Information processing program, information processing device, and information processing method

Abstract

An information processing program, an information processing device, and an information processing method are provided that can assist a company in responding to a questionnaire created in an arbitrary format. [Solution] The information processing program relating to the disclosed technology causes a computer to execute a questionnaire acquisition procedure (ST110) for acquiring questionnaire data, a conversion procedure (ST120) for using the questionnaire data to acquire data indicating a modified questionnaire in which the format of the questionnaire identified by the questionnaire data has been modified into a predetermined format, and a response acquisition procedure (ST140) for acquiring data indicating responses to questions included in the modified questionnaire using company information held by a predetermined company for answering questions included in the modified questionnaire and data indicating the modified questionnaire.

Description

[Technical Field]

[0001] The present disclosure relates to an information processing program, an information processing device, and an information processing method. [Background technology]

[0002] Conventionally, techniques have been proposed that enable companies to use computers to obtain various information from their business partners. For example, Patent Document 1 discloses a technique that enables a business entity (company) in a supply chain to use a computer to obtain information related to greenhouse gas target reduction rates or greenhouse gas emissions from multiple business entities that make up the upstream or downstream processes of the supply chain. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7518573 Summary of the Invention [Problem to be solved by the invention]

[0004] As disclosed in Patent Document 1, there have been technologies in the past that allow a company in a supply chain (hereinafter referred to as the "acquiring company") to use a computer to obtain various information from a company (hereinafter referred to as the "acquiring company") located upstream or downstream in the supply chain. However, for example, if the acquiring company uses a computer system different from that used by the acquiring company, it may be difficult for the acquiring company to obtain information via the system. In such cases, the acquiring company may send the acquiring company data such as a specified form or questionnaire (hereinafter collectively referred to as the "questionnaire") and request that the acquiring company enter responses to the questionnaire. Here, the acquiring company may send a wide variety of questionnaires to the acquiring company as data, and the formats may be arbitrary, which may make it a burden for the acquiring company to respond to these questionnaires.

[0005] The disclosed technology is intended to solve the above-mentioned problems, and provides an information processing program, an information processing device, and an information processing method that can assist companies in responding to questionnaires created in any format. [Means for solving the problem]

[0006] The information processing program according to the present disclosure includes a questionnaire acquisition step for acquiring questionnaire data, and a modified questionnaire in which a format of a questionnaire specified by the questionnaire data is modified into a predetermined format using the questionnaire data acquired in the questionnaire acquisition step. a modified questionnaire including a question field and an answer field indicating a correspondence relationship with the question field; a questionnaire conversion procedure for acquiring data indicating the modified questionnaire; company information held by a predetermined company, the company information being used to answer questions included in the modified questionnaire identified by the data acquired in the questionnaire conversion procedure; and data indicating the modified questionnaire acquired in the questionnaire conversion procedure. By inputting it into a machine learning model , included in the modified questionnaire Responding to questions Answers to questions The answer corresponding to the answer columnand a response acquisition procedure for acquiring data indicating the answer. [Effects of the Invention]

[0007] The information processing program according to the disclosed technology has the above-described technical features and has the effect of being able to assist companies in responding to questionnaires created in any format. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a block diagram illustrating an example of the configuration of an information processing system according to the disclosed technology. [Figure 2] FIG. 10 is a diagram illustrating a data structure for an information processing program according to the disclosed technology. [Figure 3] FIG. 10 is a diagram showing an example of a questionnaire identified by questionnaire data in the disclosed technology. [Figure 4] FIG. 10 is a diagram showing an example of an altered questionnaire identified by altered questionnaire data in the disclosed technology. [Figure 5] FIG. 10 is a diagram showing an example of a modified questionnaire in which answers and reasons for the answers have been entered according to the disclosed technology. [Figure 6] 1 is a diagram illustrating an example of operation of an information processing device according to the disclosed technology and the flow of information. [Figure 7] FIG. 1 is a diagram illustrating a hardware configuration of an information processing device according to the present disclosure. [Figure 8] 1 is a flowchart illustrating a processing procedure of an information processing method according to the disclosed technique. DETAILED DESCRIPTION OF THE INVENTION

[0009] The disclosed technology can be applied to surveys in various business fields, but here we will explain the technology in particular as it is applied to a survey on ESG, which has recently become an important indicator for companies aiming for sustainable growth. For example, in recent years, ESG management has attracted global attention as a management policy essential for sustainable corporate growth, and an increasing number of companies are working to implement it. Furthermore, various disclosure systems have been established in countries and regions around the world that require companies involved in activities within their region to disclose information in accordance with disclosure standards from an ESG (an acronym for Environment, Social, and Governance) perspective. There are also ESG rating organizations that evaluate whether organizations are taking ESG into consideration in their activities. In this way, ESG is gaining attention, just like the SDGs. Furthermore, in addition to the aforementioned "ESG management," phrases that include ESG, such as "ESG investment," "ESG report," and "ESG analytics," are becoming more commonly used.

[0010] Some important terms used in the present disclosure are defined as follows:

[0011] ESG Evaluation Items ESG evaluation items are defined as items related to ESG, and even if the descriptions of the items differ depending on the ESG evaluation organization, if there is a common concept or idea, the evaluation items that share that common concept or idea are considered to be the same item. ESG evaluation items include, for example, four items: ESG classification, code (number), major classification, and minor classification. The ESG classification for ESG evaluation items indicates whether each piece of data falls into the category of Environment, Social, or Governance, and letters such as "E," "S," and "G" are used to represent the ESG classification. The codes (numbers) in the ESG evaluation items are used to distinguish each subcategory, and a different code is assigned to each subcategory. These codes that distinguish between subcategories are also called "classification codes." The major and minor categories of ESG assessment items are written names that allow people to easily understand the content of the data representing ESG information. These major and minor categories are generalized or standardized versions of the ESG assessment items used by each ESG assessment organization. Examples of major categories include "general environment," "climate change," "water," "resource circulation," "biodiversity," "supplier (environment)," "environmental opportunities and impacts," "human rights," "labor practices," "diversity," "human resource development," "employee health and safety," "corporate citizenship," "product quality and product safety," "corporate governance," "risk management," and "information security and privacy." Examples of minor categories include "environmental management policy" under "general environment," and "climate change policy," "policy to address climate change," "policy to reduce energy consumption," and "carbon pricing" under "climate change."

[0012] <ESG evaluation information> In this specification, the term "ESG evaluation information" may include the following two types of information: One is the results of evaluations by each rating agency for each company, broken down by ESG evaluation item (hereinafter referred to as "Company-specific Evaluation Information"). The evaluation results by each rating agency included in the Company-specific Evaluation Information include, for example, whether the evaluation criteria adopted by each rating agency are met or not (hereinafter referred to as "satisfied / not met or not met"), and the ESG rating and score assigned by each rating agency (hereinafter simply referred to as "score"). The other is information indicating the evaluation criteria adopted by each evaluation institution in response to each ESG evaluation item (hereinafter referred to as "evaluation criteria information"). ESG evaluation information includes at least evaluation information for each company. The company evaluation information and evaluation criteria information may be stored for each company in a company information database, which will be described later, in association with the corresponding ESG evaluation items.

[0013] Evaluation Agency Rating agencies are third-party organizations that evaluate companies' ESG initiatives. Specifically, rating agencies include the Dow Jones Sustainable Index (DJSI), FTSE, MSCI, Sustainalytics, CDP, etc. There are two types of evaluation methods used by rating agencies: the so-called questionnaire type, in which the rating agency interviews or surveys the target company, and the so-called arbitrary rating type, in which the rating agency evaluates without conducting interviews or surveys. DJSI and CDP use the so-called questionnaire type evaluation method, while FTSE, MSCI, and Sustainalytics use the so-called arbitrary rating type evaluation method.

[0014] <Disclosure Standards and Information> Disclosure standards are standards that stipulate the obligation to disclose ESG information through laws, regulations, or frameworks. Examples of disclosure standards include the GRI Standards and the ISSB (IFRS Foundation Sustainability Disclosure Standards). Disclosure standards are generally compiled as standards documents. The aforementioned evaluation criteria adopted by each evaluation agency are formulated based on these disclosure standards, and the two are closely related. Disclosure standards information is electronic data of standards. Disclosure standards information for disclosure standards established by various laws and regulations or frameworks is made public on the Internet.

[0015] Embodiment 1 The first embodiment shows a case where the disclosed technology is realized as an information processing device 100. 1 is a block diagram showing an example of the configuration of an information processing system according to the disclosed technique. As shown in FIG. 1, the information processing system is a system in which an information processing device 100 and a user terminal UT are connected via the Internet. 1 shows three user terminals UT, but the number of user terminals UT may be any number equal to or greater than 1. Also, in FIG. 1, the information processing device 100 and the user terminal UT are connected via the Internet, but this connection is not limited to the Internet and any telecommunications line may be used.

[0016] Information processing device 100 The information processing device 100 supports the response creation work of a company to which a user belongs (hereinafter referred to as a "user company") when creating responses to a questionnaire. A user company is, for example, a company located in an upstream or downstream process in a supply chain, and is a acquiring company that responds to a questionnaire sent as data from an acquiring company. The questionnaire that the user company responds to is, for example, a questionnaire about ESG. The information processing device 100 supports user companies in creating answers to questionnaires, for example, using a RAG (Retrieval Augmented Generation) mechanism, and includes a questionnaire acquisition unit 110, a questionnaire conversion unit 120, a search unit 130, and an answer acquisition unit 140. The information processing device 100 also includes a company information database 150. Note that the information processing device 100 is not limited to this configuration, and the company information database 150 may be provided in an external device that can be connected to and communicated with from the information processing device 100, such as a server managed by a user company. The information processing device 100 provides users with a service related to the disclosed technology, namely, a service that uses the RAG mechanism to assist user companies in creating answers to questionnaires (hereinafter referred to as the "answer creation support service"). The information processing device 100 is, for example, any computer such as a server or a terminal used by a user. The answer creation support service may be provided, for example, in the form of SaaS (Software as a Service). In this case, the information processing device 100 is owned or managed by a provider of the answer creation support service (hereinafter referred to as a "service provider"). This specification provides details of a case where the information processing device 100 according to the disclosed technology is, for example, a server owned or managed by a service provider, and the information processing device 100 includes a corporate information database 150.

