specific users. By considering this related, supplementary, and similar information, the information collection module 10B collects information from the internet and other sources, thereby interpreting the meaning of the natural language-based question content and collecting a sufficient amount of raw data for responses.

[0203] Then, the collected raw data for answers is converted into a predetermined number of characters and format, retaining important information while removing redundant information, by the information generation module 40 and the information generation AI service systems 7A-7C, which are information generation means, to generate answer information for the specific user who asked the question. An information generation system that has a question information analysis means and generates answer information to provide to users who perform information searches by supplementing, extending, and similarly enhanced question information of a specific user's natural language question information can overcome the limitations of keyword search, where search results do not appear unless they match the keywords.

[0204] [Embodiment 4] Next, Embodiment 4 of the present invention will be described in detail. The information generation system according to this fourth embodiment is an example of a time-series data information generation system that can handle time-series data in which numbers change moment by moment depending on the date and time.

[0205] Time-series data has the property of changing over time. Because this time-series data fluctuates with the passage of time, reports based on figures from a month ago often become worthless. For example, technical indicators change daily, so reports analyzing figures from a month ago become meaningless.

[0206] Furthermore, much of the time-series data consists of numerical data that changes over time. Since this changed numerical data often becomes meaningless if it's only a month old, it needs to be constantly updated. However, a challenge with typical economic reports and analyst reports is that, for example, if the report was created in November 2023, the numerical data remains unchanged regardless of how much it has changed since then.

[0207] The benefits of an information generation system that automatically generates reports by promptly incorporating numerical data that changes over time are enormous. This is especially true for reports based on time-series data, which quickly become outdated, making this system particularly effective.

[0208] For example, a US economic report would include a large amount of time-series data, such as US employment statistics, the Dow Jones Industrial Average, interest rates, and exchange rates. Traditionally, economists would update various statistical data and provide a report once a month.

[0209] The system of Embodiment 4 acquires time-series data such as statistical data and numerical data, associating each time-series data with the date and time information in which it occurred, imports it into a database, and automatically generates reports, thus offering high updateability. For example, travel requires a large amount of time-series data, such as temperature, precipitation, snowfall, train and bus schedules, opening hours and closing times of various facilities, and restaurant hours. These are often subject to change due to weather. Experienced tour conductors can respond sensitively to these changes, but individual travelers and family travelers often struggle to adapt. This is because travel schedules cannot accommodate changes in time-series data such as temperature and precipitation. However, with the system of Embodiment 4, it is possible to automatically generate travel plan reports while continuously importing time-series data. Therefore, if heavy rain is forecast for tomorrow, the travel plan will be changed accordingly, and a travel plan report will be provided with updated schedules, destinations, recommendations for indoor travel destinations, etc., which is highly effective. Based on heavy rain forecasts, and considering how time-series data changes, the system automatically generates reports by performing vector searches to find travel plans that can be enjoyed even in the rain, as well as facilities and tourist spots that can be enjoyed even in the rain.

[0210] Reports containing time-series data must adapt to changes in the numerical data within that time series. This presents a significant challenge, as it is extremely time-consuming and laborious when done manually. For example, travel plans need to be significantly revised depending on whether it's raining or sunny, and stock reports need to change from buy recommendations to sell recommendations based on changes in stock price levels and technical indicators.

[0211] With the system of Embodiment 4, the content of the report itself can be automatically changed (without human intervention) in response to changes in time-series data. For example, in response to a change in time-series data, such as the US Federal Reserve (FRB) shifting to raising interest rates, it becomes possible to change the chapter structure to include sections such as the impact of the FRB's interest rate hikes, the impact on exchange rates, past stock prices during interest rate hikes, and the impact on exchange rates, and generate text according to each chapter.

[0212] Furthermore, changing text in accordance with changes in time-series data is extremely labor-intensive for humans. Even a single economic report will have vastly different content and chapter structure depending on whether it's during an interest rate hike phase or a rate cut phase, and a travel planner will also have significantly different text depending on whether it's a rainy day or a scorching hot day. Stock reports will also have vastly different text content and chapter structure depending on whether they are buy or sell reports. There are functions to match the text content by including chapters on the importance of profit taking and stop-loss in the case of sell reports.

[0213] In the system of Embodiment 4, there are cases where changes in numerical data are automatically captured and reports are automatically generated, and cases where reports are automatically generated that include changes in chapters and text data that change due to changes in search results over time. However, there is a close relationship between the numerical data of time-series data and the text data in the report. This relationship can be expressed as a function, making it possible to relate changes in numerical data to changes in text.

[0214] Furthermore, time-series data often changes as numerical data. This changed numerical data often has meaning and is frequently paired with text. For example, with temperature, there is often text that corresponds to the numerical data, such as "dangerous" if it exceeds 40 degrees and "beware of frost and freezing" if it falls below 0 degrees. By turning this into a function, it becomes possible to generate a warning text if the temperature exceeds 40 degrees. In the case of technical indicators, one example is a buy signal if the RSI falls below 30%. Any example where ~(change in numerical data) leads to ~(change in text content) can be turned into a function.

[0215] By not only providing numbers but also the meaning of the changes in those numbers, and passing this information as text to the information generation module (information generation module) 40 or the information generation AI service, it becomes possible to manage the response results based on the magnitude of the numbers and ensure accurate interpretation. This has a significant effect. It also makes it possible to freely generate text content that is free from misunderstandings.

[0216] For example, regarding temperature, in response to the question "What's the weather like this week?", it's possible to identify and warn about the days and times when heatstroke is likely. Similarly, in financial reports, if market interest rates (e.g., the yield on 10-year government bonds) significantly exceed policy interest rates (official discount rates or Fed interest rates) (based on changes in time-series data), a report could be automatically generated that informs the date of the next Fed or Bank of Japan policy meeting and warns of the need to pay close attention to significant market changes. This is possible by pairing numerical data with text data and changing the text data in response to changes in the numerical data.

[0217] The system of Embodiment 4 can automatically generate various reports, including financial reports, analyst reports, economic reports, and industry reports. For example, if reports are generated based on financial data and statistical data, the system can automatically generate financial reports and economic reports. If reports are generated based on securities reports, earnings presentation materials, and stock price data, the system can automatically generate analyst reports. If reports are generated based on travel data such as timetables and local weather, the system can automatically generate travel reports. If reports are generated based on care data such as long-term care insurance and care facilities, the system can automatically generate care reports. If reports are generated based on advertising data such as access data, ad types, and ad results, the system can automatically generate advertising reports. If reports are generated based on data such as customer purchase history and login history, the system can automatically generate customer reports.

[0218] The nature or type of report automatically generated by the system is determined by the data stored in the system's memory. Various information can be retrieved from the memory using keyword searches, vector searches, SQL searches, and, including internet searches, it becomes possible to answer a wide range of questions. Furthermore, by using functions, numerical data can be converted into text data, and by passing this data to the information generation module 40 or the language information generation AI service for summarization and evaluation, it becomes possible to automatically generate various types of reports.

[0219] The information generation system has a unique effect: it changes significantly depending on the data, text, and files stored in the memory area. For example, in the case of a caregiving report, it is not easy for a man in his 40s facing the challenge of caring for his parents to research what to do and how to do it. It is not easy to find information on local care facility fees and facilities, care level certification, city hall offices, procedures, and documents. The benefit of automatically generating a report tailored to the user by inputting these variables (parent's age, children's education level and family structure, region) is immeasurable. In this case, the memory area would store PDF and Excel files from city and ward offices, as well as summaries of legal revisions and the long-term care insurance system, and brochures for care facilities. These are then vectorized to enable vector searches. Questions are vectorized for similarity searches. Keyword searches and database searches are performed, and the search results are sent to the information generation module 40 and the language information generation AI service. The information generation module 40 and the language information generation AI service comprehensively summarize, summarize, and list the information, outputting it by chapter theme. All kinds of reports are automatically generated through this process.

[0220] Furthermore, in the case of stock reports, there are various factors that indicate buying, selling, or waiting, and the challenge lies in how to make a final decision. For example, technical indicators may show a buy signal, but the company's earnings show expanding losses, making a final decision requires a sophisticated judgment. The information generation system makes these individual judgments and search results using functions and various searches, and passes these multiple results to the information generation module 40 and the language information generation AI service, giving them the role of making an overall judgment, thereby deriving a unified, rather than disparate, result.

[0221] The reports automatically generated by the system are preferably structured in chapters, and each chapter may contain paragraphs. In this case, for example, each chapter corresponds to a single theme or issue, and vector searches, keyword searches, database searches, internet searches, etc., are performed for each chapter. Based on these search results, the results are passed to a function or to the information generation module 40 or the language information generation AI service, thereby forming one or more paragraphs that solve the issues for each chapter.

[0222] The results of solving problems in each chapter are ultimately sent to the information generation module 40 and the language information generation AI service, which then comprehensively evaluate these results and assign roles to determine what actions to take and how to make comprehensive decisions, thereby providing the user with a comprehensive judgment. Here, prompts play a crucial role; in the case of travel, for example, assigning the role of an excellent tour conductor enables sophisticated decision-making.

[0223] The ability to automatically generate various reports, depending on the stored data, chapter structure, and role prompts given to the information generation module 40 and the language information generation AI service, brings immeasurable benefits. For example, reports suited to various purposes, such as stock reports, economic reports, travel plans, care plans, advertising reports, and customer reports, can be automatically generated.

[0224] A typical stock report would be structured into chapters such as company overview, recent trends in company performance, industry trends, market share, segment trends, current stock price position, technical indicator trends, trend analysis, and overall evaluation. The issues for each chapter of the stock report are then analyzed using various searches and functions to output results. These results are then passed to the information generation module 40 and the language information generation AI service, which act as a skilled securities analyst (expert) to organize the overall structure of the stock report and generate the final report.

[0225] Time-series data changes numerically over time. One of its features is the ability to capture these changes in numerical data and generate reports accordingly. With conventional technologies, capturing these changes in time-series data requires constant monitoring, and even if changes are detected, it is often difficult to generate reports immediately.

[0226] Furthermore, the data stored in memory is constantly being updated. Therefore, it can capture various changes. One of its features is that you can freely set the rules for these changes. For example, you can monitor time-series data such as technical indicators calculated from stock price data and capture the relevant changes. It becomes possible to trigger an action when the price enters a buy zone and buy signals are lit up in three indicators.

[0227] Catching various changes is extremely difficult. The benefits of automating the detection of these changes and automatically generating reports tailored to those changes are immeasurable. For example, if stock A, which previously had a neutral rating regarding buying or selling, undergoes a significant change and the rating shifts to buying, the ability to automatically generate a report that immediately captures the reasons, background, changes in technical indicators, and changes in trends is a huge advantage.

[0228] Furthermore, for example, customer reports can be used to detect triggers such as a customer reaching the cancellation page. Based on this, a report on the customer's reasons for cancellation is automatically generated. The report is automatically generated and includes chapters containing customer attribute data (date of birth, phone number, address, etc.), customer purchase history (purchase date, frequency, number of months of continuous use, etc.), and customer phone call history. Based on this report, measures to prevent cancellation can be taken.

[0229] Furthermore, there are various types of technical indicators. Granville's rules, for example, are well-known, but they have eight buy / sell patterns and complex conditions. For instance, a buy signal is generated when the price breaks above the moving average after it has fallen, then leveled off or turned upward. However, few investors can utilize such signals, and managing them is extremely difficult. This becomes even more challenging when multiple indicators are involved. Information generation systems can generate results by defining such complex conditions as functions.

[0230] Furthermore, by passing these generated results to the information generation module 40 and the language information generation AI service, the ability to create a comprehensive evaluation report that takes into account multiple responses provides a significant benefit to the user. For example, stock price data is retrieved from memory, various moving averages are calculated based on that data, and the positional relationships and changes of these multiple numerical data are calculated to determine whether it meets conditions such as Granville's Law. Based on this determination, the information generation module 40 and the language information generation AI service make an overall judgment. For users, using changes in indicators that include multiple conditions such as Granville's Law to help with investment is very cumbersome and has a high barrier to entry. With such an information generation system, calculations and judgments are performed automatically, and a report is automatically generated after comprehensively diagnosing these results, which has the special effect of allowing users to immediately refer to it by reading it.

[0231] Specific examples include calculating and determining conditions for technical indicators such as RSI, Granville's Law, and Bollinger Bands, as well as measuring and determining trends in price movements over the past month. In addition to calculating and determining stock price data, it can also be used to calculate and determine financial data (for example, the divergence between policy interest rates and market interest rates). In advertising data, it can be used to calculate when the conversion rate falls below a certain level, determine the level of care needed, and calculate the arrival times of buses and trains at any given time.

[0232] Generally speaking, a report is a document or statement that presents information in an organized format for a specific audience and purpose. It may include images. It can be in written form as well as audio. It can also include essays and papers, and while it is usually composed of paragraphs, headings, and chapters (including the concepts of sections and subsections), these are not essential requirements for a report.

[0233] In particular, time-series data reports are reports that summarize data observed over time (time-series data). Time-series data reports require especially high update speed. The faster the updates, the better, and the immediate reflection of data brings immeasurable benefits.

[0234] To capture changes in time-series data, methods such as memory updates, API integration, and crawling are used. Based on the changed numerical and text data, various searches and calculations using functions are performed, and the results are returned to the information generation module 40 and the language information generation AI service, thereby realizing an information generation system with excellent time-series data update capabilities.

[0235] The term "report" in this specification encompasses a wide variety of documents, including reports, news articles, and other similar materials. The ability to hierarchically structure chapters into sections and subsections is highly effective in generating more targeted texts and paragraphs. Chapters can also be rephrased as sections, headings, or subsections. Reports include articles containing chapters and paragraphs, regardless of format. Articles and news articles with headings are also included, regardless of whether they have headings or chapters. It also includes audio, and even audio delivered by robots.

[0236] Furthermore, the type of time-series data is not restricted. For example, data that changes over time, such as the number of orders per month, the total sales per day, the average temperature per hour, the daily temperature, monthly company sales, the country's GDP per year, daily stock price data, quarterly company performance data, minute-by-minute train timetables, the rate of revision of nursing care fees which changes as it occurs, and daily advertising expenses, are all considered time-series data in this specification.

[0237] Let's touch upon the hierarchical structure of chapters, sections, and subsections (major headings, subheadings, minor headings). For example, if we make the overview of Company E's segment information a chapter (major heading), the most recent segment financial results a section (subheading), and the latest profitability for each segment a subsection (minor heading), the hierarchy gradually deepens. For example, one method of narrowing down a keyword search would be "stock E" AND "overview of segment information" AND "most recent segment financial results" AND "profitability." With a vector search, sentences similar to the profitability of stock E's segment information would be retrieved. Alternatively, the latest profitability for each segment could be retrieved as a search result from the quarterly financial results database of stock E.

