Human Capital Management Scoring Systems and Programs

Abstract

To provide a human capital management scoring system and program that can recommend measures that are thought to be effective to users who wish to improve their evaluation of human capital management. [Solution] In the health management scoring system 10, which is a human capital management scoring system, a perspective evaluation request means 33 obtains perspective evaluation result data from a large-scale language model (LLM) that indicates whether each of multiple perspectives is met using one of two values, and further, a score calculation means 34 uses the perspective evaluation result data to calculate a health management navigation score, which is a human capital management navigation score, using TF-IDF, and an output means 45 displays information related to human capital management (health management) on the screen using the human capital management navigation score (health management navigation score).

Description

[Technical Field]

[0001] The present invention relates to a computer-based human capital management scoring system and program that presents information related to health management or other human capital management, and can be used, for example, to analyze data describing companies' health management initiatives and present measures that are thought to be effective in addressing the health management challenges each company faces, with the aim of enabling employer companies of health insurance associations to obtain certification as outstanding health management corporations or to receive a health management brand. [Background technology]

[0002] The Health and Productivity Management Survey published by the Ministry of Economy, Trade and Industry (METI) not only quantifies each company's health and productivity management status, but also provides textual information such as the details of the challenges, the results of the measures implemented, and the results of the effectiveness assessment (see Figure 2 below). The overall health and productivity management assessment is calculated by weighting four items (Aspect 1: Management Philosophy and Policy), Aspect 2: Organizational Structure, Aspect 3: Systems and Measures Implementation, and Aspect 4: Evaluation and Improvement) in a 3:2:2:3 ratio. Because it is considered difficult to significantly change the management philosophy, improving the overall assessment requires improving the assessment score for Aspect 4: Evaluation and Improvement. Furthermore, Aspect 4: Evaluation and Improvement is assessed primarily based on textual information about each company's challenges, the results of the measures implemented, and the results of the effectiveness assessment. Therefore, to receive the Certified Health and Productivity Management Organization Award or to receive the Health and Productivity Management Brand designation, it is important for each company to implement health and productivity management measures that lead to an increase in the overall assessment and Aspect 4: Evaluation and Improvement.

[0003] A dashboard system developed by the present applicant is known as a system that analyzes information related to health and productivity management and presents the results to users (see Patent Document 1). This dashboard system executes a topic estimation process for each theme using multiple response sentence data obtained from questionnaire data such as health and productivity management questionnaire data, determines the topic distribution of each response sentence data, determines a topic vector using the response sentence vector of the response sentence data with the highest topic value, calculates the similarity between the company sentence vector of multiple company sentence data created from company document data such as integrated report data and the topic vector of each sub-theme to obtain a relevance score, and displays the company sentence data on the screen using the relevance score.

[0004] The dashboard system described in Patent Document 1 performs a topic inference process, but the present invention, as will be described in detail below, differs in that it does not perform topic inference but instead uses Large Language Models (LLMs) such as a Chat Generative Pre-trained Transformer (ChatGPT) to obtain binary answers indicating whether or not each of multiple set challenges is met, and then uses these binary answers to calculate Term Frequency-Inverse Document Frequency (TF-IDF) to obtain a Health and Productivity Management Navigation Score or other Human Capital Management Navigation Score (Human Capital Management Navigation Score is a broader term than Health and Productivity Management Navigation Score).

[0005] Also known is an information processing device that enables visualization of indicators related to various performances of personnel (see Patent Document 2). This information processing device has a first acquisition unit that acquires information indicating a user's motivation from responses to a survey for the user inputted on a terminal device, a second acquisition unit that acquires information indicating the user's health condition acquired on the terminal device in the survey, a calculation unit that calculates indicators associated with the organization based on the information indicating the user's motivation and the information indicating the user's health condition, and an output unit that outputs the indicators.

[0006] The information processing device described in Patent Document 2 differs from the present invention in that it does not perform processing using a large-scale language model (LLM) or processing to calculate TF-IDF. Furthermore, this information processing device does not perform clustering processing. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Patent No. 7534511 Publication (Abstract) [Patent Document 2] JP 2024-135377 A (Abstract) Summary of the Invention [Problem to be solved by the invention]

[0008] In order to obtain the Health and Productivity Management Excellent Corporation Certification or Health and Productivity Management Brand designation, a recommendation that meets the following three conditions is required for users (companies or their personnel who wish to improve the evaluation of their company's health and productivity management initiatives).

[0009] The first condition is to recommend measures that are deemed effective in addressing the health management challenges faced by the company.

[0010] The second condition is to recommend the results of implementing a variety of measures, because it is believed that users will be able to choose a measure that is more suited to their company's situation if they can select and implement one of the various recommended options.

[0011] The third condition is to recommend the implementation results of measures that receive a high overall evaluation. This is because, in order to receive the "Health and Productivity Management Excellent Corporation Certification," a company must rank highly in the overall evaluation.

[0012] In this case, it is possible to simply recommend the results of implementing measures that address the issues of other companies that are similar to the issues faced by the company, but this method would not be able to meet the second and third of the three conditions above.

[0013] Furthermore, the overall evaluation score for health and productivity management is calculated based on the details of the issues, the results of the measures implemented, and the results of the effectiveness assessment provided by companies in the Health and Productivity Management Survey. Therefore, in order to increase the overall evaluation score, it is important to receive high evaluations for the details of the issues, the results of the measures implemented, and the results of the effectiveness assessment. However, because these are provided in text, quantitative evaluation is extremely difficult. In other words, the overall evaluation is calculated based on the standards adopted by the survey body (in the case of the Health and Productivity Management Survey, the Ministry of Economy, Trade and Industry). However, it is not always clear what specific details will result in what evaluation, or in other words, what content will receive a high evaluation. Therefore, if it were possible to quantitatively grasp the evaluation of each company's efforts provided in the text and clarify the relationship between this quantified value and the overall evaluation score, it would be possible to recommend measures that are deemed effective.