[0017] The information processing device 100 has an item table (not shown) in addition to the aforementioned corporate information database 150, etc. The item table is a table in which each ESG evaluation item, the disclosure standards corresponding to those items, the evaluation standards of each evaluation organization corresponding to those items, and the importance of those items are described in association with each other. The item table is set in advance by the service provider and is stored in, for example, a storage unit (not shown) that the information processing device 100 can access. The ESG evaluation items are created, for example, by a service provider. The disclosure standards are created, for example, by a service provider by collecting disclosure standard information published on the Internet and analyzing the correspondence between each disclosure standard and the ESG evaluation items. The importance indicates the degree of importance given to each ESG evaluation item among multiple ESG evaluation items, and is set, for example, by the service provider using a default calculation formula, or by a user receiving the response creation support service. The importance included in the item table is set by default among these.

[0018] <<Corporate Information Database>> The company information database 150 holds company information held by companies including user companies. The company information held by the company information database 150 includes, for example, information on company names, ESG disclosure information of companies related to ESG evaluation items, CDP (Credit Default Swap) information, and ESG evaluation information, which is evaluation information of companies by ESG evaluation organizations. ESG rating agencies are organizations that evaluate whether companies are taking ESG into consideration when conducting their activities. Corporate information also includes, for example, information disclosed by each company as information corresponding to each indicator stipulated by various disclosure systems that require disclosure in accordance with ESG-related disclosure standards.

[0019] Disclosure Information The disclosure information held in the company information database 150 is disclosure information of one or more companies. The disclosure information of each company is obtained, for example, by crawling. Crawling is a well-known technique in which a program called a crawler periodically visits multiple websites to obtain and store information.

[0020] Here, for example, first, the information processing device 100 periodically visits the websites of multiple companies designated in advance by the service provider to acquire electronic data of disclosure documents of the multiple companies. The information processing device 100 associates the acquired electronic data of the companies' disclosure documents with information identifying the companies that have disclosed the data (hereinafter referred to as "company identification information") and stores the data in the company information database 150. The company identification information is, for example, the official name of the company, the abbreviation of the company, an ID or code specific to the company, etc. Hereinafter, the company identification information is assumed to be the company name such as the official name or abbreviation of the company. Furthermore, the information processing device 100 may store the acquired electronic data of the company's disclosure materials in the company information database 150 in association with information that identifies the disclosure materials (hereinafter referred to as "material identification information"). The document identification information is, for example, the official name of the disclosed document, the abbreviation of the disclosed document, the type of disclosed document, or an ID or code specific to the type of disclosed document. Hereinafter, the electronic data of the disclosed document is stored in association with the document identification information, and the document identification information is the name of the disclosed document, such as the official name or abbreviation. The electronic data of the acquired disclosure documents may be stored in the company information database 150 in a manner that overwrites the electronic data of the disclosure documents already stored for a particular company, or the electronic data of the newly acquired disclosure documents may be added to the electronic data of the disclosure documents already stored, for example, after linking it to information on the time when the electronic data of each disclosure document was acquired.

[0021] The electronic data of the disclosure materials stored in the company information database 150 is processed so that it can be easily handled by the information processing device 100 according to the technology disclosed herein. For example, machine learning technology may be used to process the electronic data of the disclosure materials stored in the corporate information database 150. Specifically, when the electronic data of the disclosure materials is input, a trained model may be used that has been trained to output, for each corresponding portion, the content itself of a portion corresponding to one of the subcategories of the ESG evaluation items (hereinafter referred to as the "corresponding portion"), information indicating the location where the corresponding portion is written, information on a directory or link that points to the description location as a transition destination (hereinafter referred to as the "link destination information"), and a subcategories code that identifies the subcategories of the ESG evaluation items to which the corresponding portion corresponds. The trained model may be, for example, an artificial neural network or a large-scale language model. Hereinafter, information indicating the location of description and link destination information will be collectively referred to as "location information." Information indicating the location of description includes the name of the disclosed document (document name), the page number where the corresponding section in the disclosed document is described, or the paragraph number on the page indicated by the page number (for example, a number indicating the paragraph from the top on that page), or the number of a figure or table, etc.

[0022] The disclosed technology may use a large language model (LLM) among trained models as a means for extracting information from disclosed materials. When a large-scale language model is used, for example, the information processing device 100 inputs electronic data of disclosure documents that have been newly acquired by crawling and stored in the corporate information database 150 into the large-scale language model. The prompt at this time may include, for example, the electronic data of the disclosure documents themselves or information specifying the electronic data of the disclosure documents, and the information indicating the ESG evaluation items themselves or information specifying the item table. Furthermore, the prompt may include instructions to extract corresponding sections corresponding to any of the subcategories of the ESG evaluation items and their locations, instructions to generate link destination information that generates different links for each location and links the locations to the corresponding sections, instructions to generate information linking each link to each subset code (hereinafter referred to as "item link link information"), and instructions including specific link generation rules, etc. Prompts to be input to a large-scale language model can be automatically generated. For example, a prompt template is set by a service provider and stored in a storage unit (not shown) accessible to the information processing device 100. When processing electronic data of a disclosure document, the information processing device 100 can automatically generate the above-mentioned prompt by retrieving the template from the storage unit and embedding necessary information, such as information specifying the electronic data of the disclosure document.

[0023] The large-scale language model may be provided in an external server or in the information processing device 100. For example, an existing large-scale language model provided as a service by an external server may be used. In this case, the information processing device 100 can input a prompt to the large-scale language model and obtain a response to the input from the large-scale language model through API (Application Programming Interface) cooperation.

[0024] When the information processing device 100 receives an output from the trained model that includes corresponding locations, location information, and minor category codes corresponding to minor categories of ESG evaluation items, the information processing device 100 associates the output with company-specific information and stores the associated output as disclosure information in the company information database 150. Note that, for example, one disclosure document for one company may typically include multiple pieces of disclosure information. Therefore, the company information database 150 typically stores multiple pieces of disclosure information for one company, each associated with the other. In this manner, in this embodiment, the "disclosure information" is specified by including at least the corresponding location, location information, and classification of the ESG evaluation items (here, minor classification code). Here, in the above example, the classification of the ESG evaluation items is specified by a classification code (minor classification code), but the classification of the ESG evaluation items may also be specified by a classification name and a classification code. Furthermore, the corresponding section may include any of a sentence, a figure, a table, or a page in the disclosure document. The corresponding section may be the entire description of a certain sentence or a portion of that description. For example, the corresponding section may be identified as a description of a portion of a paragraph among multiple paragraphs or as a description of multiple paragraphs. The diagrams as the specific corresponding portions may be, for example, all or some of the diagrams themselves when there are multiple diagrams. Furthermore, the page as the corresponding portion may be the description of all pages when there are multiple pages, or the description of some pages. The corresponding portion may include multiple types of text, figures, or pages. For example, the corresponding portion may be identified as a part of text (e.g., a paragraph) and parts of multiple figures, a part of a page among multiple pages and a part of a figure, or a page among multiple pages and a paragraph on that page.

[0025] The location information may also include the name of the disclosed material, the page or page and paragraph where the corresponding section in the disclosed material is described, and a link to the page or page and paragraph where the corresponding section is described. In this case, the page can be specified by the page number, and the paragraph can be specified by the page number and the paragraph number on the page indicated by the page number (for example, a number indicating the number of the paragraph from the top on the page).

[0026] The electronic data of the disclosure materials may be processed using a trained model as described above, a so-called rule-based algorithm, or manually. Regardless of the processing method used, it is sufficient that the resulting disclosure information is stored in the corporate information database 150.

[0027] Prior to the search described below, the corporate information database 150 stores in advance, in an organized state, the corresponding locations, location information, and classifications of ESG evaluation items (in this case, minor classification codes) for the disclosed information for each company using the method described above. An example of a data structure for an information processing program according to the disclosed technology is shown in Figure 2. As shown in Figure 2, the data may include, as data elements for each piece of disclosure information, "company name," "document name," "text described in the document," "diagram described in the document," "link information to the described content (page, paragraph number, etc.)," "ESG classification," "ESG evaluation item (major classification)," "ESG evaluation item (minor classification)," and "ESG evaluation item (code)." Of these, "Company Name," "Document Name," "ESG Classification," "ESG Evaluation Item (Major Category)," "ESG Evaluation Item (Minor Category)," and "ESG Evaluation Item (Code)" are as described above. "Text in document" refers to the content of the text itself when the corresponding part of the disclosure information is a sentence or a page. Therefore, this data element stores character data or the like that indicates the content of the text itself. "Diagrams in documents" refers to the content of a diagram or table in the corresponding location of the disclosure information, if that diagram or table is the relevant location. Therefore, this data element stores image data or the like showing the diagram or table itself. "Link information for the content (page, paragraph number, etc.)" is location information. As mentioned above, location information includes information indicating the location of the corresponding section and link destination information, so this data element stores a link to the disclosed information and information on where the disclosed information is located.