[0238] This information can be compiled and passed to the information generation module 40 or the language information generation AI service so that the information generation module 40 or the language information generation AI service can make decisions based on prompts.

[0239] Stock reports can be adapted for other financial reports. Identifying stock codes can be replaced with country or region codes, and technical and trend indicators can be replaced with economic indicators such as GDP and employment statistics, thus creating a macroeconomic report. Replacing stock codes with industry codes or theme IDs, technical indicators with industry market share, and trend indicators with industry market size can create an industry report. Replacing stock codes with mutual fund codes and technical indicators with mutual fund performance creates a mutual fund report. The same applies to ETFs, FX, and cryptocurrencies.

[0240] In addition to financial reports, travel plans can also be modified by changing stock codes to countries or regions, technical indicators to weather indicators (such as precipitation and temperature), trend indicators to train and bus timetables, and performance trends to restaurant and tourist facility information.

[0241] The information generation system of Embodiment 4 will be described in more detail below. In the information generation system of Embodiment 4, for example, if user G makes a request to the information generation system saying, "I want a US economic report," the information generation system first collects past US economic reports that are as suitable as possible to the user G's request trends from external information sources published on the internet. The first information regarding the collected US economic reports is stored in the collected information storage unit 72 of server 3.

[0242] However, the collected US economic reports are often created a month or even a few weeks ago. Therefore, the information generation system analyzes the collected US economic reports to identify time-series data that may have already changed, such as US employment statistics, the Dow Jones Industrial Average, interest rates, and exchange rates. The information generation system then collects the latest time-series data, such as the latest US employment statistics, the Dow Jones Industrial Average, interest rates, and exchange rates, from the internet and stores this latest time-series data as second information for the US economic report in the collected information storage unit 72 of server 3. The collected recent time-series data is collected in association with the date and time information of when each time-series data was generated and stored in the collected information storage unit 72.

[0243] Subsequently, the information generation system reads the first and second pieces of information concerning the US economic report from the collected information storage unit 72 and generates a third piece of information concerning the US economic report by combining the first and second pieces of information. The generated third piece of information, along with user identification information that identifies user G and date and time information, is stored in the response information storage unit 74 within the storage unit 3C.

[0244] There are various methods for combining the first and second pieces of information. For example, one method involves retaining the time-series data from the first US economic report created on January 1, 2024, such as US employment statistics, the Dow Jones Industrial Average, interest rates, and exchange rates, as they were in the original US economic report. Then, the second piece of information from the US economic report, which contains the latest time-series data collected on April 1, 2024, such as US employment statistics, the Dow Jones Industrial Average, interest rates, and exchange rates, is added in parentheses in the appropriate places to generate a third US economic report.

[0245] It is desirable to include in the third U.S. economic report the date and time information for the first U.S. economic report and the date and time information for the second information collection.

[0246] The third US economic report will be as follows: "US Employment Statistics as of January 1, 2024: Unemployment Rate 4.1% (3.8%), Dow Jones Industrial Average: 40,398 dollars (41,235 dollars), Interest Rate: 3.0% (3.5%), Exchange Rate: 145 yen against the dollar (151 yen) *Note: The numbers in parentheses are data as of 9:00 AM JST on April 1, 2024."

[0247] Another method of combining the first and second pieces of information is to replace the time-series data in the first piece of information with the time-series data in the second piece of US information. In this case, it is desirable to include the date and time information of when the first US economic report was created and the date and time information of when the second piece of information was collected as a note in the third US economic report, and to clearly state that each piece of time-series data has been updated to reflect the date and time when the second piece of information was collected.

[0248] Beyond time-series data such as statistical and numerical data related to fluctuating products, travel information generation systems, for example, require a large amount of time-series data to generate travel information, including weather information, temperature, precipitation, snowfall, train and bus schedules, opening hours and closing times of various facilities, and restaurant hours. In travel, itineraries are often forced to be changed due to the weather. Individual and family travelers often struggle to cope when faced with changes in weather. This is because travel schedules cannot automatically adapt to changes in time-series data such as weather, temperature, and precipitation.

[0249] However, according to the information generation system of Embodiment 4, it is possible to automatically generate and update travel-related information such as travel plans by continuously, appropriately, and regularly incorporating time-series data such as weather information, temperature, precipitation, snowfall, train and bus timetables, closing and opening hours of various facilities, and restaurant opening hours. For example, if the weather forecast for tomorrow predicts heavy rain, the travel plan will be changed to suit that weather, and the schedule, destinations, and recommendations for indoor travel spots will be modified accordingly, which has a significant effect. This is effective when the information processing module analyzes the information collected by the information collection module and extracts trigger information such as "Tomorrow's weather forecast: Heavy rain."

[0250] The trigger module 30 of the information generation system constantly checks the information collected by the information collection module 10B. The trigger module 30 is activated when it identifies abnormal information, dangerous information, or fortunate information among the collected information, and sends the identified abnormal information, dangerous information, or fortunate information, along with the overall information of the collected information, to the information generation module 40, causing the information generation module 40 to generate countermeasures information.

[0251] For example, trigger module 30 is activated by a heavy rain forecast extracted by information gathering module 10B. Information gathering module 10B then performs a vector search for travel plans that can be enjoyed even in the rain, as well as facilities and tourist spots that can be enjoyed even in the rain. The collected information is sent to information generation module 40, or to an AI information generation service (ChatGPT, Midjourney, AITalk, etc.) located on the internet outside the information generation system. A report information, which is a new travel plan adapted to the heavy rain weather, is then automatically generated.

[0252] [Examples] Next, embodiments of the present invention will be described with reference to Figure 4. Figure 4 is a flowchart showing the processing flow in the system according to the embodiment.

[0253] When a specific user, such as an investor, sends a question to the server 3 via LINE or their My Page using the user terminal 2, this question information is received by the interface module 50 of the server 3 (S1). The server 3 stores the received question information as user-specific information in the storage unit 3C and processes the information to collect investment-related information based on that question information (S2). For example, the information processing module 20 of the server 3 can perform a keyword search by breaking down the specific user's question information by keyword, or perform a vector search on the specific user's question information itself. The search scope may include, for example, information on the internet (information from external sources), or it may include information from internal sources such as databases and files stored in the storage unit 3C of the server 3 (for example, the database storage unit 77).

[0254] Furthermore, the information gathering process (S2) also includes collecting the latest numerical data (time-series data), such as the stock price of an investment target (e.g., Company X) identified from the question information, from external and internal sources, and calculating various indicators, including technical indicators.

[0255] Next, the information processing module 20 of server 3 integrates the collected information (search results and calculation results) and executes a process to generate prompt information to be passed to the information generation module 40 or the information generation AI service of an external server (S3). For example, when generating prompt information to be passed to an information generation AI service (a large-scale language conversation type AI service such as ChatGPT), the information processing module 20 generates text (prompt information) based on the collected information (search results and calculation results), for example, taking on the role of a securities analyst, which instructs the generation of text to provide answers to questions from a specific user or to make investment decisions regarding stock investments. At this time, the report structure, such as the number of chapters (headings), the number of characters in each chapter, and the total number of characters, may also be instructed.

[0256] By inputting the prompt information generated in this way into the information generation module 40 or the information generation AI service of an external server (S4), a report for answering the specific user who asked the question is generated (S5). Thereafter, the server 3 transmits the generated report to the specific user who is the questioner as answer information through the interface module 50 (S6).

[0257] So far, the description has been made based on FIG. 4. Now, a case where an S1b step is provided between the S1 step and the S2 step in the flowchart will be described. Based on the question information from the specific user in the S1b step, it is also possible to generate prompt information for causing the information generation module 40 to collect public information published on the Internet 4. In this case, step 2 becomes "information collection based on prompt information". The information collection module 10B collects the necessary public information from the public information sources on the Internet 4 for the information generation module 40 based on the prompt information. Thereafter, the process proceeds to step 3 in FIG. 4.

[0258] FIG. 5 is an explanatory diagram for explaining an example of the structure of an automatically generated report. The illustrated example is an example of a report automatically generated by performing the processing along the flowchart shown in FIG. 4 when a specific user asks a question "What is the recent situation of Company X?".

[0259] In the report example of FIG. 5, first, as the first major item, "(1) Report Summary" is described. This "Report Summary" is a part where the current share situation and technical trend analysis of Company X are described in text, and it is divided into six chapters: "1. Stock Price and Trend", "2. Evaluation of Technical Indicators", "3. Volatility and Market Trends", "4. Business Expansion and Future Possibilities of Company X", "5. Overview of Financial Situation", and "Conclusion", and the text, charts, etc. corresponding to each title are described. The chapter structure of the "Report Summary" and the amount of text in each chapter, etc., are automatically generated by the information generation module 40 for each user based on user-specific information such as the user's desired information.

[0260] Furthermore, the next major section describes "(2) Results of the Questions." In this example, a specific user asked the question, "What is the current situation of Company X?", so the report is divided into three chapters: "Strengths and Challenges of Company X's Main Business," "Technological Innovation and Future Perspectives of Company X," and "Business Environment and Risk Management of Company X." Each chapter contains text, diagrams, and other relevant information. The text content of this major section corresponds to the specific user's question information (user-specific information), and therefore will differ for each user. In particular, if user-specific information such as the specific user's attribute information is also taken into account when generating the report, even if the question information is the same, a report with different text content will be generated for each user who asked the question. In addition, the chapter structure and the amount of text in each chapter will also be changed for each user based on user-specific information such as user requests.

[0261] The final major section is titled "(3) Detailed Report." This section is divided into seven chapters: "1. X Company's Product Development Strategy and Market Launch," "2. Strategic Deployment in the Global Market," "3. Financial Status and Future Revenue Outlook," "4. Technological Innovation and Future Outlook," "5. Industry Positioning and Brand Enhancement," "6. Current Assessment from an Investment Perspective," and "7. Conclusion and Future Strategic Guidelines." Each chapter contains text appropriate to its title. This major section also includes further chapters such as "Technical Analysis Results," "Trend Analysis," "Basic Price Data," "Moving Average Data," "Detailed Technical Analysis," "Summary of Search Results," and "Analysis of the Latest Trends in the X Company Group," each containing text and charts appropriate to its title. Finally, this major section concludes with "Conclusion" and "Search Results."

[0262] Furthermore, the report example in Figure 6 includes "technical signal count values" as investment advice on how to make investment decisions based on the values ​​of technical indicators, representing the results of judgments based on technical indicators. Here, "technical signal" refers to the judgment result, evaluation result, or information or indicator that shows the status of the technical indicator, such as whether to buy, sell, or hold the stock of the company being judged, based on the numerical data of the technical indicator. A technical indicator that results in a judgment of buying is a "buy signal," a technical indicator that results in a judgment of selling is a "sell signal," and a technical indicator that results in a judgment of holding is a "neutral signal." Normally, it is difficult for users to make appropriate judgments by looking at the values ​​of these technical indicators, but according to this example, users can easily obtain judgment results and evaluation results based on the values ​​of the technical indicators.

[0263] Traditionally, it has been difficult to obtain appropriate judgments from the values ​​of individual technical indicators, and even more difficult to obtain appropriate judgments by combining the values ​​of multiple technical indicators. For example, as shown in Figure 6, it is advisable to display the count value for each of the five technical indicators, which is the result of counting the judgments derived from the values ​​of each technical indicator for every three signals. This makes it easier for users to obtain appropriate judgments by combining the values ​​of multiple technical indicators. For example, by looking at the "Technical Signal Count" in the report, they can determine that they should buy the company's stock by confirming, for instance, that there are many technical indicators indicating a "buy signal."

[0264] Furthermore, the example report in Figure 5 includes an image of the stock price chart of Company X, which is relevant to the question, as an illustration showing the temporal changes in time-series data or the numerical values ​​of technical indicators. Automatically generated reports may include not only text but also such charts and graphs.

[0265] The report generated by the information generation system, as illustrated in Figure 6, includes customized text based on specific numerical values ​​of technical indicators, such as investment advice on how to make investment decisions based on the numerical values ​​of technical indicators (e.g., whether to buy, sell, or hold the stock in question), and explanatory text (documents) based on at least one technical indicator, such as explanatory text, descriptive text, and answers to questions. The report may also show specific technical indicators or trend indicators for a particular stock and explain or describe that stock. Of course, the information generation system can also output information explaining, describing, or answering questions about a particular stock without showing specific technical indicators or trend indicators to an information storage device such as a printer or hard disk drive, send it to the user's or questioner's email address, or display it on the user's or questioner's terminal 2. Such text may be displayed in chapters within the main body of the report, or, for example, as shown in Figure 6, it may be displayed as a pop-up 81 when the cursor 80 is placed over the stock price chart of company X. In this example, customized text is displayed based on the values ​​of technical indicators calculated at the point in time corresponding to the cursor's position on the chart. For example, it is possible to check whether X company's stock was a buy or a sell at a past point in time.

[0266] Furthermore, the example report in Figure 5 includes a section for supplementary information. By including such a section for supplementary information, it is possible to, for example, reduce the amount of text in the main body of the report to make it easier to read, while also making it easier for users to access information they may want to check, such as URL information or sources that identify the basis of the information in the main body of the report.

[0267] Next, based on the first prompt information, the information generation module 40 collects information from publicly available information on the internet and other sources using the information collection module 10B, and obtains the first collected information. The first collected information includes information such as that there is a business partnership between Company X and Company T, and that Company X is producing batteries for Company T at a gigafactory in another state N.

[0268] The first collected information gathered by the information gathering module 10B is stored in the collected information storage unit 72. Next, the information generation module 40, based on the second prompt information, causes the information gathering module 10B to perform a second information gathering operation from information publicly available on the internet, thereby obtaining the second collected information. The second collected information includes information not included in the first collected information, such as the names of X Company's business partners, "Y Company" and "W Company," the state of "M" where X Company's lithium-ion battery factory is located, and "subsidies from M State."

[0269] The second set of collected information, gathered by the information gathering module 10B, is stored in the collected information storage unit 72. Subsequently, the information generation module 40 reads both the first and second sets of collected information from the collected information storage unit 72, combines them, and generates the third set of collected information. The third set of collected information is stored in the collected information storage unit 72.

[0270] The third set of collected information is more information-rich than the first and second sets of collected information, but it may be too much information or too redundant. Therefore, the information generation module 40 can create a fourth set of collected information by removing redundant information from the third set of collected information and summarizing it. For example, if a specific user requests information of about 250 characters, the information generation module 40 will generate a fourth set of collected information of about 250 characters.

[0271] An example of the fourth type of information to be collected is as follows: "Company X's automotive battery business has partnered with Company T, leveraging its lithium-ion battery technology. It produces batteries for Company T at its Gigafactory in State N, giving it massive production capacity. Furthermore, Company X's automotive battery business is growing through strong partnerships with Companies Y and W. Utilizing subsidies from State M and the federal government, it is constructing a factory in State M and establishing a system to meet the future needs of the EV market. As a result, Company X is expected to play a crucial role in the advancement of electrification."

[0272] The fourth collected information is stored in the collected information storage unit 72. The information processing module 20 transmits the fourth collected information generated by the information generation module 40 to the user terminal 2 of a specific user as answer information to the specific user's question information via the interface module 50 and the communication unit 3A. The fourth collected information generated by the information generation module 40 is stored in the answer information storage unit 74 within the storage unit 3C, along with user identification information that identifies the specific user and date and time information.

[0273] Furthermore, the information generation module 40 confirms the potential for further utilization of the answer information generated in response to a specific user's question. Specifically, the information generation module 40 analyzes the individual user information (such as user attribute information) of all users stored in the memory unit 3C to find other specific users who need the answer information generated in response to a specific user's question.

[0274] For example, the information generation module 40 finds another specific user from the user information storage unit 71 in the storage unit 3C who owns 5,000 shares of Company W and has a strong interest in electric vehicles. In this case, the information generation module 40 generates information either as is, or by adjusting the response information sent to the specific user to match the other specific user's preferences in terms of information volume and format. The generated information for the other specific user is then sent to the other specific user's user terminal 2 via the interface module 50 and the communication unit 3A.

[0275] Instead of collecting information solely from natural language-based question information created by each user, the system also performs vectorization (vector processing) of specific users' natural language-based question information to generate related, supplementary, and similar information that is not present in the question information created by those specific users. By considering this related, supplementary, and similar information, the information collection module 10B collects information from the internet and other sources, thereby interpreting the meaning of the natural language-based question content and collecting a sufficient amount of raw data for responses.

[0276] Then, the collected raw data for answers is converted into a predetermined number of characters and format, retaining important information while removing redundant information, by the information generation module 40 or the information generation AI service systems 7A to 7E, which are information generation means, to generate answer information for the specific user who asked the question. An information generation system that has a question information analysis means and generates answer information to provide to a user who performs information retrieval by supplementing, extending, and similarly enhanced question information of a specific user's natural language question information can overcome the limitations of keyword search, where search results do not appear unless they match the keywords.

[0277] The following is a further description of the embodiments. In some embodiments, an AI-based interactive financial information provision system and method may also be provided. This system and method utilizes AI to search and analyze financial data and investor relations materials and provides personalized information through interaction with the user. In particular, it relates to an interactive system that provides information in response to price fluctuations of financial products such as stocks, mutual funds, forex, and cryptocurrencies, and that supports automated trading and risk management.

[0278] In the financial industry, providing personalized information to individual investors has been limited in conventional systems. Many information - providing systems offer information uniformly to all users and are unable to respond flexibly according to the individual needs of users. As a result, investors, who are users, have to select a vast amount of information based on their own judgment, often leading to a decrease in the efficiency of investment decisions. Furthermore, in dialogue - based systems using platforms such as LINE and personal pages, it is difficult to take into account the situations and past interactions of individual investors, and improving user satisfaction has been an issue.

[0279] To solve the above problems, a system is provided that searches for and analyzes financial data and IR materials, and uses AI to provide information according to the investment styles and interests of each user. In particular, it realizes personalized report generation, advice, and automated trading based on the user's dialogue history and investment behavior, and improves the efficiency of investment decisions and report provision. Also, a trigger system is constructed according to the individual needs of users, and by equipping it with a function to automatically resume the dialogue when specific conditions occur, it supports the investment behavior of users.

[0280] An example of a stock price movement clustering system will be described. The price movements of stocks vary for each stock and each period. It is not easy to fit into a certain pattern. However, it is possible to classify them into groups with somewhat similar movements, groups with completely opposite movements, etc., into groups that are somewhat similar. Clustering refers to classifying a dataset into several groups (clusters) based on specific rules. In particular, a method of grouping similar things based on the similarity between data is typical, and it is possible to realize a system that clusters price movements by utilizing one of the "unsupervised learning" in machine learning. Here, classification by supervised learning and manual classification methods are also included in this clustering system.

[0281] For example, you could divide a period and collect stock price data for all listed stocks over a one-year period, then determine the features. For stock price movement clustering, the features could include the type of trend for each period, the strength of the trend, and the deviation rate from the moving average. For example, you could divide the collected data into 20 groups and cluster or classify them. By labeling and naming the features of each clustering result and adding explanations, you can quickly understand the characteristics of the price movements of that group.

[0282] Understanding the price movement characteristics of the stocks one holds and identifying groups that exhibit similar price movements brings various special benefits to investors. This includes the ability to choose investment targets with different price movements rather than those with similar movements, or to make optimal selections for better investment targets within the same cluster, resulting in significant advantages for investors.

[0283] As a concrete example, suppose there is a cluster of stocks labeled as "stocks with a strong upward trend," where the deviation rate from various moving averages remains positive over the long term, and the upward trend has continued for more than a year with an average monthly increase rate exceeding 5%. There are 20 stocks in this cluster, and it is possible to cluster them such that construction stocks, telecommunications stocks, and real estate stocks make up a large proportion. By performing clustering using machine learning, it is possible to cluster from a large number of features. Principal component analysis and other methods can be used to extract particularly important features from among many features.

[0284] This section describes an example of a metadata search system for presentation materials and financial statements. Financial statements such as presentation materials and securities reports contain important information for understanding a company. Processing this information and making it easily understandable requires manpower and effort. However, by converting this information into metadata such as file name, creation date, creator, summary, chapter structure, chapter titles and keywords relevant to each chapter, and paragraph summaries, and storing it in the information storage unit 72 as a text-based JSON (JavaScript® Object Notation) file, it becomes possible to search paragraphs (blocks of text) using keyword searches.

[0285] Metadata is "data about data" used to describe other data. For example, PDFs internally record information such as the creator and last modified date. Metadata is defined as data such as "when" and "who" created it, as well as its contents such as "headings," "page numbers," "page titles," and "paragraph summaries." Structuring this PDF data with metadata enables search results that cannot be obtained with simple full-text searches.

[0286] This service converts PDF data, text documents, and presentation materials into text and extracts metadata information. Metadata can be extracted by company, file, page, and paragraph. Financial results presentation materials, for example, often have inconsistent formats and are unstructured, making them difficult to understand. Such unstructured data often doesn't yield satisfactory results when stored in databases or subjected to full-text searches because the meaning of the text isn't captured. By converting unstructured data into metadata and creating a text-based JSON file, it becomes possible to search by title, summary, and other criteria.

[0287] Even if a keyword search yields results, humans need to read the surrounding text to determine if the content answers their question. When keyword searches are performed using metadata, if the results match the page's key keywords and headings, the page can be considered highly important. Furthermore, more recently created information can be weighted, as metadata such as creation and release dates are also included.

[0288] For example, in response to a question such as, "Please tell me about the operating status and operational status of F Airlines' domestic flights," it would be possible to search for the title of the most recent earnings presentation materials in F Airlines' investor relations (IR) documents, identify the page that explains the operating status and operational status of domestic flights, and then display a summary of the information on that page.

[0289] This section describes an example of a vector search system for presentation materials and financial statements. It is well known that stock prices and corporate performance are closely related. However, interpreting securities reports and other documents requires a high level of expertise in accounting and other fields, making it difficult to fully utilize them. Extracting information crucial for investment from a vast amount of data is not easy. There are 3,800 Japanese stocks alone, and compiling quarterly financial statements, business-specific briefing materials, and monthly reports, which are updated every three months, is an extremely difficult task.

[0290] The presentation materials are downloaded from each company's IR (Investor Relations) materials. The metadata information of these materials (file name, release date, page title, summary of page content, etc.) is stored in a database as a JSON file. The stored JSON files are then vectorized and quantified using methods such as BERT. The questions and triggers are also vectorized. Based on the vectorized questions, a vector search is performed on the stored JSON files, similarity is calculated, and the vectorized data is converted back into text format to obtain the search results for presentation materials relevant to the questions.

[0291] User questions are diverse, encompassing variations in wording, typos, omissions, and usage, making it extremely difficult to capture (and display accurate search results for) these issues using keyword or full-text searches. Unlike keyword searches, vector searches understand the meaning of the text before searching, resulting in a special effect that allows for the display of search results directly related to the question. In particular, the ability to perform cross-sectional searches by creating metadata from IR materials such as presentation materials and utilizing vector searches is extremely beneficial.

[0292] For example, if the question is "Regarding X Company's automotive batteries," a vector search can display results such as X Company's lithium-ion batteries, market trends for lithium-ion batteries, the operating status of their Kansas battery factory, and their relationship with companies W and Y. With a keyword search, if the term "automotive batteries" is not included, no results will appear, and the intent of the question will not be understood.

[0293] The differences become even clearer with more advanced questions. For example, a question like "Compare the current status of the copier businesses of Company R and Company P" can be compared by performing a vector search of metadata-based data containing the question information. This allows for comparison using recent financial results presentation materials and business presentation materials (the most recent materials can be identified using metadata). This has a significant effect, as it can dramatically improve understanding of the companies and the industry. Understanding the meaning of the question and the content of the paragraphs allows for a more relevant answer, which is highly beneficial.

[0294] This section describes an example of a corporate performance clustering system. Securities reports and financial statements are highly correlated with stock prices, but they are difficult for beginners to understand and approach. Therefore, after obtaining securities reports and financial statements from information sources using API integration functions, features are selected from these reports and financial statements. These features include revenue growth rate, profit growth rate, dividends, and earnings per share growth rate. By clustering based on these features, clusters with similar characteristics are generated.

[0295] Even within the same industry, it's possible to cluster companies based on their different features, strengths and weaknesses within the industry, connections with other industries, and similar trends. For example, there might be 30 stocks belonging to a cluster with strong revenue drivers, such as revenue growth of over 30%, profit growth of over 30%, and increased dividends resulting in an operating profit margin of over 30%. Another concrete example is a cluster with revenue growth of over 30% but reduced profits due to increased costs, resulting in an operating profit margin of 20% or less, or having segments with a loss-making structure. Depending on how features are selected, various clusters are possible, such as a cluster with five consecutive quarters of increased revenue and profit, an improving operating profit margin, a stable advertising ratio, and increasing personnel costs that are absorbed by sales.

[0296] One method for selecting features is to determine their correlation with stock prices using correlation coefficients. One approach is to calculate the correlation coefficient between features such as revenue growth rate, dividend growth rate, and profit growth rate and stock prices, and then select the features that show a correlation. By using figures that are highly correlated with stock prices within a company's performance as features, clustering based on company performance can be performed, which has the special effect of making it easier to identify which cluster a portfolio of stocks belongs to and which stocks to focus on.

[0297] This section explains the extraction, explanation, and classification methods (numerical cases) for trend analysis. From the stock price moving average and closing price of a specific stock, it is possible to calculate features such as the moving average deviation rate and the positional relationship between the moving average and the closing price. These features are used to calculate the trend of stock price fluctuations for each period in accordance with the stock price fluctuations, which are time-series data. Based on the calculated results, explanatory results (for example, if the trend over the past month has been a recovery trend, but over a period of six months it has been a strong downward trend, an explanatory result indicating that a sell signal for a rebound has emerged) are passed to an information generation AI such as ChatGPT, which performs natural language processing to generate text data that can be used as one of the factors for decision-making.

[0298] Using methods such as the K-means algorithm, we identify clusters with common characteristics from features used to analyze various trends. For example, we might classify stocks with similar tendencies into Cluster 1 if they are in an upward trend with a strong upward trend that has continued for a long time (both the 5-day and 25-day moving averages are on a strong upward trend, the current price is above the 5-day moving average, and the deviation is low).

[0299] The k-means algorithm is so named because it creates k clusters of any specified number. The GPT in ChatGPT stands for "Generative Pre-trained Transformer." This is a type of artificial intelligence designed to understand and generate human-like text based on the input it receives. Natural Language Processing (NLP) is a technology that enables computers to interpret and appropriately understand the meaning of human everyday conversation (natural language). It is a field of artificial intelligence (AI) and is used in chatbots, speech recognition AI, character recognition AI, and more. A moving average is a technique for smoothing time-series data. It is used in a wide range of technological fields, not only in digital signal processing such as audio and images, but also in finance, meteorology, hydrology, and other measurement fields. The three main types are simple moving average, weighted moving average, and exponential moving average.

[0300] This section explains extraction and classification methods (numerical cases) for corporate performance analysis. The system calculates features such as revenue growth rate, profit growth rate, dividend yield, and P / E ratio, which are indicators and figures related to corporate performance. These features are calculated in accordance with the fluctuations in corporate performance, which are time-series data. The explanatory results based on these calculations (for example, if revenue growth rate exceeds 30% but profits continue to decline, it means that the cost burden is increasing and caution is needed in investment, and the explanatory results show what is increasing by presenting the breakdown of costs) are passed to a system like GPT, which performs natural language processing to generate text data that can be used as one of the factors for decision-making.

[0301] Using methods such as the K-means algorithm, we identify clusters with common characteristics from features used to analyze various corporate performance trends. For example, one classification method might be to classify a group of companies with high performance, such as those that have achieved a 30% increase in revenue and profit for two consecutive years, and have also increased dividends for two consecutive years, into cluster 1 if the group has a low P / E ratio.

[0302] This section explains methods for extracting, interpreting, and classifying technical indicators (numerical cases). Technical indicators such as RSI, Psychological Line, MACD, Ichimoku Kinko Hyo, Stochastic, and RCI are used as features to calculate features. These features are calculated in accordance with the fluctuations of stock prices, which are time-series data. The explanatory results based on the calculated results (for example, an explanation that if the RSI value falls below 30%, a buy signal has been generated) are passed to GPT, which performs natural language processing to generate text data that can be used as one of the decision-making factors.

[0303] Using methods such as the K-means algorithm, we identify clusters with common characteristics from the features used to analyze various technical indicators. For example, cluster 1 might be defined as RSI below 30% and Stochastic Slow %D between 0 and 20%.

[0304] This section explains text analysis methods, including extraction, explanation, and classification (text-based). Text information such as company performance results, segment information, and news necessary for investment decisions is extracted and summarized according to the content of the question. This text is extracted using various search methods, identifying the stock and content based on the user's question and the trigger described later. The explanatory results based on this text information (for example, an explanation that Company X's automotive batteries are struggling in the US) are passed to GPT, which performs natural language processing to generate text data that can be used as one of the factors in determining the value of a variable product.

[0305] Let me explain the trigger system. The trigger system consists of a trigger module 30 and automatically resumes interaction with the user when specific conditions occur. For example, the trigger is activated and the user is notified when the trigger module 30 detects that a stock price has exceeded a certain threshold, a specific technical indicator has changed, or an upward revision of a company's earnings forecast has been announced. A typical example of this is when, based on company earnings data obtained through API integration, the company's forecast has been significantly revised upward and stored in the database, at which point the trigger is activated, and the company's stock is listed as a stock to watch, users who hold that stock are notified in an interactive manner, or the interaction is resumed.