[0028] Alternatively, the disclosure information may be stored in the company information database 150, for example, as follows. For example, first, the information processing device 100 periodically visits the websites of multiple companies designated in advance by a service provider to acquire disclosure information from the multiple companies. The information processing device 100 converts the acquired disclosure information from the companies into text data using optical character recognition (OCR), divides the converted text data into predetermined ranges, and converts the text data for each divided range into vectors. The information processing device 100 then associates the vectors obtained by the conversion with information identifying the companies disclosing the disclosure information (company identification information), and stores the vectors in the company information database 150. If the disclosure information is created in PDF format, text data may already exist in the disclosure information. In this case, the information processing device 100 may not need to convert the disclosure information into text data using the optical character recognition (OCR) described above. Note that the disclosure information of multiple companies acquired by the information processing device 100 may be created in HTML format. When dividing the text data into predetermined ranges, the information processing device 100 can divide the data by page, paragraph, or number of characters, for example. When dividing the text data by number of characters, the information processing device 100 can set an arbitrary number of characters, such as a maximum of 400 characters, as the division unit. In this case, the information processing device 100 may divide the text data so that an arbitrary number of characters, such as 100 characters, overlap before and after the divided range.

[0029] <ESG evaluation information> As described above, the ESG evaluation information includes company-specific evaluation information and evaluation criteria information. Of the ESG evaluation information, original information on the evaluations made by each evaluation institution for each company (hereinafter referred to as "original evaluation information"), which is necessary for creating company-specific evaluation information, is uploaded to the information processing device 100, for example, by a company as a user having the original evaluation information operating a user terminal UT. In other words, the original evaluation information is information indicating the evaluations made by each evaluation institution for each company before being classified by ESG evaluation item. The process of extracting the evaluations made by each evaluation institution in the original evaluation information by ESG evaluation item, that is, the process of extracting evaluation information by company (hereinafter referred to as "extraction process") may be performed manually or may be performed using machine learning technology by the information processing device 100. The procedure by which the information processing device 100 performs the extraction process is referred to as the "evaluation information extraction procedure." When machine learning technology is used, specifically, a trained model may be used that has been trained to separate the ratings made by each rating agency into ESG evaluation items when the original rating information is input. The trained model may be constructed using an artificial neural network or a large-scale language model (LLM) such as a generative pre-trained transformer (GPT).

[0030] The prompts input to the large-scale language model include, for example, the original rating information itself or information specifying the original rating information, and information indicating ESG evaluation items itself or information specifying an item table. The information specifying the original rating information is, for example, information indicating the directory and file name of the original rating information when the original rating information is stored in a storage unit (not shown) accessible to the information processing device 100. Similarly, the information specifying the item table is information indicating the directory and file name of the item table. Furthermore, the prompt contains instructions to extract corresponding sections from the original evaluation information, which are sections containing descriptions corresponding to the ESG evaluation items, and to extract evaluation results such as scores contained in the corresponding sections. Note that the processing based on these instructions may be performed multiple times using multiple types of prompts. For example, first, only the corresponding part is extracted using a prompt that describes only the instruction to extract the corresponding part, and then, the evaluation results, such as scores, included in the extracted corresponding part are extracted using a prompt that describes the instruction to extract the evaluation results. The prompt may also include an instruction to extract evaluation criteria information contained in the corresponding portion. Prompts can be automatically generated when a large-scale language model is used as the trained model. For example, prompt templates are set by a service provider and stored in a storage unit (not shown) accessible to the information processing device 100. When the information processing device 100 performs the task of separating the ratings performed by each rating agency in the original rating information by ESG rating item, the information processing device 100 can automatically generate the above-described prompts by retrieving the templates from the storage unit and embedding necessary information, such as information specifying the original rating information.

[0031] The large-scale language model may be provided on an external server or may be included in the information processing device 100. For example, an existing large-scale language model provided as a service by an external server may be used. In this case, the information processing device 100 can input a prompt to the large-scale language model and obtain a response to the input from the large-scale language model through API cooperation.

[0032] The extraction process may be performed using a trained model, a rule-based algorithm, or entirely by hand, as described above. Alternatively, the extraction process may be performed semi-automatically, using a trained model such as the large-scale language model described above, and then finally confirmed by a human through visual inspection. Furthermore, among the ESG evaluation information, the evaluation criteria information may be extracted in association with the ESG evaluation items in the extraction process as described above and stored in the company information database 150, or after the company-specific evaluation information is created, the evaluation criteria information may be created in association with the ESG evaluation items by referring to the item table and stored in the company information database 150. This operation may be performed automatically by the information processing device 100 or may be performed manually.

[0033] Survey Acquisition Section 110 The questionnaire acquisition unit 110 acquires questionnaire data from the user terminal UT. The questionnaire data is data indicating a questionnaire to be answered by the user. The questionnaire includes, for example, one or more question fields and answer fields corresponding to the question fields. A user can operate the user terminal UT to access the information processing device 100 and display a screen for using the answer creation support service on a display (not shown) of the user terminal UT. For example, if the answer creation support service is provided in the form of SaaS, the screen is displayed on a browser. For example, a questionnaire data registration window can be displayed on the screen. The user can operate the user terminal UT to input information for identifying any survey data into this survey data registration window. When the user inputs information for identifying any survey data into the survey data registration window, the user terminal UT transmits the survey data identified by the information input by the user to the information processing device 100.

[0034] For example, when a user responds to a questionnaire sent by Company A, a business partner of the user's company, the user operates the user terminal UT to input information indicating the storage location in the user terminal UT of the questionnaire data sent from Company A into the questionnaire data registration window. When the user inputs the above information into the questionnaire data registration window, the user terminal UT transmits the questionnaire data stored in the storage location indicated by the above information to the information processing device 100. The survey acquisition unit 110 of the information processing device 100 acquires the survey data transmitted from the user terminal UT. Upon acquiring the survey data transmitted from the user terminal UT, the survey acquisition unit 110 outputs the acquired survey data to the survey conversion unit 120. Note that the survey data also includes information (company identification information) that identifies the company to which the user who transmitted the survey data belongs.

[0035] Survey conversion unit 120 The questionnaire conversion unit 120 acquires questionnaire data from the questionnaire acquisition unit 110. When the questionnaire conversion unit 120 acquires the questionnaire data from the questionnaire acquisition unit 110, it uses the acquired questionnaire data to acquire data indicating an altered questionnaire in which the format of the questionnaire specified by the questionnaire data has been altered into a predetermined format (hereinafter referred to as "altered questionnaire data"). Specifically, first, the questionnaire conversion unit 120 checks the data format of the acquired questionnaire data. If the data format of the acquired questionnaire data is not a predetermined data format, the questionnaire conversion unit 120 analyzes the questionnaire data and acquires the analysis results. Here, the predetermined data format is a data format that allows a first machine learning model, into which the questionnaire data is input, to process the questionnaire data. The first machine learning model will be described later. Here, the predetermined data format is assumed to be, for example, PDF (Portable Document Format).

[0036] If the data format of the acquired questionnaire data is not PDF as a result of checking the data format of the questionnaire data, the questionnaire conversion unit 120 performs the following analysis, for example. For example, if the data format of the questionnaire data is a table format created using spreadsheet software such as Excel, the questionnaire conversion unit 120 analyzes the questionnaire data using a known library (e.g., openpyxl) for analyzing information in cells contained in a table, and extracts information regarding the arrangement of the question and answer fields contained in the questionnaire, as well as information regarding the content of the questions contained in the questionnaire (e.g., question text, etc.), as analysis results. Furthermore, when the data format of the questionnaire data is a document format created using word processing software such as Word, the questionnaire conversion unit 120 analyzes the questionnaire data using a known library, in the same way as when the data format of the questionnaire data is a table format, and extracts information regarding the content of the questions included in the questionnaire as the analysis result. Furthermore, if the data format of the questionnaire data is not the predetermined data format, PDF, and the questionnaire data contains image data, the questionnaire conversion unit 120 converts the data format of the questionnaire data into the predetermined data format, PDF. For example, if the data format of the questionnaire data is the above-mentioned table format created by spreadsheet software such as Excel, or an office format including a document format created by word processing software such as Word, and the questionnaire data contains image data, the questionnaire conversion unit 120 uses known data format conversion software to convert the data format of the questionnaire data into the predetermined data format, PDF.

[0037] Next, the questionnaire conversion unit 120 generates a prompt (hereinafter referred to as an "modification prompt") that instructs the user to modify the format of the questionnaire identified by the acquired questionnaire data, and that is to be input to the first machine learning model. Specifically, the questionnaire conversion unit 120 generates a modification prompt that instructs the user to modify the format of the questionnaire identified by the questionnaire data into a format that can be processed by the second machine learning model described below. In the following description, for simplicity, the questionnaire after its format has been modified will be referred to as an "modified questionnaire."

[0038] Here, the modification prompt includes, for example, the following content: (1) Questionnaire data acquired by the questionnaire acquisition unit 110 (2) When the data format of the questionnaire data is not PDF, the analysis result of the questionnaire data by the questionnaire conversion unit 120 (3) Instructions to identify the answer fields of the questionnaire identified by the above questionnaire data and the question fields corresponding to the answer fields (4) An instruction to identify the category of each question included in the questionnaire identified by the above questionnaire data and to output the identified category. (5) An instruction to specify the answer format for each question included in the questionnaire specified by the above questionnaire data and to output the specified answer format. (6) An instruction to output data showing the modified questionnaire so that the number of question columns and answer columns in the modified questionnaire are the same.

[0039] The questionnaire conversion unit 120 inputs a modification prompt including the above contents (1) to (6) into the first machine learning model. Note that the questionnaire conversion unit 120 may input at least one of the above (1) and (2) directly into the first machine learning model without including it in the modification prompt. Furthermore, if the data format of the questionnaire data acquired from the questionnaire acquisition unit 110 is a predetermined data format (here, PDF), the questionnaire conversion unit 120 may omit the above-described analysis of the questionnaire data. In this case, the questionnaire conversion unit 120 may generate an alteration prompt including at least the contents of (3) to (6) excluding the analysis result of the questionnaire data exemplified in (2) above, as an alteration prompt, and input the generated alteration prompt to the first machine learning model. This is because, if the data format of the questionnaire data is a predetermined data format, the first machine learning model can analyze the questionnaire data as is. In other words, the first machine learning model is trained in advance so as to be able to handle questionnaire data created in a predetermined data format (here, PDF). On the other hand, if the data format of the questionnaire data is not a predetermined data format, the questionnaire conversion unit 120 may analyze the questionnaire data as described above and input the analysis result to the first machine learning model.