[0306] Trigger timings include when data is stored in the database or tables, when files are retrieved, and when clustering results are saved. Who to notify is determined by extracting user information. How to notify includes chat systems, notifications on user pages, and email.

[0307] The information generation module (information generation module 40) will be described below. The information generation module (information generation module 40) is an information generation means that generates information about volatile products, such as stock reports, which include company overviews, financial summaries, recent news, technical indicator analysis, trend analysis, and overall evaluations, in natural language based on the analysis and commentary results from the information processing module 20. This enables the provision of information in natural language optimized for the user. The chapter structure of the information reports can be freely determined by the administrator and can be changed for each user according to the user's needs and requests (for example, more technical indicators, more company performance analysis, more industry analysis, etc.).

[0308] When generating a stock report (numerical data) containing information about a specific stock, which is a volatile asset, the specific stock is identified, the stock code is identified, and the most recent stock price data for that stock (preferably real-time stock price data) is collected from the internet, for example, from the timely disclosure information and stock price distribution service of the Tokyo Stock Exchange, using API integration functionality. Various numerical values ​​such as technical indicators and trend indicators are calculated from the collected stock price data for that stock.

[0309] Specific technical indicators include Moving Average (SMA), Exponential Moving Average (EMA), RSI (Relative Strength Index), MACD, MACD Histogram, Bollinger Bands, Stochastic Oscillator, ATR (Average True Range), etc. For example, based on stock price data collected on October 25, 2024, for a specific stock X, the technical indicators produced by the information generation module 40 were as follows: SMA: 13544.25 EMA:13395.24 RSI: 48.26 MACD: 46.85 MACD Histogram: 93.27 Bollinger Bands (3σ): 14694.20 Stochastic: 39.53 ATR: 395.80

[0310] The information generation module 40 can generate a stock report, which is explanatory information for a specific stock X, including the calculated technical indicators and their specific numerical values. However, even a general large-scale language model information generation AI like ChatGPT cannot judge, evaluate, explain, or describe numerical data that is not text data. In other words, it cannot explain or describe the meaning of specific numerical data. Therefore, it cannot generate textual information that explains or describes the meaning and situation indicated by the numerical values ​​of specific technical indicators.

[0311] Therefore, this problem can be solved by providing a conversion means for converting specific numerical data into character information in the information generation system 1 shown in Figure 3. This conversion means for converting specific numerical data into character information can be composed of multiple conditional expressions or a conversion table.

[0312] This document explains a method for converting specific numerical data into text data using the "Moving Average Deviation Rate," an oscillator-based technical indicator used to determine whether a market is overbought or oversold.

[0313] For example, suppose the information generation module 40 calculates that the 25-day moving average deviation rate, a technical indicator for a specific stock X, is 25%. Since 25% is numerical data, the information generation module 40 can include "25-day moving average deviation rate for specific stock X = 25%" in the stock report for specific stock X, but it cannot include textual information in the stock report that explains or clarifies the meaning of that numerical data.

[0314] Therefore, a conversion means for converting specific numerical data into character information is provided inside the information generation module 40 or inside the storage unit 3C. Here, we will explain the case where there is a conversion means for converting specific numerical data consisting of multiple conditional expressions into character information.

[0315] <Conversion means for converting specific numerical data into character information> Condition 1 = "25-day moving average deviation > 20%" → Text information: The market is overbought. Condition 2 = "25-day moving average deviation rate < -20%" → Text information: The market is oversold. If condition 3 = "-5% ​​< 25-day moving average deviation rate < 5%", then the textual information is: the state is normal.

[0316] Suppose the information generation module 40 calculates that the 25-day moving average deviation rate is 25%. The information generation module 40 inputs this specific numerical data, 25% 25-day moving average deviation rate, into the aforementioned conversion means and obtains the textual information "overbought," which indicates the meaning of this specific numerical data. As a result, the information generation module 40 becomes able to recognize that "the stock price of specific stock X has a 25-day moving average deviation rate of 25%, and the stock price situation of specific stock is overbought." Consequently, the information generation system can explain and comment on the stock price situation of specific stock X based on the specific numerical data of 25% for the 25-day moving average deviation rate of specific stock X, create a stock report for specific stock X, and provide answer information to questions from investor users.

[0317] As a result, the information generation module 40 can generate text information such as the following. "The current share price of stock X, 13,320 yen, is at a 25-day moving average deviation of 25%, indicating an overbought and overheated condition. If you own stock X, now is a good time to sell. If you are considering buying stock X, you should wait for a while."

[0318] This document, generated by the information generation system, contains textual information such as stock price information for a specific stock, specific numerical indicators related to the stock price of that specific stock, semantic and explanatory information which are primary textual information of the specific numerical indicators related to the stock price, and proposed or advice information which are secondary textual information based on the specific numerical indicators.

[0319] Furthermore, if the user information storage unit 71 of the information generation system stores information indicating that user Y's investment stance is "I want to invest with safety as the priority," the information generation module 40 can generate user Y-specific text such as the following. "The current share price of stock X, 13,320 yen, is 25% above its 25-day moving average, indicating an overbought and overheated condition. Investing in this stock now carries a very high risk, and I cannot recommend it to you, Y. Waiting until the share price of stock X falls below its 25-day moving average would be more in line with your investment strategy."

[0320] This document, generated by the information generation system, contains information such as stock price information for a specific stock, specific numerical indicators related to the stock price of that specific stock, semantic and explanatory information which is primary textual information of the specific numerical indicators related to the stock price, and suggestion or advice information based on information of a specific user which is secondary textual information based on the specific numerical indicators.

[0321] If a conversion table is used to convert specific numerical data into textual information for the 25-day moving average deviation rate, the X-axis of the table will plot the 25-day moving average deviation rate in 1% increments from -100% to +100%. The Y-axis of the table will plot textual information that represents the meaning of the numbers plotted on the X-axis. For example, the Y-axis corresponding to +25% to +100% on the X-axis will have the textual information "Overbought." plotted on it. And the Y-axis corresponding to -25% to -100% on the X-axis will have the textual information "Oversold." plotted on it.

[0322] It is desirable to provide at least one conversion mechanism for converting specific numerical data into textual information. Furthermore, it is desirable to provide as many such conversion mechanisms as possible. For example, it is desirable to provide a conversion mechanism for converting specific numerical data into textual information for each technical indicator.

[0323] Furthermore, regarding stock prices, if a conversion means is provided to convert specific difference data, such as the difference between the previous day's closing price and the current price, or the difference between the closing price one week ago and today's closing price, into textual information, the information generation system will be able to generate explanatory and descriptive information regarding stock price fluctuations as well.

[0324] For example, let's consider a case where a conditional expression is used to define a conversion method for converting specific numerical data into textual information, specifically the difference between the closing price one week ago and the closing price today.

[0325] <Method for converting numerical data to character data> Condition 1 = "Difference between last week's closing price and today's closing price > +1000 yen" → Text information: The stock price has risen significantly. Condition 2: "Difference between last week's closing price and today's closing price < -1000 yen" → Text information: The stock price has fallen significantly. If condition 3, "The difference between last week's closing price and today's closing price = 0 yen," then the text information is: The stock price has not actually changed.

[0326] Suppose the information generation module 40 calculates that the difference between the closing price of stock X last week and the closing price of stock X today is 0 yen. In that case, the information generation module 40 inputs the specific numerical data, "difference between the closing price of stock X last week and the closing price of stock X today = 0 yen," into the aforementioned conversion means that converts specific numerical data into text information, and generates text information that indicates the meaning of the numerical data, "difference between the closing price of stock X last week and the closing price of stock X today = 0 yen," such as "the stock price has not substantially changed." As a result, the information generation module 40 can create a stock report for a specific stock X that includes text information that means, explains, or describes the fact that "the stock price of stock X has not substantially changed," or it can provide answer information to questions from investor users.

[0327] As a result, the information generation module 40 can generate text information such as the following. "The closing price of stock X today, 13,320 yen, is essentially unchanged from last week's closing price. Whether you're selling or buying stock X, you should wait and see for a while."

[0328] In this context, the textual information, "Today's closing price of stock X, 13,320 yen, is essentially unchanged from last week's closing price," is primary textual information generated by the information generation system, while the textual information, "Whether you are selling or buying stock X, you should wait and see for a while," is secondary textual information generated by the information generation system based on the primary textual information. It is desirable to have multiple conversion means of different types for converting specific numerical data into textual information.

[0329] Furthermore, the conversion means for converting specific numerical data into textual information is not limited to conversion means for converting specific numerical data related to variable products into textual information. For example, it may be a conversion means for converting specific numerical data related to temperature into specific textual information, or a conversion means for converting specific numerical data related to the price of a specific product into specific textual information.

[0330] The analysis results are generated based on the numerical data of the stock in question (analysis results of numerical data). According to this, because the numerical data is based on stock price data, which is dynamically changing time series data, the numerical values ​​change dynamically over time even for the same stock. Therefore, the numerical values ​​from one year ago, one week ago, and today have all changed over time, and the text content also changes with these changed numerical values, so the report content also changes dynamically. For example, a report that included text encouraging selling when the stock price was 1000 yen and the RSI was 80% one year ago might now include text suggesting that buying is a good idea when the stock price is 500 yen and the RSI is 15%. Both the numerical values ​​and the analysis text change dynamically as the stock price changes, resulting in a report that is automatically generated that changes dynamically.

[0331] In the case of stock reports (text data), the stock is identified, the stock code is identified, and text data for that stock (e.g., presentation materials, internet search results, Vector search results, etc.) is extracted. From the extracted text data for that stock, text related to the answer to the question is extracted, and an explanatory result for the question is generated based on the answer data for that stock. As a result, the content to be conveyed (such as discussions of company performance, discussions of revenue drivers, or news necessary for investment decisions) will differ depending on the stock and the chapter extracted.

[0332] Furthermore, the content of the report can be changed according to the user's interests (user information stored in the user information storage unit 71). In other words, the response results can be modified to the user's preferences, or explanations can be added for user A regarding company performance, user B regarding revenue drivers, and user C regarding industry trends.

[0333] In the case of trigger-based stock reports, the trigger extracts stocks, identifies them, and so on, similar to the stock report generation described above. Instead of questions, explanations of the triggers, what triggers were activated, and why the report was generated will be added.

[0334] The combination of the information gathering module 10 and the information generation module 40 allows for variations in the processing of the information gathering module 10 depending on the purpose, question content, and trigger content. For example, for data containing numerical values ​​such as technical indicators, the information gathering module 10 can collect stock price data from the stock exchange; for data containing text data such as corporate performance, it can collect securities report data from the Financial Services Agency; and for text data such as segment information, it can collect data from business report briefing materials in corporate IR materials. For example, if the question is "What are the latest developments in Company X's automotive battery business?", the system will determine that searching the briefing materials in the Company X folder, where PDF files are stored in the storage unit 3C (internal information source) of the server 3, will yield the best information, and will collect the information. The company name can be searched not only by the official name of Company X, but also by abbreviations and English names. After identifying the company through natural language processing, the system will perform internet searches and folder searches for specialized terms such as "recent developments in the automotive battery business," prioritizing those containing the best information. The ability to freely utilize various search methods and information sources, such as internet searches, vector searches, JSON searches, and natural language searches, is beneficial. The ability to determine which search method is appropriate and which data is important in answering a question yields particularly valuable results.

[0335] Furthermore, the content of the report will differ depending on whether the collected data is numerical or text. When the collected data is numerical, the information processing module is characterized by its focus on time-series data that changes over time, such as stock price data and performance data. Since the numerical data calculated from this data also changes over time, the information processing module calculates this numerical data at each point in time, and the explanatory results will differ depending on the calculated values. The information generation module 40 then generates the report for each part based on these explanatory results. In the case of numerical data, the information processing module 20 changes the explanatory text based on the range and magnitude relationships. For example, the information processing module 20 could compare the 25-day moving average with the current value and generate text such as, "If it exceeds 30%, the deviation rate has exceeded 30%, indicating overheating."

[0336] When the information processing module 20 and the information generation module 40 work together, after processing by the information processing module 20, natural language processing such as GPT is added, and importance is judged along with other results to perform a comprehensive technical analysis. Sometimes, if an item is of lower importance than other explanatory results, other explanatory results are displayed, or if it is of higher importance, it is emphasized and a detailed explanation is added; this is the role of the information generation module 40. While a function outputs the same text when a certain numerical value is calculated, the information generation module 40 has the characteristic of changing flexibly depending on the user's preferences and other numerical values ​​and text. Therefore, when these two work together, the explanation is made in language that is very easy for the user to understand and follows the flow of conversation.

[0337] Let's consider an example where the information processing module 20 outputs two explanatory results. For instance, suppose the current price reaches the 200-day moving average, and a golden cross occurs where the 5-day moving average is above the 25-day moving average. These two results are contradictory: a sell signal in the long term and a buy signal in the short term. However, the information generation module 40 takes other factors into account and generates a report that emphasizes the long-term sell signal. This involves a judgment that is close to human judgment. However, this judgment is only possible with the information processing module 20. The same applies to comprehensive judgment results that combine corporate performance analysis, technical analysis, and current trends.

[0338] There is also a combination of Trigger Module 30 and Information Generation Module 40. There are various triggers, but for example, let's say there is a trigger for an upward revision of a company's earnings forecast that significantly exceeds expectations. In this case, the system identifies the stocks that meet the conditions of this trigger, and a report calculated using the current stock price of those stocks is automatically generated and provided to the user. According to this, the type of report to be generated by Trigger Module 30 is determined, and the report generated accordingly is provided. The effect of instantly reporting the reason why the trigger was activated, the expected effect of the trigger, the current stock price position, and whether there is any overheating based on technical indicators is extremely significant. The effect of the report, which is generated not just as a simple trigger, such as an article about an upward revision, but also with technical analysis, trend analysis, and characteristics of recent stock price movements after upward revisions based on the market price at that time, is remarkable. A major feature is that the report generation can be changed depending on the type of trigger. For example, let's say a trigger occurs for stock A that indicates three important buy signals have been activated. Then, when the trigger is activated by the occurrence of a golden cross, the moving average deviation rate being positive for all four indicators, and two weeks having passed since the announcement of increased revenue, profits, and dividends, the information generation module 40 automatically generates a report that includes an explanation of these three conditions.

[0339] Furthermore, there is also a combination of the information generation module 40 and the interface module 50. By combining it with tools that can identify individuals, such as LINE or My Page (linked with login information, etc.), the information generation module 40 can generate various types of information depending on the user. For example, by identifying the user, it becomes possible to generate different reports for each user based on user information such as the user's past questions, conversation history, and survey responses. For example, by tailoring the report structure to the user's preferences, or by allowing the user to specify or select the number of characters (distribution) and wording of each chapter, it becomes possible to generate reports that suit each user's purpose and preferences. It is also possible to create a report structure that is of high interest to the user based on survey responses and dialogue records. The information that users seek varies. Some users who are new to stocks may require explanations of stock terminology in an easy-to-understand way, or explanations that do not use too much technical jargon such as technical indicators, while users who are good at short-term trading may require reports that quickly and thoroughly convey changes in technical indicators and trends. Users are no longer satisfied with centralized reports and articles; they demand information tailored to their individual needs and preferences. Those with limited funds and existing holdings are primarily interested in the performance of their existing holdings, while those with ample funds and looking to buy stocks want buy recommendations above all else. Therefore, a personalized report system would be highly effective for users.

[0340] For example, we can provide a system that allows users to customize the chapter structure to their preferences. Specifically, we can provide an automated report generation system that allows users to select chapters such as company overview, 5-year company performance, explanations of indicators such as PER, explanations of industry trends, explanations of segment information, explanations of our company's revenue drivers, stock price trend analysis, technical analysis, and overall analysis results, and adjust the number of chapters and word counts. In addition, we can provide an automated report generation system that allows users to change the chapter structure and composition according to their choices, preferences, and experience, such as using polite language or avoiding technical jargon to make it easy for beginners to understand.

[0341] These reports can be customized by allowing users to choose or instruct on the customization options. For example, one possible approach is to include a chapter explaining the company using its most familiar products. The themes for each chapter are treated as questions, processed using natural language processing, vectorized using methods such as Bert, searched using vectors, and the results are summarized and compiled using methods like GPT based on their similarity, enabling the generation of such interactive reports.

[0342] To summarize the above points, it can be said as follows: A system for searching and analyzing stock price data, corporate performance data, technical indicators, and IR materials, (a) Extract specific features from the data, (b) Includes an information processing module that performs data analysis based on the features, (c) The information processing module is a system characterized by performing at least one of the following: clustering, vector search, JSON search, database search, API search, file search, and internet search. Furthermore, the system is characterized in that the information processing module, based on the features, weights the results according to the user's investment style and provides personalized search results. The system is further characterized in that the information processing module includes an information generation module that generates a stock report based on the analysis results, which includes one of the following: company overview, financial results summary, recent news, technical indicator analysis, trend analysis, and overall evaluation. Furthermore, the system is characterized by including a trigger system that notifies the user and resumes the interaction when specific conditions occur. Furthermore, the system is characterized by including an interface module and providing stock reports generated by the information generation module in an interactive or non-interactive format. Furthermore, the system includes an automated trading module and is characterized by automatically performing trades based on user settings and market conditions.

[0343] It can solve specific problems with existing technologies (e.g., insufficient personalization, delays in real-time response). Unlike GPT, it provides a system with features such as real-time capabilities, personalization, industry specialization, and a trigger system that allows users to proactively engage with the system. The trigger system is different from the email distribution system of securities companies. It can initiate a dialogue, from which questions can be asked, and it is not just a simple notification; it can be used as a hook to resolve various questions, which is a special effect that can be expected. Furthermore, scenarios that respond to the latest investment trends can be considered, such as use in prediction markets using AI, or trend analysis in the cryptocurrency market. A system is provided that can answer all questions about individual stocks using AI-powered natural language processing. For example, technical indicators are very difficult for beginners to use, and even for people familiar with stocks, even if technical indicators are displayed in conjunction with charts, it is very difficult to interpret how to think about them or how to use them in combination with other indicators. However, many users find them useful, and there is a very high need for a comprehensive evaluation of individual technical indicators, including their meaning, according to the situation at the time.

[0344] LINE offers features for individual and group chats, but these are not being effectively utilized for communication with businesses. The same applies to user profiles; providing specific, individualized information is time-consuming, so user-specific data is often limited to purchase history (such as stock trading history) and user attributes (name, address, contact information). This is because engaging in individual conversations with each customer to meet their needs would be highly inefficient and burdensome for businesses. While there are interactive forms of communication, such as surveys, they struggle to address individual concerns and challenges. Chatbot systems, exemplified by ChatGPT, are innovative products and services due to their conversational format, but they are often unable to address the specific challenges and concerns of individuals. This is because each individual's situation and challenges are diverse, making it difficult to continuously remember and respond to that personal information. It is also difficult to provide solutions based on the latest information (usually learned several years or months ago), difficult to derive solutions based on industry-specific information, difficult to have a conversation tailored to the individual based on various pieces of information like a human would, and one-way communication that does not progress unless the user initiates the dialogue, thus failing to notice potential problems.

[0345] The solution to each of these five barriers and challenges is the AI ​​concierge service (information generation system and method). An AI concierge service that overcomes these five barriers will provide personalized support for each user's challenges and questions, based on the latest and most specialized information, weighted according to the user's attributes. This enables an AI-based concierge service that not only responds to user inquiries but also provides up-to-date answers as circumstances change.

[0346] Remembering and utilizing personal information (remembering each individual's personal information and engaging in conversations based on that memory) is extremely important. To achieve this, the memory of conversations and individualized conversations based on records of those conversations are crucial. This is especially true in the financial industry, and in the world of stock investment in particular, where information related to money is involved, making these individualized conversations tailored to each person's situation extremely important because there are so many options. To narrow down from a large number of options, the ability to record and utilize conversations makes it possible, for example, to present a list and allow for filtering as needed. For example, the effect of being able to have a series of conversations such as, "Which stocks are above 5%?", "Present the list," "Among those, which stocks have a trading volume of 2 billion yen or more?", "Narrow down the list and present it," "Among those, which stocks are projected to have increased revenue and profits?", "Present 3 stocks, shall I present their company profiles, corporate performance, and technical indicators?", "Yes, please," "Provide the report" is significant. It differs greatly from normal screening, allowing users to narrow down from a large number of options as they see fit, and is expected to have a special effect of greatly improving user convenience. This kind of narrowing down is possible precisely because it involves dialogue and because the previous conversation is remembered.

[0347] Beyond the screening mentioned above, individual consultations and problem-solving regarding specific stocks, holdings, trading activity, and differences between one's own results and those of others are extremely important. Solving these requires a consultation service tailored to each individual's style, investment results, holdings, stock performance, asset status, and age. Typically, such consultations are conducted face-to-face by sales representatives who have experience handling various user inquiries. This is because customer satisfaction is higher when they can respond while remembering the user's personality, family structure, assets, and history. The most important aspects of this consultation service are the ability to remember past interactions and the user's background, and the ability to offer suggestions tailored to the user's situation based on that memory. Therefore, dialogues and suggestions based on records accumulated from various interactions, where each user's conversation is documented, are an indispensable service in the financial industry. The most crucial element of this is the recording of each user's conversations and the use of those records in subsequent conversations.

[0348] The services offered can include advice, order placement, or simply current situation analysis and reports without going as far as advice. The scope can also be narrowed or broadened, such as support only for existing holdings, support only for stocks of interest, support only for US stocks, or advice only for NISA accounts. If we create a three-dimensional model with financial products on the horizontal axis, investment styles on the vertical axis, and service content on the height, the horizontal axis represents the number of financial products (US stocks, Japanese stocks, FX, cryptocurrencies, interest rate information, investment trusts, ETFs, REITs, etc.), and the vertical axis represents the number of investment styles (short-term trading, medium- to long-term investment, regular contributions, small amounts, large amounts, etc.). The height represents the content of the service. This ranges from simple report and information provision to dialogue triggered by reports and information, to specific advice on stock selection, quantity, buy or sell recommendations, to the initiation of dialogue triggered by advice, and even automated order placement, triggered by automated order placement.

[0349] The financial industry has unique personal characteristics (asset composition, size, preferences, risk tolerance, and necessary information (NISA, stock information, interest rate information, exchange rate information, etc.) which vary greatly from user to user). By efficiently storing this information in databases and files for each user, and also by storing records of conversations with each user, it becomes possible to conduct conversations and provide information based on user records. For example, for a user with stock assets of 100 million yen or more, who has significant unrealized gains and is highly interested in undervalued stocks, services could include delivering reports on companies with low P / E ratios, high dividends, and a strong desire for dividend increases, providing buy advice on specific stocks, whether to buy or sell, and the quantity to buy, placing automatic orders, and initiating conversations accordingly. On the other hand, for a user who is trying to increase their capital of 500,000 yen through margin trading, services could include introducing stocks that have shown strong buy signals from technical indicators targeting short-term arbitrage, providing advice, placing automatic orders, and initiating conversations accordingly.

[0350] The difference between providing a service and initiating a dialogue is that providing a service is a one-time event, while initiating a dialogue involves accepting questions after the service has been provided. This is a typical example of how the information required varies depending on the user's investment style. This user-specific information can be obtained through dialogue or surveys. Importantly, this technological development is extremely valuable to users because it enables dialogue that provides different suggestions, support, reports, advice, and automated order placement for each user through these records. Normally, information is provided uniformly rather than differentiating it for each user. Even in the user management screens of securities companies, while trading records and order records change for each user, information based on dialogue records, survey results, and investment style is not provided, and users have to find it themselves. In particular, there are no systems that generate answers to questions for each user and immediately answer questions about held stocks or stocks of interest for each user. Questions and doubts arise on a wide range of topics, such as information about stock investments, how to use member sites, and questions about financial products, but questions and answers about stock investments in particular are in demand, but it is a difficult area to meet the demand. This is because it's impossible to answer questions about 3,800 Japanese stocks alone, and over 10,000 stocks including US and other international stocks, manually. Therefore, the information provided ends up being a list of reports and news articles, resulting in a uniform and undifferentiated presentation.

[0351] The information that is needed is diverse, including company performance, daily stock prices, technical indicators, and company news. However, this can only be achieved through collaborative computer work such as chat, database technology, and file operations. In particular, the fact that the stock information needed differs from user to user is crucial. In other words, user A needs information on when to sell stock A, while user B wants to know when to buy the same stock A. The information each person needs regarding stock A is vastly different, which is a characteristic unique to financial products (especially volatile products whose value fluctuates from time to time, such as stocks). This stems from the fact that the value of financial products fluctuates daily. Therefore, the following special effects can be expected from dialogue that changes for each user regarding these price-fluctuating financial products: Daily price fluctuation information, changes in the information required due to changes in value. For example, the information required when the stock price is 1000 yen is vastly different from the information required when the stock price has fallen to 500 yen. This is a characteristic unique to price-fluctuating financial products. That's why there's a strong need for information that incorporates the ever-changing circumstances, tailored to each user and time period.

[0352] Regarding the scope of information described in this statement, all wording in this statement includes all information, whether or not it involves price fluctuations. However, it can be limited to financial products that involve price fluctuations, limited to stock investments, limited to financial products in general, or it can include all kinds of information, such as travel information and caregiving information. For example, in the case of travel information, it means that the requirements will differ for each user, depending on their age, family structure, who they are traveling with, location, purpose, budget, etc.

[0353] For example, information-generating AI requires a learning period, making it weak when it comes to the most recent information. While there are ways to compensate for this through web access, currently, for slightly difficult, recent problems, it's often recommended to encourage users to access the latest information via the web. This is partly because many websites prohibit scraping, limiting robot access to recent information and making it difficult to include recent information in dialogue. In this respect, human web access often has the advantage.

[0354] However, this drawback can be compensated for by accessing APIs and necessary industry-specific up-to-date information, and by creating databases of this information. For example, with weather forecasts, it is possible to obtain the latest information by linking with APIs of external weather forecasting services to retrieve the latest weather forecast data. Methods for acquiring the latest information include API integration, API integration and database creation, file downloads and uploads, and web access. For information such as stock price information, the latest financial results, and the latest technical indicators, it is necessary to store real-time and up-to-date information in the storage unit 3C of server 3, and to accurately retrieve and utilize this information in response to user inquiries. There is a great need for the distribution of information and dialogue that takes this up-to-date information into account. To achieve this, it is necessary to store and accumulate this information in databases and files by communicating with specific external servers using APIs or crawling methods, and to create various triggers (to determine whether a certain condition is met) so that it can be used for dialogue at any time. At the same time, it becomes possible to build triggers.

[0355] One example involves using API integration and periodic crawling to access an external server, and then storing the collected information on server 3. During this process, changes are detected based on triggers when data is entered into the database. And, The user is notified. Another example is to perform API integration or periodic crawling on an external server, and instead of storing the data on server 3, capture changes based on a trigger and notify the user. Another example is to perform API integration or periodic crawling on an external server, collect the latest information on server 3, and notify the user. Another example is to perform API integration or periodic crawling on an external server, collect the latest information on server 3, and then resume or start interaction with the user. Another example is to perform API integration or periodic crawling on an external server, collect the latest information on server 3, and then resume or start interaction with the user when a trigger is activated.

[0356] A trigger is a cause or trigger that brings about a specific event. It can also mean an activation or activation device. In the IT and web industries, the term "trigger" is often used to refer to software that performs a specific process when a particular event occurs. In the stock market, there are various triggers, such as a specific company announcing its earnings, a specific stock's RSI (Relative Strength Index) falling below 30%, a specific company revising its forecast upwards, a specific company's stock price rising for three consecutive days, a weaker yen, or rising oil prices. To catch this trigger information, it is necessary to catch the latest information, and it is extremely valuable when the trigger module 30 of the information generation system (server 3) detects this trigger information, initiating a dialogue between the information generation system and the investor user, allowing for questions and answers. For example, if the trigger is activated that stock A has risen for three consecutive days, the reason for the rise will be reported, a dialogue about that stock will begin, and being able to ask about technical indicators and get answers, or ask about the company overview and get answers, will be extremely helpful in making investment decisions.

[0357] A trigger, meaning "a trigger that initiates a process," refers to a clause that is activated when a predetermined condition is violated. For example, if the dividend yield exceeds 5%, the condition is met. These triggers are often linked to database recording, which is convenient, but of course, other methods such as files or APIs can also be used instead of a database. A trigger system only makes sense when the interaction is based on the latest information. With this trigger system, it becomes possible to proactively initiate conversations in response to user requests, provide advice and guidance, automate ordering in conjunction with the ordering system, and lead to the next action.

[0358] For example, a trigger system can be activated, and a LINE push notification can be sent to resume the conversation. When an event of interest to the user occurs, the conversation can be resumed, initiating a discussion about that event and prompting user action. For example, when a dividend increase is announced, and this is converted to a 5% increase in the current stock price, a trigger is a trigger that causes or initiates a specific event. It also includes the meaning of operation or activation device. In the IT and web industries, a trigger is often used to describe software that performs a specific process when a specific event occurs. A new trigger is activated, and reports such as company overviews, technical indicators, and corporate performance related to the stock are sent, which can have a significant impact on the user's investment decisions.