[0040] The first machine learning model may be, for example, an artificial neural network, and there may be multiple first machine learning models. Furthermore, if there are multiple first machine learning models, the first machine learning models may include a large-scale language model (LLM). In this case, the first machine learning model is a machine learning model trained to output data indicating a modified questionnaire as a response when given the above-described modification prompt as input (hereinafter referred to as "modified questionnaire data").

[0041] The first machine learning model including the large-scale language model may be provided in an external server or in the information processing device 100. Furthermore, the large-scale language model may be, for example, an existing large-scale language model provided as a service by an external server. In this case, the information processing device 100 can input a modification prompt to the large-scale language model through API cooperation and obtain a response to the input from the large-scale language model. The first machine learning model generates altered questionnaire data based on the given alteration prompt and outputs the generated altered questionnaire data to the information processing device 100. The questionnaire conversion unit 120 of the information processing device 100 acquires the altered questionnaire data output from the first machine learning model. Upon acquiring the altered questionnaire data from the first machine learning model, the questionnaire conversion unit 120 outputs the acquired altered questionnaire data to the search unit 130 and the response acquisition unit 140.

[0042] An example of a questionnaire identified by the questionnaire data is shown in Fig. 3. Also, an example of a modified questionnaire identified by the modified questionnaire data is shown in Fig. 4. 3, the questionnaire includes a plurality of question fields in which questions are written, answer fields corresponding to each question field, a description field for writing the answers in text, a field for inputting the department of the person in charge of answering the questions, and a field for inputting the name of the person in charge. The modified questionnaire includes, for example, a field (coordinate) indicating the question identifier, a question number field (question_number), a field (question_category) in which the question category is written, a field (question_text) in which the question text is written, and a field (answer_options) in which answer options are written. When questionnaire data indicating a questionnaire such as that shown in Fig. 3 is input together with the modification prompt described above, the first machine learning model outputs modified questionnaire data indicating a modified questionnaire such as that shown in Fig. 4. Although not shown in Fig. 4, in the modified questionnaire such as that shown in Fig. 4, an answer column for entering answers may be provided to the right of the column in which answer options are described, and a column for describing the reasons for the answers may be provided further to the right of that. 4, a field for describing the answer format may be provided to the right of the field for describing the question. The answer format indicates, for example, the format in which the user will answer, and may be, for example, a selection format using radio buttons or a multiple-choice format using check boxes.

[0043] When the first machine learning model receives the questionnaire data and the modification prompt as input, it may generate modified questionnaire data in accordance with the modification prompt so that the number of question fields is equal to the number of answer fields for writing answers. Note that in this case, it is assumed that the number of answer fields in the questionnaire identified by the questionnaire data is equal to the total number of questions in the questionnaire. Furthermore, if the first machine learning model analyzes the input questionnaire data and obtains an analysis result indicating that one question field contains multiple questions, the modified questionnaire may divide the question field and answer field for each of these questions, and generate modified questionnaire data in which the question field and answer field are divided for each question. 3 and the modified questionnaire shown in FIG. 4 are merely examples, and the questionnaire and the modified questionnaire may be generated with content and formats other than those described above. However, it is desirable that the modified questionnaire be in at least a format that allows the second machine learning model described below to accept and process the input when modified questionnaire data is input to the second machine learning model.

[0044] Search Section 130 The search unit 130 acquires the modified questionnaire data from the questionnaire conversion unit 120. When the search unit 130 acquires the modified questionnaire data from the questionnaire conversion unit 120, it acquires company information from the company information database 150 to answer the questions included in the modified questionnaire. For example, the search unit 130 acquires information indicating questions included in the modified questionnaire (hereinafter referred to as "question information") based on the acquired modified questionnaire data, and generates a search query for searching for company information necessary to answer the questions included in the modified questionnaire based on the acquired question information. Here, the question information includes, for example, text data describing the question sentence and text data describing the question category. The search unit 130 can acquire the text data describing the question sentence from the above-mentioned field (question_text) in which the question sentence is described, and can acquire the text data describing the question category from the above-mentioned field (question_category) in which the question category is described. When the search unit 130 generates the search query, it performs a vector search of the company information database 150 based on the generated search query, and obtains, as the search result, company information necessary for responding to the modified questionnaire.

[0045] The search unit 130 may arbitrarily set an upper limit on the number of pieces of company information to be acquired from the company information database 150 based on the generated search query as a search condition when searching the company information database 150. For example, the search unit 130 may set an upper limit on the number of pieces of company information to be acquired as a search condition, such as acquiring, as search results, up to the top 10 pieces of company information that have a high rate of matching the search query. Here, the company information acquired by the search unit 130 includes, for example, information about the company name, disclosure information disclosed by the company, CDP information, and evaluation information about the company by an ESG evaluation organization. Each piece of information included in the company information may be assigned predetermined tag information. In this case, the search unit 130 may acquire company information that matches the search query based on the tag information. The search unit 130 outputs the acquired company information to the response acquisition unit 140. In the following explanation, for simplicity, the company information acquired from the company information database 150 and necessary to respond to the modified questionnaire will be referred to as "specific company information."

[0046] 《Answer acquisition part 140》 The response acquisition unit 140 acquires the modified questionnaire data from the questionnaire conversion unit 120. The response acquisition unit 140 also acquires the specific company information from the search unit . The response acquisition unit 140 acquires modified questionnaire data from the questionnaire conversion unit 120 and acquires specific company information from the search unit 130, and then uses the acquired specific company information and modified questionnaire data to acquire data indicating answers to questions included in the modified questionnaire. Specifically, the answer acquiring unit 140 generates a prompt (hereinafter referred to as an "answer prompt") that includes an instruction to generate an answer to a question included in the acquired modified questionnaire and is to be input to the second machine learning model. At this time, the answer acquiring unit 140 may include in the answer prompt an instruction to cause the second machine learning model to output the reason for the generated answer. In addition, if the answer format of the questions included in the modified questionnaire is multiple choice, that is, if the answer format requires the respondent to select an appropriate option from multiple options, the answer acquisition unit 140 may include information about the multiple options in the answer prompt.

[0047] When the answer acquisition unit 140 generates the answer prompt, it inputs the generated answer prompt, the modified questionnaire data acquired from the questionnaire conversion unit 120, and the specific company information acquired from the search unit 130 into a second machine learning model. Note that the answer acquisition unit 140 may input the answer prompt, the modified questionnaire data, and the specific company information individually into the second machine learning model, or, if the modified questionnaire data and the specific company information can be included in the answer prompt, it may generate an answer prompt by including the modified questionnaire data and the specific company information, and input the generated answer prompt into the second machine learning model.

[0048] The second machine learning model may be, for example, an artificial neural network, or there may be multiple second machine learning models, and if there are multiple second machine learning models, the second machine learning models may include a large-scale language model (LLM). In this case, the second machine learning model is a machine learning model trained to output data indicating answers to questions included in the modified questionnaire (hereinafter referred to as "answer data") when given the above-mentioned answer prompt, modified questionnaire data, and specific company information as input. Note that the second machine learning model may also be a machine learning model trained to output data indicating the reason for the answer (hereinafter referred to as "answer reason data") in addition to the above-mentioned answer data when given the above-mentioned answer prompt, modified questionnaire data, and specific company information as input.

[0049] The second machine learning model including the large-scale language model may be provided in an external server or in the information processing device 100. Furthermore, the large-scale language model may be, for example, an existing large-scale language model provided as a service by an external server. In this case, the information processing device 100 can input answer prompts, modified questionnaire data, and specific company information to the large-scale language model through API linkage, and can also obtain responses to the input from the large-scale language model. The second machine learning model generates answer data, or answer data and answer reason data, based on the given answer prompt, the modified questionnaire data, and the specific company information, and outputs the generated data to the information processing device 100. The answer acquisition unit 140 of the information processing device 100 acquires the answer data, or the answer data and answer reason data, output from the second machine learning model.

[0050] In the above description, an example has been described in which the answer acquiring unit 140 acquires answer data, or answer data and answer reason data, from the second machine learning model. However, the answer acquiring unit 140 may acquire, from the second machine learning model, modified questionnaire data in which these answers, or answers and reasons for the answers, are entered in the modified questionnaire. In other words, the answer acquiring unit 140 may acquire, from the second machine learning model, modified questionnaire data including answer data, or answer data and answer reason data. In this case, the answer acquiring unit 140 may include, in the answer prompt provided to the second machine learning model, an instruction to cause the second machine learning model to output modified questionnaire data in which answers are entered in the answer field, or modified questionnaire data in which answers are entered in the answer field and reasons for the answers are entered in the reason field. Furthermore, the above-described first machine learning model and second machine learning model may be the same machine learning model, and the same machine learning model may be the same large-scale language model (LLM). Furthermore, in the above example, the search unit 130 may perform a search using a technique called "few shot." For example, the search unit 130 generates only a company name as a search query and acquires company information matching the search query from the company information database 150. The answer acquisition unit 140 may then include all of the company information acquired by the search unit 130 in an answer prompt, prompting the second machine learning model to generate answer data, or answer data and answer reason data. In this case, the information processing device 100 can acquire optimal answer data, etc., using only a small amount of information (here, company names).

[0051] Fig. 5 shows an example of a modified questionnaire in which answers and reasons for the answers have been entered. As shown in Fig. 5, the modified questionnaire includes, for example, a field (question_id) indicating the question identifier, a question number field (question_number), a field (question_category) describing the question category, a field (question_sentence) describing the question sentence, a field (answer_format) describing the answer format, a field (options) describing answer options, an answer field (answer), and a reason field (reason) for describing the reason for the answer. Also, as shown in Fig. 5, answers to questions are entered in the answer field, and reasons for the answers are entered in the reason field. The reason for the answer, as shown in Figure 5, describes which page of the company information the reason for the answer entered in the answer field is described on, in other words, which page of the company information the answer was based on. For example, the first line, which displays "1" as the question identifier, describes the question category as "1. Environment (E) Area / 1 GHG Initiatives (Energy)," and the question text as "Is your company taking steps to reduce energy consumption?" The answer to this question is a radio button selection format, with the options being "Yes" or "No." The answer to this question is "Yes," and the reason given is "The Corporate Governance Report 2023 (p. 12) states that the company is promoting the expansion of renewable energy use and energy-saving activities." This reason is generated based on the specific company information acquired by the search unit 130. Additionally, lines 2 through 5, labeled "2.1" through "2.4," display separate options for the same question. Lines 2 through 5 display the question category as "1. Environment (E) Area / 1. GHG Initiatives (Energy)," and the question reads, "Q2 is for business partners who answered "Yes" to Q1. What initiatives are you taking? Please select all that apply." The answer format for this question is a multiple-choice checkbox format, with the options being "1. We have set a total CO2 target within the company," "2. We are implementing PDCA management toward the total target," "3. We have publicly announced our goal of achieving carbon neutrality by 2050," and "4. We are building a company-wide system to strengthen our environmental initiatives in the medium and long term." The answer to each of these options is "Yes," and the reasons for selecting each option are described. These reasons are also generated based on the specific company information acquired by the search unit 130.

[0052] Note that the reasons for answers shown in Figure 5 are merely examples, and as long as the specific reasons for the answers are clear, they do not have to be written in sentences, for example. Also, the reason column for writing the reasons for the answers, as shown in Figure 5, is not an essential component of the modified questionnaire, and may be omitted. However, by providing a reason column for writing the reasons for the answers in the modified questionnaire, the user can easily understand the reasons for the answers. Therefore, it is preferable to provide a reason column in the modified questionnaire. 5 illustrates an example in which the second machine learning model outputs modified questionnaire data in which answers and reasons for the answers are entered into the modified questionnaire. However, the second machine learning model may output only the answer data, or only the answer data and reason for answer data, in association with a question identifier or question number, rather than inputting the answers and reasons for the answers into the modified questionnaire and then outputting the modified questionnaire data. In this case, the answer acquiring unit 140 acquires only the answer data, or only the answer data and the answer reason data, from the second machine learning model, and the acquired answer data, or the answer data and the answer reason data, may be integrated into the modified questionnaire data. That is, the answers, or the answers and the reasons for the answers, may be input into the modified questionnaire. In this case, the data integration may be performed manually by, for example, a provider of the answer creation support service (service provider), or may be performed by a data integration unit (not shown) included in the information processing device 100. Furthermore, the answer acquisition unit 140 may output the answer data or the modified questionnaire data in which the answer data and the answer reason data are integrated to the user terminal UT. In this case, the user terminal UT acquires the modified questionnaire data from the answer acquisition unit 140. When the user terminal UT acquires the modified questionnaire data from the answer acquisition unit 140, it displays the modified questionnaire identified by the acquired data on a display of the user terminal UT. For example, the user terminal UT displays the modified questionnaire on a screen for using the answer creation support service that is displayed on the display.

[0053] Next, an example of the operation of the information processing device 100 in the information processing system and the flow of each piece of information will be described with reference to FIG. 6. In the following description, it is assumed that survey data has been sent in advance to the user company from multiple business partner companies. In the following description, it is assumed that a company information database 150 has been constructed in advance by the service provider and the information processing device 100. In addition, it is assumed that the first machine learning model and the second machine learning model described above are different machine learning models. First, the user operates the user terminal UT to input information for identifying any questionnaire data into the above-mentioned questionnaire data registration window. When the user inputs the above information into the questionnaire data registration window, the user terminal UT transmits the questionnaire data identified by the input information to the information processing device 100 (reference numeral (1) in FIG. 6). Next, in the information processing device 100, the questionnaire acquisition unit 110 acquires the questionnaire data transmitted from the user terminal UT (reference number (2) in FIG. 6). Upon acquiring the questionnaire data from the user terminal UT, the questionnaire acquisition unit 110 outputs the acquired questionnaire data to the questionnaire conversion unit 120. Next, the questionnaire conversion unit 120 acquires the questionnaire data from the questionnaire acquisition unit 110. The questionnaire conversion unit 120 uses the questionnaire data acquired from the questionnaire acquisition unit 110 to acquire data (altered questionnaire data) indicating an altered questionnaire in which the format of the questionnaire specified by the data has been altered into a predetermined format. Specifically, the questionnaire conversion unit 120 generates the above-mentioned modification prompt and inputs the generated modification prompt to the first machine learning model 160 (reference number (3) in FIG. 6). The first machine learning model 160 generates modified questionnaire data based on the input modification prompt and outputs the generated modified questionnaire data to the information processing device 100 (reference number (4) in FIG. 6). The questionnaire conversion unit 120 of the information processing device 100 acquires the modified questionnaire data output from the first machine learning model. Upon acquiring the modified questionnaire data from the first machine learning model, the questionnaire conversion unit 120 outputs the acquired modified questionnaire data to the search unit 130 and the response acquisition unit 140. Next, the search unit 130 acquires the modified questionnaire data from the questionnaire conversion unit 120, and acquires company information (specific company information) for answering the questions included in the modified questionnaire from the company information database 150 (reference number (5) in FIG. 6). The search unit 130 outputs the acquired specific company information to the response acquisition unit 140. Next, the response acquisition unit 140 acquires the modified questionnaire data from the questionnaire conversion unit 120, and acquires the specific company information from the search unit 130. The response acquisition unit 140 uses the acquired specific company information and the modified questionnaire data to acquire data indicating answers to questions included in the modified questionnaire. Specifically, the answer acquisition unit 140 generates an answer prompt including instructions to generate an answer to a question included in the acquired modified questionnaire, and inputs the generated answer prompt to the second machine learning model 170 (symbol (6) in Figure 6). The second machine learning model 170 generates answer data based on the input answer prompt and outputs the generated answer data to the information processing device 100 (reference number (7) in FIG. 6). At this time, the second machine learning model 170 may generate answer reason data along with the answer data and output the generated data to the information processing device 100. Furthermore, the second machine learning model 170 may include the answer data, or the answer data and the answer reason data, in the modified questionnaire data and output it to the information processing device 100. The answer acquiring unit 140 acquires the data output from the second machine learning model 170. Although not shown in FIG. 6, the answer acquiring unit 140 may output the data acquired from the second machine learning model 170 to the user terminal UT. The information processing device 100 operates as explained above in the symbols (1) to (7), and transmits and receives each piece of data. Here, the RAG mechanism is realized by the symbols (3) to (7).

[0054] In the above example, the information processing device 100 transmits and receives information to and from the second machine learning model 170 once. However, when the information processing device 100 uses a large-scale language model as the second machine learning model 170, the information processing device 100 may transmit and receive information to and from the large-scale language model multiple times (multi-turn). For example, in the first turn of the information processing device 100, the answer acquisition unit 140 generates an answer prompt as described above and inputs the generated answer prompt to the large-scale language model. Then, the answer acquisition unit 140 acquires answer data from the large-scale language model. In the second turn, the answer acquisition unit 140 generates a prompt for the large-scale language model, which prompts the large-scale language model regarding the probability or success or failure of the answer indicated by the answer data generated in the first turn, and inputs the generated prompt to the large-scale language model. The answer acquisition unit 140 then acquires information regarding the probability or success or failure of the answer output from the large-scale language model, and outputs the acquired information to the user terminal UT. In this case, the user can not only easily create an answer to the question, but also easily understand the probability or success or failure of the answer, and can appropriately modify the answer depending on the probability or success or failure of the answer.

[0055] Furthermore, when outputting answer data obtained from the second machine learning model 170 to the user terminal UT, the answer acquisition unit 140 may output multiple pieces of answer data to the user terminal UT as answer candidates. For example, after outputting answer data to the user terminal UT, the information processing device 100 collects feedback from the user regarding the answer data and stores information regarding the collected feedback as a log in a storage unit (not shown) in advance. The feedback from the user includes, for example, corrected answer data or other answer data that are possible responses to the answer data. When outputting the answer data obtained from the second machine learning model 170 to the user terminal UT, the answer acquisition unit 140 refers to the log stored in the storage unit, generates multiple pieces of answer data based on the log, and outputs the generated answer data to the user terminal UT as answer candidates. In this case, the information processing device 100 can present multiple answer candidates to the user at one time, and can reduce the number of times data is sent and received to the user terminal UT, leading to a reduction in communication volume. The information processing device 100 may also have a function for allowing a user to modify an answer. For example, the information processing device 100 may be configured to receive a modification of the answer data from the user via the user terminal UT before the answer acquisition unit 140 outputs the answer data obtained from the second machine learning model 170 to the user terminal UT. The answer acquisition unit 140 may then output the answer data modified by the user to the user terminal UT. In this case, the user can modify the answer data in advance on the information processing device 100 before acquiring the answer data, which leads to improved efficiency in the answer creation work. Furthermore, when the information processing device 100 receives a correction to the answer data from the user, the information processing device 100 may use the corrected answer data as re-learning data for the second machine learning model 170. In this case, the second machine learning model 170 becomes able to output highly accurate answer data that better reflects the user's intention, and the information processing device 100 can obtain, from the second machine learning model 170, highly accurate answer data that better reflects the user's intention.