[0359] Furthermore, for example, an order can be placed with a securities company upon activation of a trigger system. By specifying the stock and quantity to place an order in response to an event of the user's particular interest, the order related to that event is automated, resulting in a special effect of automating the user's investment actions. For example, if a dividend increase is announced and, when converted to the current stock price, this triggers a new event of over 5%, an order related to that stock is automated, potentially having a significant impact on the user's decision-making.

[0360] Furthermore, for example, the system can be triggered to resume dialogue and advise on placing orders for the relevant stock. The system has a special effect in that it can initiate dialogue related to events of the user's interest, prompting user action. For instance, a new trigger could be activated when a dividend increase is announced, resulting in a 5% increase in the current stock price. This triggers advice on placing orders for the stock, allowing the user to significantly influence their investment decisions. These can be implemented at any level of the three-dimensional model described earlier. For example, in a stock investment model (horizontal axis), a user with a strong short-term trading orientation (vertical axis) could receive advice (height axis) triggered by short-term technical indicators. Triggers are optional, but their presence makes the process clearer.

[0361] For example, in the stock market, information such as stock prices, company performance, and charts is of high importance. For users who need to care for their parents, information about laws, long-term care insurance, and care needs assessment is important. For users planning a trip, information such as timetables, weather, local tourist event information, and access methods to recommended tourist spots is important when planning how to spend their time at their destination. The information needed varies greatly depending on the problem or concern. It is extremely difficult to expect an information-generating AI to answer all of these questions. It is necessary to create a database of industry-specific information and retrieve information according to the purpose and concerns. For example, by including the latest timetables, it becomes possible to create travel plans that take into account information such as trains and buses. The convenience is greatly increased by including the latest long-term care remuneration, trends in legal revisions according to the level of care needed, and support information from each local government. Significant progress can be made by incorporating technical indicators that are in line with the latest financial and stock market conditions, as well as news and changes in economic conditions.

[0362] For example, in response to an inquiry like, "Which stocks hit a new year-to-date high today?", it becomes possible to answer, "There are 10 stocks that hit a new year-to-date high today." "Among those 10 stocks, are there any with particularly low P / E ratios and high dividends?" "Stock A has a 5% dividend and a P / E ratio of 10, and there is the following news about it." This kind of exchange is possible because the conversation continues with each user, the information is up-to-date, the user is familiar with industry-specific information, and the format is conversational.

[0363] It is necessary to weight multiple sources of information and make comprehensive judgments, rather than relying on just one. In other words, the ability to assign industry-specific weights to various pieces of information, select and discard them, and engage in dialogue is required. The amount of information is often large, and it is difficult to weight it according to individual concerns and challenges, leaving the decision of what to discard challenging. However, this challenge can be overcome by selecting appropriate search methods (such as vector search, JSON search, and various SQL queries). Furthermore, it is necessary to form dialogues and provide information by comprehensively considering multiple sources, rather than simply summarizing from a single source. Investors with a strong short-term trading orientation who place weight on technical indicators and investors who place weight on corporate performance and invest in undervalued stocks will handle the same information differently. Even if both consider technical indicators, the weighting will differ significantly. Forming a dialogue while taking these factors into account requires advanced techniques, and this is something that can only be achieved through collaborative work with computers. For example, it is possible to allow users to select weights, and the system can also determine where to focus based on dialogue records and investment styles. By categorizing stock information such as technical indicators, company performance, macroeconomic environment, and the investment behavior of other investors, and varying the weighting of each category according to the user and investment style, it becomes possible to change the weighting of information. This weighting information can also be woven into the dialogue. For example, in the previous example, investors who conduct margin trading with 500,000 yen or those with a strong short-term orientation need to have a very high weighting of buy and sell signals from short-term technical indicators, while those with assets of 100 million yen and a medium- to long-term orientation do not need such information, or it should be provided with a very low weighting. This makes it possible to simply add a little of this information to other information. The weighting of information sources and categories according to style can be stored as information (e.g., in a database), and this can be obtained by referring to it in the dialogue with the user (referencing the database information). This applies not only to stock information, but also to other examples, such as how the weighting of information on individual stocks and the overall market differs greatly between users who are strongly inclined towards ETFs and mutual funds and those who are strongly inclined towards individual stocks when seeking NISA information.Travel information, such as gourmet information, sightseeing information, and marine sports information, is an example of this weighted information, as its weighting changes depending on the user's preferences and purpose. It is similar to user information and is often used together, but while individuals usually make their own judgments and unconsciously weight these pieces of information when searching, quantifying and explicitly displaying them allows for a more concierge-like approach to information, as it helps to gauge importance and facilitates dialogue. Weighted information is crucial when accumulating information for each user and selecting from a large amount of information. This weighting is a vital factor in selecting and filtering information.

[0364] The drawback that the chatbot won't progress unless the user takes action to engage in dialogue can be overcome by using it in conjunction with tools such as LINE, My Page, and email distribution from the management screen. Chatbots typically begin dialogue in response to a user's question. However, the trigger isn't limited to user questions; it can also be triggered by some change or event, leading to the resumption or commencement of dialogue. This refers to various changes such as stock prices, weather, company performance, legal changes, and changes in the macro environment, which can initiate dialogue.

[0365] Furthermore, the system relates to the communication application "LINE" (registered trademark), which has individual and group dialogue functions, and "My Page," which is accessed by logging in. In particular, it is a system designed to be effectively used for dialogue with companies. Currently, it is inefficient for companies to engage in dialogue with each customer individually, and it places a heavy burden on the companies. Therefore, the system provides a way to efficiently and effectively facilitate dialogue tailored to the individual concerns and challenges of each customer.

[0366] While conventional chatbot systems offer innovative product and service features through their conversational format, they are considered to have difficulty providing specific solutions tailored to individual circumstances and challenges. In particular, their use is limited in situations where personalized conversations based on various information, similar to those of a human, are required, as are the presentation of solutions based on the latest information and responses based on industry-specific information.

[0367] The above system can solve the following five problems. 1. Remembering and utilizing personal information 2. Providing solutions based on the latest information. 3. Providing solutions based on industry-specific information. 4. The ability to weight and select various pieces of information for dialogue. 5. Problems that do not progress unless the user initiates the interaction.

[0368] To address these challenges, the above system provides a system that enables individualized dialogue, specifically by collecting and storing information tailored to each individual's situation, and based on that information, facilitating efficient and effective communication between customers and companies.

[0369] Another example is an interactive intelligent concierge system and method. This system and method relates to a technology that stores and utilizes user information in a database via LINE or My Page, and then uses APIs and the accumulated database to provide a conversational concierge service tailored to each individual's situation. The conversational personalized concierge service provides information tailored to each individual's situation by storing user information in a database and utilizing it at the appropriate time. This technology is in particularly high demand in the field of stock investment. Conventional stock investment information systems often provide uniform information to users and do not adequately address individual holdings, hobbies, preferences, or investment tendencies. As a result, users had to select and filter information based on their own judgment, which required a lot of time and effort. To solve this problem, there is a need for technology that collects user information using APIs, accumulated databases, and files, and constructs responses in a conversational format based on the accumulated information. Specifically, there is a demand for the provision of private equity information that provides investment information and advice optimized for each user, based on information about the user identified by their ID or email address (e.g., hobbies, preferences, investment tendencies, stock holdings, etc.).

[0370] Furthermore, investment information needs to change continuously based on the user's holdings, age, investment style, and other factors. Therefore, a system that continuously learns based on conversational history and input information is crucial. This system utilizes diverse data such as stock price information, technical analysis, company performance information, and macroeconomic information as input, enabling it to provide the most useful information to the user in real time. This allows users to easily obtain investment information based on their own investment style and areas of interest, enabling them to make investment decisions efficiently. Consequently, personalized concierge service technology tailored to individual needs overcomes the shortcomings of conventional stock investment information systems and achieves a higher level of information provision.

[0371] Conventional stock investment information systems provide a large amount of information uniformly to all users, making it difficult to adequately address individual preferences, tastes, and investment tendencies. As a result, users must select and filter the necessary information based on their own judgment, which requires a significant amount of time and effort, leading to a decrease in the efficiency of investment decisions. The system described above stores and accumulates individual user information (e.g., hobbies, preferences, investment tendencies, holdings, etc.) in a database and utilizes this information through APIs and conversational responses to provide optimal investment information and advice tailored to the user's situation. Furthermore, by continuously learning based on conversation history and input information, the system can realize a personalized concierge service that meets the individual needs of each user. This allows users to efficiently acquire investment information based on their own investment style and areas of interest in stock investment, enabling them to make quick and accurate investment decisions. Moreover, this technology is applicable not only to the field of stock investment but also to many other fields, such as FX, cryptocurrencies, investment trusts, elderly care issues, pension issues, and travel itinerary consultations, in order to provide information tailored to the individual's situation.

[0372] The system described above is a system that accumulates individual user information in a conversational format on platforms such as LINE and My Page, through the input of an ID, password, email address, etc. that can identify individual users. By using the accumulated individual information to facilitate conversations, it provides users with a private, concierge-like service. Depending on the purpose of the service, the stock investment concierge service uses stock price information and financial information as input information, the nursing care concierge service uses nursing care information and administrative information, and the travel concierge service can provide information tailored to the user's needs, such as their individual purpose (family trip, presence of small children, elderly couple, lover, women's group, etc.), location (Hokkaido, Okinawa, Kyoto, etc.), budget, hobbies, and preferences.

[0373] In the system described above, personalized investment advice generation algorithms, automated trading algorithms, real-time notification and alert systems, conversational interfaces and contextual understanding algorithms, user profile management and personalization technologies, investment recommendations, news article similarity search, customer support, and investment target similarity search technologies can provide investment advice and automated trading optimized for individual users, dramatically improving user engagement. This technology can improve users' investment results, significantly reduce time and effort through automated trading, and increase investment efficiency. Furthermore, the real-time notification system supports users' investment decisions by providing timely information, contributing to differentiation from competitors. In addition, the conversational interface and contextual understanding algorithms can dramatically improve the user experience by smoothing user interaction and generating contextually accurate responses. These technologies are widely applicable across the financial services industry, and have significant industrial applicability, particularly in services for individual investors.

[0374] A concrete example is an AI concierge service that uses user input such as their holdings and investment style, along with conversation records, to identify specific stocks and then uses push notifications or emails to inform users of changes in technical indicators or financial results. For instance, this service could notify users via push notifications or emails when the RSI (Relative Strength Index) of a stock they hold or are interested in falls below 30%.

[0375] Another specific example is an AI concierge service that identifies stocks with changes in technical indicators or financial results based on user input such as holdings and investment style, as well as dialogue records. This service then uses these changes as triggers to notify the user via push notifications or email. For example, this could be used to notify users with a high dividend yield preference about stocks that have recently announced increased dividends and now have a dividend yield of 5% or more, via push notifications or email.

[0376] These specific examples are implemented using a database that stores user information, along with a program that stores stock price information and financial results information in the database, a trigger system that detects changes, and a transmission system that notifies users of those changes. An original trigger system can also be created to suit the user's preferences. The initiation of the conversation through this trigger is crucial.

[0377] To summarize the above system, it is a concierge system that stores individual user information in a database and provides personalized concierge services through dialogue. It is characterized by building dialogues based on the accumulated personal information and information obtained from dialogues, according to each individual's hobbies, preferences, characteristics, investment tendencies, and holdings, and changing investment information based on the interaction history and input information. Furthermore, the concierge system is characterized by acquiring stock price information, technical information, corporate performance information, macro information, and economic information as input information using APIs, databases, file systems, etc., and forming conversations based on information such as conversation history, personal age, investment style, holdings, and stocks of interest.

[0378] Furthermore, automated trading and risk management systems can also be mentioned. Specifically, this concerns a system that uses an automated trading algorithm that automatically executes trades based on user settings and market conditions, as well as a risk management tool for evaluating and managing investment risk. This system analyzes input data such as stock price data, technical indicators, and financial figures, along with the user's individual confidence and importance of technical indicators and the importance of financial data, based on their interaction history. It then generates buy and sell signals, automatically sends these signals, and notifies users via LINE and other means.

[0379] Existing data management systems are not designed for interactive user interaction and often rely on uniform interfaces and procedures. Furthermore, these systems do not provide information based on technical metrics or numerical data trusted by individual users. Therefore, data management tailored to individual user needs and levels of trust is difficult, and they lack flexibility. Consequently, there is a need for a new data management system that provides highly reliable data while meeting individual user requirements.

[0380] Existing data management systems do not offer interactive operation and do not provide information based on technical indicators or numerical data that individual users trust, resulting in the problem of uniform information provision. This makes it difficult to provide personalized data that reflects individual users' investment styles, hobbies, preferences, experience, and know-how. Therefore, there is a need for a new system that can provide notifications based on the user's personal information, such as changes in technical indicators, macro indicators, and financial figures, based on the user's trusted technical indicators and numerical data. The above system will provide customizable notifications according to the user's investment style and preferences through various channels such as LINE, My Page notifications, and email notifications.

[0381] The automated trading and risk management system described above includes a communication unit as a means of receiving user questions, issues, and commands, a storage unit for storing this information, and a notification unit that notifies the user when specific conditions occur. After notification, it includes a dialogue unit for further instructions and order execution in a conversational format, thereby increasing the user's investment efficiency and providing a secure environment. Real-time notification and alert systems and automated trading and risk management systems have a significant impact on investment activities in financial markets. Notification algorithms provide real-time notifications of stock price fluctuations and important market events, enabling investors to make quick and appropriate decisions. Furthermore, customizable alert systems can provide personalized alerts according to user settings, realizing information provision that meets individual investment needs. In addition, the technology of automated trading algorithms that automatically execute trades based on user settings and market conditions can dramatically increase investment efficiency. It is also possible to provide a safer investment environment by evaluating and managing investment risks with risk management tools. This allows for appropriate decisions on selling held stocks and buying unheld stocks, greatly improving the reliability and effectiveness of investment activities. Therefore, the above system has the potential to be widely adopted in the financial services industry, and its industrial applicability is very high.

[0382] A concrete example is an AI concierge service that identifies stocks with changes in technical indicators or financial results based on user input such as holdings and investment style, as well as conversation records, and then executes automated trades of those stocks by specifying the quantity. For example, for a user with a strong preference for technical indicators, this service could automatically execute buy orders for stocks where the RSI is below 20% and the 75-day moving average deviation is 20% or more, based on newly applicable conditions.

[0383] This specific example is implemented using a database that stores user information, along with a program that stores stock price information and financial results information in the database, a trigger system that detects changes, and an order system that executes those triggers. Original trigger systems and order systems can also be created to suit the user's preferences.