[0056] Also, in the above description, an example has been described in which the search unit 130 acquires information (question information) indicating questions included in the modified questionnaire based on the modified questionnaire data acquired from the questionnaire conversion unit 120, and generates a search query for searching for company information necessary to answer the questions included in the modified questionnaire based on the acquired question information. However, the search unit 130 is not limited to this, and may acquire question information based on questionnaire data acquired by the questionnaire acquisition unit 110, and generate a search query based on the acquired question information. In other words, the search unit 130 may generate a search query based on data indicating a questionnaire before its format was altered, rather than data indicating the modified questionnaire after its format has been altered. In the above description, an example has been described in which the search unit 130 searches for company information stored in the company information database 150. However, the search unit 130 is not limited to this, and may search for, for example, response data from past responses to questionnaires by user companies, or response data and response reason data (hereinafter, these are referred to as "responded data"). In this case, the responded data may be stored in advance in the company information database 150 as included in the company information or as information separate from the company information. When the company information database 150 is included in the information processing device 100, the answered data may be uploaded to the information processing device 100 by, for example, the user operating the user terminal UT, and stored in the company information database 150. When the company information database 150 is provided on a server managed by the user company, the answered data may be stored in the company information database 150 provided on the server by, for example, the user operating the user terminal UT. When the answered data is stored in the company information database 150 as information separate from the company information, the search unit 130 outputs the company information (specific company information) and the answered data acquired from the company information database 150 to the answer acquisition unit 140. After generating the answer prompt, the answer acquisition unit 140 inputs the generated answer prompt, the modified questionnaire data acquired from the questionnaire conversion unit 120, and the specific company information and the answered data acquired from the search unit 130 to a second machine learning model. Note that the answer acquisition unit 140 may input the answer prompt, the modified questionnaire data, the specific company information, and the answered data individually to the second machine learning model, or, when the modified questionnaire data, the specific company information, and the answered data can be included in the answer prompt, the answer acquisition unit 140 may generate an answer prompt by including the modified questionnaire data, the specific company information, and the answered data, and input the generated answer prompt to the second machine learning model. Furthermore, the answer acquiring unit 140 may include in the answer prompt an instruction to output answer data, or answer data and answer reason data, to the second machine learning model with reference to the answered data. This increases the likelihood that the second machine learning model will output answer data, or answer data and answer reason data, that reflects the content of the answered data, and the answer acquiring unit 140 can acquire answer data, or answer data and answer reason data, that reflects the content of answers given by the user company in the past.

[0057] In the above description, the search unit 130 generates a search query by itself, searches the company information database 150 based on the generated search query, and acquires company information necessary to respond to the modified questionnaire. However, the search unit 130 is not limited to this, and may instruct a large-scale language model (LLM) (not shown) to generate a search query and search for company information. For example, when the search unit 130 obtains modified questionnaire data from the questionnaire conversion unit 120, it generates a prompt (hereinafter referred to as a "search prompt") that includes an instruction to obtain the company information necessary to answer the modified questionnaire from the company information database 150, and that is to be input into the large-scale language model described above. After generating the search prompt, the search unit 130 inputs the generated search prompt and the modified questionnaire data obtained from the questionnaire conversion unit 120 into the large-scale language model. Note that the search unit 130 may input the search prompt and the modified questionnaire data separately into the large-scale language model, or, if the modified questionnaire data can be included in the search prompt, may generate the search prompt by including the modified questionnaire data, and input the generated search prompt into the large-scale language model. In response to the search prompt, the large-scale language model generates a search query for searching for company information necessary for answering the modified questionnaire, for example, and performs a vector search of the company information database 150 based on the generated search query to obtain the company information necessary for answering the modified questionnaire as the search result. The search unit 130 then outputs the company information obtained by the large-scale language model to the response acquisition unit 140 as specific company information. In this case, the search unit 130 can have the large-scale language model perform the search process without performing the search process itself, which is expected to reduce the processing load compared to when the search unit 130 performs the search process itself.

[0058] In the above description, the search unit 130 generates a search query by itself, searches the company information database 150 based on the generated search query, and obtains company information necessary to respond to the modified questionnaire, or the search unit 130 instructs the large-scale language model to perform these processes. However, in the information processing device 100, it is not necessarily required that the search unit 130 or the large-scale language model perform search processing. For example, the information processing device 100 can acquire in advance from the company information database 150 company information held by a predetermined company, which is used to answer questions included in a modified questionnaire identified by data acquired by the questionnaire conversion unit 120, and further, if the answer acquisition unit 140 can acquire answer data using the acquired information, search processing by the search unit 130 or a large-scale language model becomes unnecessary. In this case, the search unit 130 may be omitted from the information processing device 100.

[0059] In the above description, an example has been described in which a user inputs information for identifying desired survey data into a survey data registration window displayed on the display of the user terminal UT. Alternatively, the user may specify survey data by operating the user terminal UT, for example, by selecting desired survey data from a pull-down menu displayed in the survey data registration window. When the user selects survey data from the pull-down menu, the user terminal UT transmits the selected survey data to the information processing device 100. The pull-down menu is a menu that displays a list containing multiple survey data when the user presses a specific position, and may be, for example, a menu that displays multiple survey data that the user terminal UT has previously acquired as options.

[0060] 6, it is assumed that the search unit 130 was unable to acquire the specific company information necessary to answer the question from the company information database 150, or that the information was acquired but was insufficient. In such cases, if the second machine learning model 170 is a large-scale language model, the answer acquisition unit 140 may include in the answer prompt an instruction to have the large-scale language model output a message to the company user such as "The answer may be insufficient because the company information necessary to answer the question is insufficient. Please enter input to supplement the answer as necessary," and input this answer prompt to the large-scale language model. Furthermore, when the second machine learning model 170 is a large-scale language model and the above-mentioned answer prompt is input to the large-scale language model, the large-scale language model may generate answer data by including information indicating a message such as "The answer may be insufficient because the company information necessary to answer the question is insufficient. Please enter input to supplement the answer if necessary."

[0061] Furthermore, when outputting answer data acquired from the second machine learning model 170 to the user terminal UT, if the answer format of the answer indicated by the answer data is not a predetermined answer format, the answer acquisition unit 140 may correct the answer format so that the answer format of the answer indicated by the answer data acquired from the second machine learning model 170 matches the predetermined answer format, and then output the answer data to the user terminal UT. The predetermined answer format may be, for example, the answer format described in the field (answer_format) in which the answer format is described in the modified questionnaire. Furthermore, when outputting the answer data acquired from the second machine learning model 170 to the user terminal UT, the answer acquisition unit 140 may output the answer data in a manner that allows the user to accept input from the user terminal UT regarding the answer indicated by the answer data. This is to leave room for the user to input (edit) the answer, since it is not certain whether the answer indicated by the answer data is the answer the user desires.

[0062] In this case, when the user makes some input in response to the answer using the user terminal UT, the user terminal UT may transmit information indicating the content input in response to the answer (hereinafter referred to as "additional input information") to the information processing device 100. In the information processing device 100, an additional input information acquisition unit (not shown) acquires the additional input information transmitted from the user terminal UT and stores the acquired additional input information in, for example, an additional input information database (not shown). The additional input information database may be provided in the information processing device 100, or may be provided in an external device that can be communicatively connected to the information processing device 100. The additional input information stored in the additional input information database is used the next time the answer acquisition unit 140 outputs answer data to a question that is the same as or similar to the question related to the additional input information. For example, when outputting answer data acquired from the second machine learning model 170 to the user terminal UT, the answer acquisition unit 140 refers to the additional input information database. If additional input information related to the answer data to be output is stored in the additional input information database, the answer acquisition unit 140 reflects the content of the additional input information in the answer data to be output and outputs the answer data to the user terminal UT. For example, if additional input information related to the answer data to be output is stored in the additional input information database, the answer acquisition unit 140 may modify the answer data to be output based on the additional input information stored in the additional input information database and output the modified answer data, or may output the additional input information stored in the additional input information database together with the answer data to be output. This allows the answer acquisition unit 140 to present the user with an answer that is closer to the answer desired by the user.

[0063] The above process also applies when the answer data includes information indicating a message such as, "The answer may be insufficient because the company information necessary to answer the question is insufficient. Please enter additional information to supplement the answer, if necessary." In this case, the user is expected to check the message on the user terminal UT and, if necessary, enter some information in response to the answer using the user terminal UT. Even in this case, the user terminal UT transmits information indicating the content entered in response to the answer (additional input information) to the information processing device 100, and an additional input information acquisition unit (not shown) in the information processing device 100 acquires the additional input information transmitted from the user terminal UT and stores the acquired additional input information in an additional input information database (not shown). Hereafter, the answer acquisition unit 140, in a similar manner to the above, can use the additional input information when outputting answer data from the next time onward to present the user with an answer that is closer to the user's desired answer.

[0064] Embodiment 2 The second embodiment shows a case where the disclosed technology is realized as an information processing program. Unless otherwise specified, the same symbols as those used in the first embodiment are used in the second embodiment. Furthermore, in the second embodiment, explanations that overlap with those in the first embodiment are omitted as appropriate. The information processing program of embodiment 2 causes a computer to execute the following steps: a questionnaire acquisition procedure for acquiring questionnaire data; a questionnaire conversion procedure for using the questionnaire data acquired in the questionnaire acquisition procedure to acquire data indicating a modified questionnaire in which the format of the questionnaire identified by the questionnaire data has been modified into a predetermined format; a search procedure for acquiring company information for answering questions included in the modified questionnaire from a database that stores company information held by a predetermined company; and an answer acquisition procedure for using the company information acquired in the search procedure and data indicating the modified questionnaire acquired in the questionnaire conversion procedure to acquire data indicating answers to the questions included in the modified questionnaire.

[0065] 7 is a diagram showing a hardware configuration 200 of an information processing device 100 according to the present disclosed technique. As shown in FIG. 7, the hardware configuration 200 of the information processing device 100 includes a communication interface 210, an input / output interface 220, a processor 230, and a memory 240.