[0384] To summarize the above system, it is an automated trading and risk management system that includes an automated trading algorithm that automatically executes trades based on user settings and market conditions, characterized in that the automated trading algorithm analyzes the trading parameters entered by the user and real-time market data to generate trading signals, and automatically executes trades based on those trading signals. Furthermore, the automated trading and risk management system according to claim 1 includes a risk management tool for evaluating and managing investment risk, wherein the risk management tool evaluates risk factors related to the user's investment portfolio, generates a risk profile, recommends actions to reduce risk based on the risk profile, and automatically executes such actions as necessary. Furthermore, the automated trading and risk management system is characterized by the fact that the automated trading algorithm and the risk management tool work together on an integrated platform, thereby improving investment efficiency and providing an environment in which users can use with peace of mind.

[0385] Another example is an automated advice system. Specifically, this refers to a system that uses a management tool to automatically provide advice based on user settings and market conditions. This system analyzes input data such as stock price data, technical indicators, and financial figures, as well as the user's individual confidence level and importance of technical indicators and the importance of financial data, based on their interaction history. It then generates buy and sell signals, automatically sends these signals, and notifies users via LINE or similar platforms.

[0386] Existing data management systems are not designed for interactive user interaction and often rely on uniform interfaces and procedures. Furthermore, these systems do not provide information based on technical metrics or numerical data trusted by individual users. Therefore, data management tailored to individual user needs and levels of trust is difficult, and they lack flexibility. Consequently, there is a need for a new data management system that provides highly reliable data while meeting individual user requirements.

[0387] Existing data management systems do not offer interactive operation and do not provide information based on technical indicators or numerical data that individual users trust, resulting in the problem of uniform information provision. This makes it difficult to provide personalized data that reflects individual users' investment styles, hobbies, preferences, experience, and know-how. Therefore, there is a need for a new system that can provide notifications based on the user's personal information, such as changes in technical indicators, macro indicators, and financial figures, based on the user's trusted technical indicators and numerical data. The above system will provide customizable notifications according to the user's investment style and preferences through various channels such as LINE, My Page notifications, and email notifications.

[0388] The above automated advice system includes a communication unit as a means of receiving user questions, issues, and commands, a memory unit for storing this information, and further includes a notification unit that notifies the user when specific conditions occur. After notification, it includes a dialogue unit for giving further instructions or executing orders in a conversational format, thereby increasing the user's investment efficiency and providing a secure environment for use.

[0389] The real-time notification and alert system and automated advice system described above will have a significant impact on investment activities in financial markets. The notification algorithm provides real-time notifications of stock price fluctuations and important market events, enabling investors to make quick and appropriate decisions. Furthermore, the customizable alert system can provide personalized alerts based on user settings, realizing information tailored to individual investment needs. In addition, the automated advice technology, which automatically provides advice based on user settings and market conditions, can dramatically improve investment efficiency. This allows for appropriate decisions regarding the sale of holdings and the purchase of unheld stocks, greatly improving the reliability and effectiveness of investment activities. Therefore, the above systems have the potential for widespread adoption in the financial services industry, and their industrial applicability is very high.

[0390] A concrete example is an AI concierge service that identifies stocks with changes in technical indicators or financial results based on user input such as holdings and investment style, as well as dialogue records, and then automatically provides advice on those stocks based on the user's specified quantity. For example, for a user with a strong preference for technical indicators, this service could automatically provide buy advice for stocks where the RSI is below 20% and the 75-day moving average deviation is 20% or more, based on newly applicable conditions.

[0391] This specific example is implemented using a database that stores user information, along with a program that stores stock price information and financial results information in the database, a trigger system that detects changes, and an advice system that notifies users of those changes. Original trigger and advice systems can also be created to suit the user's preferences.

[0392] To summarize the above system, it is an automated advice system that automatically provides advice based on user settings and market conditions, characterized in that the automated advice algorithm analyzes the trading parameters input by the user and real-time market data to generate trading signals, and automatically executes advice based on those trading signals. Furthermore, the automated advice system is characterized in that the automated advice tool evaluates risk factors related to the user's investment portfolio, generates a risk profile, recommends actions to reduce risk based on that risk profile, and automatically provides advice as needed. Furthermore, the automated advisory system is characterized by the fact that the automated advisory tools work together on an integrated platform, improving investment efficiency and providing an environment in which users can use the system with peace of mind.

[0393] [Embodiment 5] Next, Embodiment 5 of the present invention will be described in detail. Embodiment 5 is a specific example of a combination of each module of System 1. The system of Embodiment 5 includes one of the following modules. A brief explanation will be given again, including any overlap with the above description.

[0394] [Information Gathering Module] The information gathering module collects structured and unstructured data from various fields (finance, elderly care, travel, education, healthcare, etc.). For example, in the financial sector, it collects stock price data, company performance, news articles, and investor relations materials, while in the elderly care sector, it collects vital signs and daily living records.

[0395] [Information Processing Module] The information processing module analyzes collected data and extracts field-specific features. It can analyze unstructured data as well; for example, in the financial sector, it can perform sentiment analysis of news articles, and in the medical sector, it can perform anomaly detection based on patients' diagnostic history.

[0396] [Trigger Module] A trigger module is a module (trigger system) that, when specific conditions (such as stock price fluctuations or abnormal health data) occur, promptly resumes interaction with the user and initiates automatic notifications or appropriate actions.

[0397] [Information generation module] The information generation module is a module that, based on analysis results, timely generates personalized reports in various fields, such as financial reports, care plans, travel plans, educational performance reports, and medical diagnosis reports.

[0398] [Automated Trading / Automated Response Module] The automated trading and response module is a module that automatically executes trades in the financial sector based on stock prices and market movements, and automatically responds in the nursing care and medical sectors based on specific health conditions and diagnostic results.

[0399] [Risk Management Module] The risk management module performs risk assessments and implements management measures to minimize risks to the user's portfolio, investment strategy, and health status.

[0400] [Interface Module] Interface modules are modules that provide users with generated reports and interactions in interactive or non-interactive formats. They integrate with existing platforms such as LINE and My Page to provide a user-friendly environment.

[0401] Next, the system of Embodiment 5 will be described in more detail. The following description provides specific examples related to supporting stock investments in the financial sector, health monitoring and anomaly response in the elderly care sector, and the suggestion of personalized travel plans in the travel sector. Each specific example details how the system of Embodiment 5 operates in its respective field and provides valuable information to the user.

[0402] The first concrete example is the financial sector. In the financial sector, there is a demand for personalized information provision and automated trading for financial products (volatile products) such as stocks, bonds, mutual funds, FX, and cryptocurrencies. Below, we show examples of system operation and module combinations in the financial sector.

[0403] One example in the financial sector is an operation that monitors stock price trends and determines and executes buy and sell orders. A system that performs this operation can be implemented using a modular configuration consisting of an information gathering module, a trigger module, an automated trading module, and an interface module.

[0404] Another example in the financial sector is the operation of generating and providing investment reports based on news analysis results by AI. The system that performs this operation can be realized with a modular configuration consisting of an information gathering module, an information processing module, an information generation module, and an interface module.

[0405] Another example in the financial sector is the operation of performing a risk assessment and adjusting through automated trading when stock risk increases. A system that performs this operation can be realized with a modular configuration consisting of an information processing module, an automated trading module, and a risk management module.

[0406] Another example in the financial sector is the operation of analyzing a company's financial statements and executing trades. A system that performs this operation can be implemented using a modular configuration consisting of an information gathering module, an information processing module, a trigger module, and an automated trading module.

[0407] Another example in the financial sector is an operation that evaluates and provides high-dividend stocks to investors. A system that performs this operation can be implemented using a modular configuration consisting of an information gathering module, an information processing module, an information generation module, and an interface module.

[0408] Another example in the financial sector is an operation that monitors technical indicators of stock prices and automatically determines and executes buy and sell timings. A system that performs this operation can be implemented using a modular configuration consisting of an information gathering module, a trigger module, an automated trading module, and an interface module.

[0409] Another example in the financial sector is the use of AI to analyze the sentiment surrounding news related to a company and generate a report on its impact. A system that implements this operation can be realized through a modular configuration consisting of an information gathering module, an information processing module, an information generation module, and an interface module.

[0410] Another example in the financial sector is an automated trading system that minimizes risk by automatically buying and selling stocks when the risk of the holdings increases. A system that implements this operation can be realized through a modular configuration consisting of an information processing module, an automated trading module, and a risk management module.

[0411] Another example in the financial sector is an operation that performs analysis based on technical indicators and provides reports including buy / sell signals. A system that performs this operation can be implemented using a modular configuration consisting of an information gathering module, an information processing module, an information generation module, and an interface module.

[0412] Another example in the financial sector is the operation of automatically executing trades based on buy / sell signals when earnings results exceed expectations. A system that implements this operation can be realized with a modular configuration consisting of an information gathering module, an information processing module, a trigger module, and an automated trading module.

[0413] Another example in the financial sector is the operation of providing reports that predict future stock market trends based on AI-generated stock price predictions. The system that implements this operation can be realized with a modular configuration consisting of an information collection module, an information processing module, an information generation module, and an interface module.

[0414] Another example in the financial sector is the operation of monitoring overall market trends and responding with automated trading when risks increase. A system that implements this operation can be realized with a modular configuration consisting of an information gathering module, a trigger module, a risk management module, and an automated trading module.

[0415] Another example in the financial sector is an operation that evaluates high-dividend stocks and suggests stocks with attractive dividend yields. A system that performs this operation can be implemented using a modular configuration consisting of an information gathering module, an information processing module, an information generation module, and an interface module.

[0416] Another example in the financial sector is the operation of automatically executing trades based on stock split and merger information. A system that performs this operation can be implemented using a modular configuration consisting of an information gathering module, a trigger module, and an automated trading module.

[0417] Another example in the financial sector is the operation of performing financial analysis and market valuation of newly issued shares and providing reports on IPO stocks. A system that performs this operation can be implemented using a modular configuration consisting of an information gathering module, an information processing module, an information generation module, and an interface module.

[0418] The second specific example is the field of elderly care. In the elderly care field, there is a need to monitor the health status and vital data of those requiring care, and to respond when abnormalities are detected. Below, we show examples of system operation and module combinations in the field of elderly care.

[0419] One example in the field of elderly care is a system that monitors the vital signs of a person requiring care and notifies the caregiver if an abnormality occurs. A system that performs this operation can be implemented using a modular configuration consisting of an information collection module, a trigger module, and an interface module.

[0420] Another example in the field of elderly care is an action that assesses a person's health status, generates a report in a timely manner, and provides it to the caregiver. The system that performs this action can be implemented using a modular configuration consisting of an information collection module, an information processing module, an information generation module, and an interface module.

[0421] The third specific example is the travel sector. In the travel sector, there is a need for user-generated travel plans and automated booking based on price fluctuations. Below, we show examples of system operation and module combinations in the travel sector.

[0422] One example in the travel sector is an operation that monitors fluctuations in flight and hotel prices and automatically executes reservations. A system that performs this operation can be implemented using a modular configuration consisting of an information gathering module, a trigger module, an automated transaction module (automated reservation module), and an interface module.

[0423] Another example in the travel sector is an operation that suggests the most suitable travel plan to the user and provides it as a report. The system that performs this operation can be implemented using a modular configuration consisting of an information gathering module, an information processing module, an information generation module, and an interface module.

[0424] The fourth specific example is the field of education. In the field of education, there is a need to monitor students' grades and learning progress, and to provide appropriate learning advice. Below are examples of system operation and module combinations in the field of education.

[0425] One example in the field of education is an operation that analyzes students' grades and assignment progress and proposes a learning plan. A system that performs this operation can be realized with a modular configuration consisting of an information gathering module, an information processing module, an information generation module, and an interface module.

[0426] The fifth specific example is the medical field. In the medical field, there is a need for anomaly detection and medical support based on patients' vital data and diagnostic results. Below, we show examples of system operation and module combinations in the medical field.

[0427] One example in the medical field is a system that monitors a patient's vital data and notifies a doctor of any abnormalities to instruct them on appropriate action. A system that performs this operation can be implemented using a modular configuration consisting of an information collection module, an information processing module, a trigger module, and an interface module.

[0428] Next, we will describe the stock investment support system according to Embodiment 5. This system uses AI to collect and analyze financial data such as stock price data and corporate performance data, and supports investors in making investment decisions by providing them with timely and personalized information. The system analyzes stock price trends and market movements, automatically notifying users of optimal buying and selling timings, and also executes automated trades based on user settings.

[0429] This system primarily consists of an information gathering module, an information processing module, a trigger module, an information generation module, an automated trading module, and an interface module. Each module works in conjunction to provide the user with optimal investment support.

[0430] The information gathering module collects stock-related data such as stock prices, company earnings reports, economic news, and technical indicators. The information processing module analyzes the collected data and extracts useful features for investment (trend analysis and risk assessment). The trigger module notifies the user or initiates automated trading when specific conditions are met (e.g., stock price exceeding a certain value or technical indicator reaching a threshold). The information generation module generates and provides personalized stock reports to the user based on the analysis results. The automated trading module automatically buys and sells based on set conditions. The interface module provides information to the user in an interactive or report format.

[0431] We will now explain an example of how this system works, following a specific investment scenario. This system first collects stock data from financial markets in real time or as it becomes available. This includes stock price data, news articles, and corporate financial data. The information collection module retrieves data from stock exchanges and financial data providers via APIs. For example, it collects the stock price, financial information (EPS, P / E ratio, etc.), and related news (earnings announcements, etc.) of a company "Company A" that the user is monitoring. In addition, technical indicators such as moving averages (MA), relative strength index (RSI), and MACD are also retrieved.

[0432] The information processing module analyzes collected stock price data and news to evaluate specific trends and risk factors. For example, it uses technical indicators to assess market overheating and extracts the basis for investment decisions. Specifically, if Company A's stock price rises above a certain period's moving average, it may be judged as a buy signal, or if the latest earnings report exceeds market expectations, it may be analyzed as positive news and incorporated into investment decisions. The information processing module weights data based on past trading history and the user's investment style, and performs analysis optimized for individual investment decisions.

[0433] The trigger module activates based on user-defined conditions. It monitors real-time market data and activates when it detects changes in stock prices or technical indicators. For example, when Company A's stock price rises 10% above its 7-day moving average, the system detects a "buy signal" and activates the trigger. Once the trigger is activated, the user receives a notification and interactive advice regarding buying and selling timing.

[0434] After the trigger is activated, the system automatically generates an investment report. The report includes information such as stock price trends (price fluctuations and technical indicators over the past 30 days), the latest financial results and their analysis, sentiment analysis of related news articles, and overall market trends (e.g., performance compared to the S&P 500). Specifically, for example, if Company A's stock price exceeds a certain technical indicator and the trigger is activated, the system will notify the user that "Company A's stock price is 10% above the moving average, so it's a good time to buy," and provide a detailed report.

[0435] The automated trading module executes trades based on conditions set by the user in advance. For example, if a user has set it to "buy 100 shares of Company A at a specified price," the system will automatically execute the order according to that instruction. Specifically, if the system detects a "buy signal" and the conditions set by the user are met, an automated order for 100 shares will be sent to the stock exchange. The risk management module can also monitor the entire portfolio and adjust trading volume and order prices to avoid excessive risk.