[0066] 8 is a flowchart showing, as processing steps, an information processing method realized by the information processing program according to the present disclosure. As shown in FIG. 8, the information processing method realized by the information processing program includes a questionnaire acquisition step ST110, a questionnaire conversion step ST120, a search step ST130, and a response acquisition step ST140. The numbers assigned to each processing step correspond to the symbols of the entities that execute the processing step. For example, the questionnaire acquisition step ST110 is executed by the questionnaire acquisition unit 110, the questionnaire conversion step ST120 by the questionnaire conversion unit 120, the search step ST130 by the search unit 130, and the answer acquisition step ST140 by the answer acquisition unit 140. As shown in FIG. 8, a questionnaire acquisition procedure ST110, a questionnaire conversion procedure ST120, a search procedure ST130, and an answer acquisition procedure ST140 are included in the loop process.

[0067] The information processing method realized by the information processing program according to the second embodiment is an information processing method executed by a computer. In this information processing method, when a loop is started, first, the questionnaire acquisition unit 110 executes a questionnaire acquisition procedure (ST110) to acquire questionnaire data transmitted from the user terminal UT. Then, the questionnaire conversion unit 120 executes a questionnaire conversion procedure (ST120) to acquire data indicating an altered questionnaire in which the format of the questionnaire identified by the questionnaire data is altered to a predetermined format, using the questionnaire data acquired in the questionnaire acquisition procedure. Then, the search unit 130 executes a search procedure (ST130) to acquire company information for answering questions included in the altered questionnaire from a company information database 150 that stores company information held by predetermined companies. Furthermore, the answer acquisition unit 140 executes an answer acquisition procedure (ST140) to acquire data indicating answers to questions included in the altered questionnaire, using the company information acquired in the search procedure and data indicating the altered questionnaire acquired in the questionnaire conversion procedure. Note that while FIG. 6 illustrates an example of loop processing, the disclosed technology is not limited to this.

[0068] The functions of the questionnaire acquisition unit 110, questionnaire conversion unit 120, search unit 130, and response acquisition unit 140 of the information processing device 100 according to the disclosed technology are realized by processing circuits. That is, the information processing device 100 includes processing circuits for executing a questionnaire acquisition procedure ST110, a questionnaire conversion procedure ST120, a search procedure ST130, and a response acquisition procedure ST140. The processing circuit is a processor 230 (also referred to as a CPU, central processing unit, processing device, arithmetic unit, microprocessor, microcomputer, or DSP) that executes an information processing program stored in a memory 240.

[0069] The functions of the questionnaire acquisition unit 110, questionnaire conversion unit 120, search unit 130, and response acquisition unit 140 are realized by software, firmware, or a combination of software and firmware. The software and firmware are written as information processing programs and stored in memory 240. The processing circuit realizes the functions of each unit by reading and executing the information processing programs stored in memory 240. That is, the information processing device 100 includes memory 240 for storing an information processing program that, when executed by the processing circuit, results in the execution of a questionnaire acquisition procedure ST110, a questionnaire conversion procedure ST120, a search procedure ST130, and a response acquisition procedure ST140. It can also be said that these information processing programs cause a computer to execute the procedures or methods of the questionnaire acquisition unit 110, the questionnaire conversion unit 120, the search unit 130, and the response acquisition unit 140. Here, the memory 240 may be, for example, a non-volatile or volatile semiconductor memory such as RAM, ROM, flash memory, or EPROM. The memory 240 may also be in the form of a disk such as a magnetic disk, a flexible disk, an optical disk, a compact disk, a mini disk, a DVD, etc. The memory 240 may also be in the form of an HDD or an SSD.

[0070] In the above description, an example has been described in which the search unit 130 searches for company information stored in the company information database 150. However, the search unit 130 is not limited to this, and may search for the above-mentioned answered data, for example, as described in the first embodiment. In this case, the answered data may be stored in advance in the company information database 150 in the form of being included in the company information or as information separate from the company information, using the procedure described in the first embodiment. Also, in embodiment 2, when the answer acquisition unit 140 generates an answer prompt, it may input the generated answer prompt, the modified questionnaire data acquired from the questionnaire conversion unit 120, and the specific company information and answered data acquired from the search unit 130 into a second machine learning model, as in embodiment 1. Furthermore, the answer acquiring unit 140 may include in the answer prompt an instruction to output answer data, or answer data and answer reason data, to the second machine learning model with reference to the answered data. This increases the likelihood that the second machine learning model will output answer data, or answer data and answer reason data, that reflects the content of the answered data, and the answer acquiring unit 140 can acquire answer data, or answer data and answer reason data, that reflects the content of answers given by the user company in the past.

[0071] In the above description, the search unit 130 generates a search query by itself, searches the company information database 150 based on the generated search query, and acquires company information necessary to respond to the modified questionnaire. However, the search unit 130 is not limited to this, and may instruct a large-scale language model (LLM) (not shown) to generate a search query and search for company information, as described in the first embodiment, for example. For example, when the search unit 130 obtains modified questionnaire data from the questionnaire conversion unit 120, it generates a prompt (search prompt) that includes an instruction to obtain the company information necessary to answer the modified questionnaire from the company information database 150, and is input into the large-scale language model described above. After generating the search prompt, the search unit 130 inputs the generated search prompt and the modified questionnaire data acquired from the questionnaire conversion unit 120 into the large-scale language model described above. In response to the search prompt, the large-scale language model generates a search query for searching for company information necessary for answering the modified questionnaire, for example, and performs a vector search of the company information database 150 based on the generated search query to obtain the company information necessary for answering the modified questionnaire as the search result. The search unit 130 then outputs the company information obtained by the large-scale language model as specific company information to the response acquisition unit 140. In this case, the search unit 130 can have the large-scale language model perform the search process without performing the search process itself, which is expected to reduce the processing load compared to when the search unit 130 performs the search process itself.

[0072] In the above description, the search unit 130 generates a search query by itself, searches the company information database 150 based on the generated search query, and acquires company information necessary to respond to the modified questionnaire, or the search unit 130 instructs the large-scale language model to perform these processes. However, even in the second embodiment, it is not necessarily required to perform search processing by the search unit 130 or the large-scale language model. For example, the information processing device 100 can acquire in advance from the company information database 150 company information held by a predetermined company, which is used to answer questions included in a modified questionnaire identified by data acquired by the questionnaire conversion unit 120, and further, if the answer acquisition unit 140 can acquire answer data using the acquired information, search processing by the search unit 130 or large-scale language model becomes unnecessary. In this case, the search step (ST130) may be omitted from the information processing program.

[0073] As described above, the information processing program of embodiment 2 causes a computer to execute a questionnaire acquisition procedure ST110 for acquiring questionnaire data, a questionnaire conversion procedure ST120 for using the questionnaire data acquired in the questionnaire acquisition procedure to acquire data indicating a modified questionnaire in which the format of the questionnaire identified by the questionnaire data has been modified into a predetermined format, and a response acquisition procedure ST140 for acquiring data indicating the responses to the questions included in the modified questionnaire using company information held by a predetermined company, which is for answering questions included in the modified questionnaire identified by the data acquired in the questionnaire conversion procedure ST120, and data indicating the modified questionnaire acquired in the questionnaire conversion procedure ST120. Therefore, the information processing program according to the second embodiment can cause a computer to execute a predetermined procedure, thereby assisting a company in responding to a questionnaire created in any format.

[0074] In addition, the information processing program of embodiment 2 causes a computer to execute a search procedure ST130 that acquires company information for answering questions included in the modified questionnaire from a company information database 150 that stores company information held by a specified company, and in an answer acquisition procedure ST140, data indicating answers to the questions included in the modified questionnaire is acquired using the company information acquired in the search procedure ST130 and data indicating the modified questionnaire acquired in the questionnaire conversion procedure ST120. As a result, the information processing program according to the second embodiment can cause a computer to execute a predetermined procedure and accurately acquire data indicating answers to questions included in the modified questionnaire.

[0075] In addition, in the information processing program according to the second embodiment, the questionnaire conversion procedure ST120 may generate a modification prompt that instructs the user to modify the format of the questionnaire identified by the questionnaire data into a predetermined format, input the generated modification prompt together with the questionnaire data into a first machine learning model, and obtain data indicating the modified questionnaire from the first machine learning model. As a result, the information processing program according to the second embodiment can cause a computer to execute a predetermined procedure and acquire data indicating an altered questionnaire whose format has been appropriately altered to a predetermined format.

[0076] Furthermore, in the information processing program according to the second embodiment, in the questionnaire conversion procedure ST120, if the data format of the questionnaire data is not a predetermined data format, before inputting the modification prompt into the first machine learning model, information regarding the arrangement of the question and answer fields included in the questionnaire identified by the questionnaire data, as well as information regarding the content of the questions included in the questionnaire, may be extracted from the questionnaire data, and the extracted information may be included in the modification prompt. As a result, the information processing program shown in embodiment 2 can cause a computer to execute a predetermined procedure and obtain data indicating a modified questionnaire whose format has been appropriately modified to a predetermined format, even if the data format of the questionnaire data is not the predetermined data format.

[0077] In addition, in the information processing program according to the second embodiment, in the questionnaire conversion procedure ST120, the modification prompt may include an instruction to output data representing the modified questionnaire to the first machine learning model so that the number of question fields and answer fields in the modified questionnaire is the same. As a result, the information processing program shown in embodiment 2 can cause a computer to execute a predetermined procedure to obtain data indicating a modified questionnaire in which the number of question fields and answer fields is the same, and can accurately obtain data indicating answers to the questions.

[0078] Furthermore, in the information processing program according to the second embodiment, in the survey conversion procedure ST120, if the data format of the survey data is not a predetermined data format and the survey data includes image data, the data format of the survey data may be converted into a predetermined data format before inputting the modification prompt into the first machine learning model, and the survey data with the converted data format may be input into the first machine learning model together with the modification prompt. As a result, the information processing program shown in embodiment 2 can cause a computer to execute a predetermined procedure and obtain data indicating a modified questionnaire whose format has been appropriately modified to a predetermined format, even if the data format of the questionnaire data is not a predetermined data format and the questionnaire data includes image data.

[0079] In addition, in the information processing program according to the second embodiment, in the search procedure ST130, a search query may be generated based on information indicating a question contained in the modified questionnaire identified by the data acquired in the questionnaire conversion procedure, and company information corresponding to the search query may be acquired from the company information database 150. As a result, the information processing program according to the second embodiment can cause a computer to execute a predetermined procedure and accurately acquire company information from the company information database 150 to answer questions included in the modified questionnaire.

[0080] In the information processing program according to the second embodiment, in the search procedure ST130, information indicating a question may be acquired from the data indicating the modified questionnaire, and a search query may be generated based on the acquired information indicating the question. As a result, the information processing program according to the second embodiment can cause a computer to execute a predetermined procedure and generate a search query for searching for company information with high accuracy.

[0081] In addition, in the information processing program of embodiment 2, in the answer acquisition procedure ST140, an answer prompt is generated, which is a prompt including an instruction to generate an answer to a question included in the modified questionnaire identified by the data acquired in the conversion procedure, and the company information acquired in the search procedure, data indicating the modified questionnaire acquired in the questionnaire conversion procedure, and the answer prompt are input into a second machine learning model, and data indicating the answer to the question included in the modified questionnaire is acquired from the second machine learning model. As a result, the information processing program according to the second embodiment can cause a computer to execute a predetermined procedure and accurately acquire data indicating answers to questions included in the modified questionnaire.

[0082] In the information processing program according to the second embodiment, in the answer acquisition procedure ST140, an instruction to output the reason for the answer to the question may be included in the answer prompt. As a result, the information processing program according to the second embodiment can cause the computer to execute a predetermined procedure and acquire data indicating the reason for the answer to the question.

[0083] In the information processing program according to the second embodiment, in the answer acquisition step ST140, if the answer format of the question is multiple choice, information about the options may be included in the answer prompt. As a result, the information processing program according to the second embodiment can cause the computer to execute a predetermined procedure, and can reflect information about the options in data indicating the answer.

[0084] Furthermore, in the information processing program according to the second embodiment, the company information acquired in the search procedure ST130 is information disclosed by the company to which the user who answers the questions included in the questionnaire belongs, and the survey data acquired in the survey acquisition procedure ST110 may be acquired from a business partner of the company to which the user belongs. As a result, the information processing program according to the second embodiment can cause a computer to execute a predetermined procedure to assist in the task of answering a questionnaire identified by questionnaire data sent from a business partner of the user company.

[0085] In addition, in the information processing program of embodiment 2, the format of the modified questionnaire may be a format that allows a second machine learning model to which data indicating the modified questionnaire has been input, process the data, and output data indicating answers to questions included in the modified questionnaire. As a result, the information processing program according to the second embodiment can cause the computer to execute a predetermined procedure, enabling the second machine learning model to output data indicating an answer.

[0086] The information processing device 100 according to the first embodiment includes a questionnaire acquisition unit 110 that acquires questionnaire data, a questionnaire conversion unit 120 that uses the questionnaire data acquired by the questionnaire acquisition unit 110 to acquire data indicating a modified questionnaire in which the format of the questionnaire identified by the questionnaire data has been modified into a predetermined format, and a response acquisition unit 140 that uses company information held by a predetermined company, the company information being used to answer questions included in the modified questionnaire identified by the data acquired by the questionnaire conversion unit 120, and the data indicating the modified questionnaire acquired by the questionnaire conversion unit 120, to acquire data indicating answers to the questions included in the modified questionnaire. Therefore, the information processing device 100 according to the first embodiment can assist a company in responding to a questionnaire created in any format.

[0087] The information processing method according to the first embodiment is an information processing method by information processing device 100, in which questionnaire acquisition unit 110 acquires questionnaire data, questionnaire conversion unit 120 uses the questionnaire data acquired by questionnaire acquisition unit 110 to acquire data indicative of an altered questionnaire in which the format of the questionnaire identified by the questionnaire data is altered into a predetermined format, and response acquisition unit 140 acquires data indicative of responses to questions included in the altered questionnaire, using company information held by a predetermined company and used to answer questions included in the altered questionnaire identified by the data acquired by questionnaire conversion unit 120, and the data indicative of the altered questionnaire acquired by questionnaire conversion unit 120. By information processing device 100 executing the above method, it is possible to assist companies in the task of answering questionnaires created in any format. [Explanation of symbols]

[0088] 100 information processing device, 110 questionnaire acquisition unit, 120 questionnaire conversion unit, 130 search unit, 140 response acquisition unit, 150 corporate information database, 160 first machine learning model, 170 second machine learning model, 200 hardware configuration, 210 communication interface, 220 input / output interface, 230 processor, 240 memory, UT user terminal.

Claims

1. A survey acquisition procedure for acquiring survey data; a questionnaire conversion step of acquiring data indicating a modified questionnaire in which the format of the questionnaire specified by the questionnaire data acquired in the questionnaire acquisition step has been modified into a predetermined format, the modified questionnaire including question fields and answer fields indicating correspondences with the question fields; and an answer acquisition step of acquiring data indicating answers to questions corresponding to the question fields included in the modified questionnaire, the data being answers to questions corresponding to the question fields included in the modified questionnaire, by inputting the company information held by a specified company, the company information being for answering questions included in the modified questionnaire identified by the data acquired in the questionnaire conversion step, and the data indicating the modified questionnaire acquired in the questionnaire conversion step, into a machine learning model; An information processing program that causes a computer to execute the above.

2. causing a computer to execute a search procedure to retrieve company information for answering questions included in the modified questionnaire from a database storing company information held by a predetermined company; In the answer acquisition procedure, acquiring data indicating answers to questions included in the modified questionnaire using the company information acquired in the search step and data indicating the modified questionnaire acquired in the questionnaire conversion step; The information processing program according to claim 1.

3. In the questionnaire conversion step, generating a modification prompt that instructs a user to modify a questionnaire format identified by the questionnaire data into the predetermined format, inputting the generated modification prompt together with the questionnaire data into a first machine learning model, and acquiring data indicating the modified questionnaire from the first machine learning model; 3. The information processing program according to claim 1.

4. In the questionnaire conversion step, If the data format of the questionnaire data is not a predetermined data format, before inputting the modification prompt into the first machine learning model, information regarding the arrangement of question fields and answer fields included in the questionnaire identified by the questionnaire data and information regarding the content of the questions included in the questionnaire are extracted from the questionnaire data, and the extracted information is included in the modification prompt.

4. The information processing program according to claim 3.

5. In the questionnaire conversion step, The modification prompt includes an instruction to output data representing the modified questionnaire to the first machine learning model so that the modified questionnaire has the same number of question fields and answer fields.

4. The information processing program according to claim 3.

6. In the questionnaire conversion step, If the data format of the questionnaire data is not a predetermined data format and the questionnaire data includes image data, convert the data format of the questionnaire data into the predetermined data format before inputting the modification prompt into the first machine learning model, and input the questionnaire data whose data format has been converted into the first machine learning model together with the modification prompt.

4. The information processing program according to claim 3.

7. In the search procedure, generating a search query based on information indicating a question included in the modified questionnaire identified by the data acquired in the questionnaire conversion step, and acquiring company information corresponding to the search query from the database; 3. The information processing program according to claim 2.

8. In the search procedure, acquiring information indicating the question from the data indicating the modified questionnaire, and generating the search query based on the acquired information indicating the question.

8. The information processing program according to claim 7.

9. In the answer acquisition procedure, generating an answer prompt, the answer prompt including an instruction to generate an answer to a question included in the modified questionnaire identified by the data obtained in the questionnaire conversion step; inputting company information for answering questions included in the modified questionnaire, data representing the modified questionnaire, and the answer prompts into a second machine learning model, and obtaining data representing answers to questions included in the modified questionnaire from the second machine learning model; The information processing program according to claim 1.

10. In the answer acquisition procedure, The answer prompt includes an instruction to output a reason for the answer to the question.

10. The information processing program according to claim 9.

11. In the answer acquisition procedure, If the question has a multiple choice answer format, the answer prompt includes information about the options.

11. The information processing program according to claim 9 or 10.

12. the company information acquired in the search procedure is information disclosed by a company to which the user who answers the questions included in the questionnaire belongs, The survey data acquired in the survey acquisition step is acquired from a business partner of the company to which the user belongs.

3. The information processing program according to claim 2.

13. The modified questionnaire format is: The format is such that the second machine learning model, to which the data indicating the modified questionnaire has been input, can process the data and output data indicating answers to questions included in the modified questionnaire.

10. The information processing program according to claim 9.

14. a survey acquisition unit that acquires survey data; a questionnaire conversion unit that uses the questionnaire data acquired by the questionnaire acquisition unit to acquire data indicating an altered questionnaire in which the format of the questionnaire specified by the questionnaire data has been altered into a predetermined format, the altered questionnaire including question fields and answer fields indicating correspondences with the question fields; and an answer acquisition unit that acquires data indicating answers corresponding to the answer fields, which are answers to questions corresponding to the question fields included in the modified questionnaire, by inputting corporate information held by a specified company, which is for answering questions included in the modified questionnaire identified by the data acquired by the questionnaire conversion unit, and data indicating the modified questionnaire acquired by the questionnaire conversion unit into a machine learning model; An information processing device comprising:

15. An information processing method by an information processing device, The survey acquisition unit acquires the survey data, a questionnaire conversion unit, using the questionnaire data acquired by the questionnaire acquisition unit, acquires data indicating an altered questionnaire in which the format of the questionnaire specified by the questionnaire data has been altered into a predetermined format, the altered questionnaire including question fields and answer fields indicating correspondences with the question fields; an answer acquisition unit inputs, into a machine learning model, company information held by a predetermined company, for answering questions included in the modified questionnaire identified by the data acquired by the questionnaire conversion unit, and data indicating the modified questionnaire acquired by the questionnaire conversion unit, thereby acquiring data indicating answers corresponding to the answer fields, which are answers to questions corresponding to the question fields included in the modified questionnaire; Information processing methods.

Images