[0436] After an automated trade is completed, the interface module notifies the user of the trade results and the latest portfolio status. It also provides interactive feedback, allowing the user to ask further questions or make further trades.

[0437] This stock investment support system enables investors to make quick and accurate investment decisions because the system analyzes data in real time and provides timely buy and sell signals. Furthermore, since the system automatically executes buy and sell orders based on user-defined conditions, it saves effort while achieving optimal trading efficiency through automation. Regarding risk management, the risk management module ensures that trades are executed according to the user's risk tolerance, optimizing the overall portfolio risk.

[0438] The following describes an investment information provision system that provides specialized investment information (expert reports) for specific stocks, which are volatile commodities, to a specific investor XX in real time, using Figures 10 to 16 as an embodiment of the system.

[0439] Figure 10 illustrates the conventional technology of investment information provision systems. It is an investment information provision service published on the Internet 4 by an investment service provider. Figure 10 displays stock information for the company "XXX Electric (6XXX)". "XXX Electric" represents the stock name, and "(6XXX)" is a four-digit number that identifies the stock. Entering a specific company name, stock code, etc., into the search box in the upper right corner of Figure 10 will display stock price data for that specific stock, as well as charts illustrating stock price fluctuations, as shown in Figure 10.

[0440] Users of investment information services include free users who use the service free of charge, and paid users who pay a predetermined amount to use the investment information service for a set period. Generally, free users have limited access to certain functions, and the data displayed is outdated compared to paid users. Figure 10 shows a screen for free users, so 2674.5 yen is displayed as the real-time stock price, but the date information is November 1st. In reality, the date and time when this free user accessed the screen in Figure 10 was November 4th, 2024, at 9:19 AM, but it can be seen that the data displayed is three days old.

[0441] In the middle section of Figure 10, there are multiple selection buttons for choosing the display function (Details, Bulletin Board, Chart, Timely Disclosure, News, Company Information, Dividend). In Figure 10, the "Details" button is pressed, and detailed information about XXX Electric's stock price (multiple stock price data and stock price chart) is displayed.

[0442] Clicking the "Chart" button in the function selection menu displays a higher-resolution chart. Clicking the "Timely Disclosure" button displays "Important Company Information" disclosed by each listed company. Clicking the "News" button displays news information related to investment. Clicking "Company Information" displays company information for that stock (fiscal year end, average annual salary, representative's name, etc.). Clicking the "Dividend" button displays dividend information for that company, such as the dividend per share (company forecast).

[0443] Pressing the "Bulletin Board" button in the function selection menu will display information from a bulletin board related to that stock. The bulletin board is operated by an investment service provider that publishes investment information on the internet, as shown in Figure 10. Users of this service can post various comments about the stock price of XXX Electric. While some user comments may contain useful information, the vast majority of comments are personal opinions or speculative information, making them very unreliable.

[0444] Many investors need reliable information about stock prices to guide their investment decisions, so investment service providers operate message boards to provide users with various kinds of information. Ideally, it would be desirable for investment experts to provide users with expert information and advice based on specific data. However, because the number of stock issues on the Tokyo Stock Exchange is approximately 4,000, and stock price information changes constantly, it has been impossible to provide users of investment support services with timely expert information created by experts.

[0445] Therefore, using Figures 11 to 16, we will explain an investment support service that can provide high-quality, expert information on investment products to users of the investment support service who are investors, in a timely manner, and the technology that makes this possible.

[0446] Figure 11 illustrates an investment support service that enables an information generation system to provide users with expert reports, similar to those provided by experts, in near real-time, based on the latest stock price information, for all stocks handled by the Tokyo Stock Exchange. The contents illustrated in Figures 11 to 16 are displayed on the display unit 2C of the user terminal 2 used by the user, or printed via the user terminal 2. In addition, while the contents illustrated in Figures 11 to 16 are displayed on the display unit 2C of the user terminal 2 used by the user, System 1, which is an information generation system explained using Figures 1 to 4, is running in the background.

[0447] In the middle section of Figures 11 to 16, similar to Figure 10, there are multiple selection buttons for selecting display functions (Details, Bulletin Board, Chart, Timely Disclosure, News, Company Information, Dividend). Furthermore, Figures 11 to 16 also include multiple selection buttons for selecting new display functions (Expert Report, Technical Analysis, Trend Analysis, Comparison with Last Week's Closing Price, 25-Day Moving Average Deviation Analysis, Chart Analysis, Signal Analysis).

[0448] First, using Figure 11, we will explain the state in which user XX clicks the "Expert Report" button and the "Expert Report" is displayed on the display unit 2C of user terminal 2. When the expert report button is clicked, System 1 begins the following task. (1) Use the Tokyo Stock Exchange's API service to collect the latest stock price information for XXX Electronics. (2) Collect information about XXX Electric (performance, business, etc.) that is publicly available on the internet. (3) Numerical indicators such as various technical indicators (SMA, EMA, RSI, MACD, etc.) and trend indicators (trend strength, trend direction, 25-day moving average deviation rate, difference between last week's closing price and the current price, etc.) are calculated from the latest stock price information of XXX Electric. Note that technical indicators and trend indicators are evaluation indicators for the stock market and specific stocks, which are volatile commodities. (4) Using a "conversion means that converts specific numerical data into character information," the numerical data of each calculated numerical indicator is used to obtain character information that represents the meaning and state of each numerical indicator. (5) Obtain the format specification information for the expert report (summary version). (6) Obtain the format specification information for the expert report (detailed version). (7) The information generation module 40 generates an expert report (summary version) and an expert report (detailed version) based on the information from (1) to (6). (8) Display the expert report (summary version) on the display unit 2C of terminal 2.

[0449] Here, we explained the method of "(1) using the Tokyo Stock Exchange's API service to collect the latest stock price information for XXX Electronics," but it is not always necessary to collect the latest stock price information. For example, if you have a system that periodically (for example, every 15 minutes) downloads data for all stocks in CSV format using the Tokyo Stock Exchange's API service and stores it in a database, you do not need to collect the latest stock price information for XXX Electronics as described in (1) above. You can substitute this by "extracting the stock price information for XXX Electronics from the data downloaded periodically."

[0450] The specific contents of the expert report (summary version) are as follows: (A) Current stock price situation and technical analysis The latest stock price for XXX Electronics (6XXX) is 2,645 yen (closing price on October 30, 2024). This price indicates a short-term correction and has fallen below medium-term and shorter-term moving averages. This information suggests that market participants' anxieties remain, and concerns about future volatility may be influencing stock prices. (B) Technical Indicator Analysis The MACD is at -1930.06, above the signal line of -2246.19, suggesting a slight upward trend. However, caution is advised as the trend strength remains weak. - Bollinger Bands: The current stock price is located between -1σ and -2σ, suggesting a possibility of a short-term rebound.

[0451] This expert report shows that the latest stock price is 2,645 yen (closing price on 2024-10-30). This latest stock price is the price at 3:00 PM on 2024-10-30, immediately before user XX asked the question (2024-10-30 15:38:46). This latest price data was obtained after user XX clicked the "Expert Report button" and the stock price information was retrieved. In this case, the data retrieval date and time information is 2024-10-30 15:39:23. In this case, the user who clicked the "Expert Report" button was Mr. XX, and time information such as Question Date and Time: 2024-10-30 15:38:46, Data Acquisition Date and Time: 2024-10-30 15:39:23, Answer Date and Time: 2024-10-30 15:43:30, and Time Taken to Answer: 0:04:44.467804 is also displayed. System 1 took 4 minutes and 44 seconds to answer, completing the information provision in a short time.

[0452] Thus, providing investors with expert report information, including the latest stock price data, multiple technical indicator values, written explanations of the meaning of those technical indicator values, and explanations of the stock's trends based on those technical indicators, within minutes has a significant impact.

[0453] The information illustrated in Figure 11 was provided to user XX, but it can also be provided to other users. For example, suppose the user information storage unit 71 of System 1 stores the email address of YY, a paid member who owns 3,000 shares of XXX Electric, and the email addresses of ZZ, a free member who owns 2,000 shares of XXX Electric. Based on this information, the information processing module 20 can send a summary version of the expert report provided to XX to the email address of free member ZZ, and send a detailed version of the expert report generated for XX to the email address of paid member YY.

[0454] At the bottom of Figure 11, there are function selection buttons such as "View Details," "View Entire Summary," "Download Details," and "Send Details via Email." When the user clicks the "View Details" button, the detailed version of XXX Electronics' expert report is displayed on the display unit 2C of the user terminal 2. The detailed version of the expert report contains information as illustrated in Figures 6 to 9. When the user clicks the "View Entire Summary" button, the entire content of the summary version of XXX Electronics' expert report is displayed on the display unit 2C of the user terminal 2. When the user clicks the "Download Details" button, an HTML or PDF file of the detailed version of XXX Electronics' expert report is downloaded. When the user clicks the "Send Details via Email" button, the detailed version of XXX Electronics' expert report is sent to user XX's email address. The email address of user XX is the one stored in the user information storage unit 71.

[0455] In this way, on the investment service screen provided by the investment service provider, an investor user can click a function selection button to receive multiple types of expert reports on a specific stock in real time, generated by an information generation AI based on the latest stock price information. Moreover, this expert report information includes calculated technical indicators such as MACD and trend indicators such as Bollinger Bands. Furthermore, it not only includes numerical data for technical indicators, but also natural language explanations of what those technical indicator values ​​mean, and even natural language explanations of future trends based on what those technical indicator values ​​mean.

[0456] Furthermore, the search box shown in the upper right of Figure 11 is labeled "Enter stock name, code, keyword, question, or instruction," prompting users to input questions and instructions in natural language text, in addition to the stock name, code, and keyword. For example, users can enter instructions such as "Create a document that provides a detailed explanation of XXX Electric's stock price" or questions such as "What is the current stock price of XXX Electric?" into this search box. The investment support service provider or investment information provider can then respond to the instructions or questions by sending a summary of an expert report or detailed information of an expert report on the specific stock, XXX Electric, as shown in Figure 11, to the user terminal 2 and displaying it on the display unit 2C, or by sending it to the email address of the user who made the instructions or questions. In addition, numerical information and natural language information, as shown in Figures 12 to 16, can be provided to the investor question...

Claims

1. An information generation system that generates information about fluctuating products, Information gathering means for collecting raw information on fluctuating products from external or internal information sources, An information generation system characterized by comprising: an information generation means that generates generated information relating to variable products based on the raw information collected by the information collection means.

2. In the information generation system described in claim 1, An information generation system characterized by having a distribution means that automatically distributes the generated information generated by the information generation means to a user when predetermined distribution conditions are met.

3. In the information generation system described in claim 1, It has a user information storage means that stores user-specific information for a specific user, The information generation system is characterized in that the information generation means generates the generated information based on the user-specific information stored in the user information storage means and the raw information collected by the information collection means.

4. In the information generation system described in claim 3, The aforementioned user-specific information includes user attribute information and request information. The information generation system is characterized in that the information generation means generates the generated information based on both the attribute information and the request information.

5. In the information generation system described in claim 3, The aforementioned user-specific information includes the user's behavioral history and holdings information. The information generation system is characterized in that the information generation means generates the generated information based on both the behavioral history information and the held stock information.

6. In the information generation system described in claim 3, The aforementioned user-specific information includes information about the delivery schedule set by the user. An information generation system characterized by having a distribution means for distributing generated information generated by the information generation means on a daily, weekly, or monthly basis based on the aforementioned distribution schedule.

7. In the information generation system described in claim 3, The aforementioned user-specific information includes information about the user's question. The aforementioned source information includes time-series data, The information generation system is characterized in that the information generation means analyzes the content of the question using vector search technology and generates the generated information based on the time series data or the results of the analysis of the time series data.

8. In the information generation system described in claim 3, The aforementioned user-specific information includes information about the user's selections or instructions regarding the chapter structure of the report. The information generation system is characterized in that the information generation means generates the generated information such that it includes a report having a chapter structure selected or instructed by the user.

9. In the information generation system described in claim 3, The aforementioned user-specific information includes the user's holdings, investment style, risk tolerance, or areas of interest. The information generation system is characterized in that the information generation means generates the generated information such that the chapter structure is dynamically changed based on the user's holdings, investment style, risk tolerance, or areas of interest.

10. In the information generation system described in claim 3, The aforementioned user-specific information includes user attribute information, risk tolerance, or securities holdings. The information generation system is characterized in that the information generation means generates the generated information such that it includes a report in which the weights of technical indicators and fundamental analysis are adjusted based on the user's attribute information, risk tolerance, or holdings.

11. In the information generation system according to any one of claims 1 to 10, The aforementioned source information includes time-series data, The information generation means is characterized by generating the generated information when the time-series data exceeds a predetermined threshold.

12. In the information generation system according to any one of claims 1 to 10, The aforementioned source information includes technical indicators, The information generation system is characterized in that the information generation means generates the generated information such that it includes a report containing text that changes dynamically based on the numerical values ​​of the technical indicators.

13. In the information generation system according to any one of claims 1 to 10, The information generation means is characterized in that it starts generating the information in response to a question or inquiry from a user.

14. In the information generation system according to any one of claims 1 to 10, The aforementioned information gathering means also collects other source information in addition to the source information relating to the fluctuating product, The information generation system is characterized in that, in addition to the first process of generating generated information relating to the fluctuating product, the information generation means also performs a second process of generating other generated information based on the other raw information collected by the information collection means.

15. In the information generation system according to claim 14, The information generation system is characterized in that the information gathering means collects the other raw information from relevant external or internal information sources, depending on the type of other generated information generated by the information generation means performing the second processing.

16. In the information generation system according to claim 14, The information generation means is characterized in that the second processing is performed by a computer executing a program to which a different algorithm is applied for each type of other generated information.

17. In the information generation system according to any one of claims 1 to 10, The aforementioned information gathering means collects the raw information from multiple information sources, The information generation means is characterized by generating generated information based on a single source of information that is generated based on a plurality of source pieces of information obtained from the plurality of source pieces of information.

18. In the information generation system according to claim 17, The information generation means is characterized by generating generated information based on a single source of information that is generated by assigning different weights to each of the multiple source pieces of information obtained from the multiple source pieces of information.

19. In the information generation system according to claim 14, The information generation means is characterized by generating generated information that integrates the generated information generated by the first process and the generated information generated by the second process.

20. In the information generation system according to any one of claims 1 to 10, The information generation system is characterized in that the information gathering means collects the raw information after receiving a question or inquiry from a user.

21. In the information generation system according to any one of claims 1 to 10, The information generation means is an information generation system characterized by creating and distributing generated information in the order of information collection date and time, information generation date and time, and viewing date and time, starting from the date and time of the user's request.

Images