[0014] The above is not limited to health and productivity management, but also applies to surveys conducted more broadly on human capital management, where questionnaires answered by companies are collected and an evaluation is made by the surveyor based on the collected questionnaires.In this application, human capital management also includes the non-financial situation of a company, so health and productivity management is a concept included in human capital management.

[0015] An object of the present invention is to provide a human capital management scoring system and program that can recommend measures that are thought to be effective to users who wish to improve their evaluation of human capital management. [Means for solving the problem]

[0016] The present invention is a human capital management scoring system configured by a computer that presents information on health management or other human capital management, a questionnaire data storage means for storing the health management survey data or other human capital management-related survey data obtained as responses from each company to a survey on health management or other human capital management, which are associated with company response identification information for identifying the company or, in cases where multiple responses from one company are permitted, for identifying the company and the response data, including issue content data consisting of text indicating the issue content included in the health management survey data or other human capital management-related survey data, policy implementation result data consisting of text indicating the results of policy implementation, and effectiveness verification result data consisting of text indicating the results of effectiveness verification; a combining means for combining the problem content data, the policy implementation result data, and the effect verification result data stored in the questionnaire data storage means, or for combining the problem content data and the policy implementation result data to create combined text data; a perspective evaluation request means for generating perspective evaluation request data, one for each of the combined text data of each company, including the combined text data and a request for a binary response as to whether the content of the efforts expressed in the combined text data satisfies each of the multiple perspectives for evaluating each company's efforts regarding human capital management, and inputting the combined text data and perspective evaluation request data, one for each of the combined text data of each company, into a Chat Generative Pretrained Transformer (ChatGPT) or other large-scale language model, and storing the perspective evaluation result data output from the large-scale language model, which indicates whether each of the multiple perspectives is satisfied with one of the binary responses, in association with company response identification information, in a perspective evaluation result storage means; a score calculation means for calculating a term frequency-inverse document frequency (TF-IDF) by considering each of the multiple perspectives as a word and the combined text data of each company as a document using two values ​​constituting the perspective evaluation result data for the combined text data of each company stored in the perspective evaluation result storage means, and for adding up or averaging the TF-IDF values ​​for each of the multiple perspectives obtained for each combined text data to obtain a total or average value, or a value obtained by normalizing or otherwise processing these totals or average values, and storing the resulting value in the score storage means in association with the company response identification information as a health management navi score or other human capital management navi score; an output means for executing at least one of the following processes for displaying information related to human capital management on a screen: a process for displaying on a screen at least one of the company name, task content data, policy implementation result data, effectiveness verification result data, and combined text data corresponding to the company response identification information, together with the human capital management navigation score stored in the score storage means; a process for displaying on a screen at least one of the company name, task content data, policy implementation result data, effectiveness verification result data, and combined text data corresponding to the company response identification information according to the results of a selection, extraction, sorting, or grouping process using the human capital management navigation score; and a process for displaying on a screen the distribution of the human capital management navigation scores of each company or the changes over time in the human capital management navigation scores of each company; The present invention is characterized by the following features.

[0017] Here, "health and productivity management or other human capital management" indicates that "health and productivity management" is a term that includes the concept of "human capital management." Similarly, "health and productivity management questionnaire data or other human capital management-related questionnaire data" indicates that "health and productivity management questionnaire data" is a term that includes the concept of "human capital management-related questionnaire data." Furthermore, "health and productivity management navi score or other human capital management navi score" indicates that "health and productivity management navi score" is a term that includes the concept of "human capital management navi score." Therefore, the "questionnaire data" handled by the human capital management scoring system of the present invention includes data related to health and productivity management only, a mixture of data related to health and productivity management and data related to financial status, and data related to financial status only.

[0018] Furthermore, "health and productivity management questionnaire data or other human capital management questionnaire data" refers to questionnaire data in which response data on a set theme is organized. The survey body is not limited to national or local government agencies such as the Ministry of Economy, Trade and Industry, but can also be a private research organization. In other words, it is sufficient if the data contains data in which each company responded according to a set theme. Therefore, even if the name of a questionnaire currently referred to by the Ministry of Economy, Trade and Industry as the "Health and Productivity Management Questionnaire" is later changed, the data from the renamed questionnaire is acceptable as long as it contains response data on similar themes. Furthermore, questionnaire data in accordance with the guidelines of ISO 30414 (guidelines for disclosure of information related to human capital) is also acceptable.

[0019] Furthermore, the phrase "company response identification information for identifying a company, or for identifying a company and response data when multiple responses from one company are allowed" indicates that in the questionnaire data, when one company can only enter (input) one response data (one issue content data, one corresponding policy implementation result data, and one effectiveness verification result data), the "company response identification information" in this application is information for identifying a company. However, since there is a one-to-one correspondence between companies and response data, the "company response identification information" in this case is also information for identifying the response data. On the other hand, when the questionnaire data allows multiple responses from one company (for example, in the Health Management Survey conducted by the Ministry of Economy, Trade and Industry, one company can write (input) two responses), the "company response identification information" of this application indicates information that can identify the company and also identify the response data, but in this case, the "company response identification information" may be identification information that is a combination of company identification information and response identification information managed separately, or may be an integrated version of these identification information, as specifically described in detail in <Company response identification information: Figure 3> in the [Form for implementing the invention] described below.

[0020] Furthermore, "normalization processing or other processing" in "score calculation means" means that normalization processing is included in the concept of processing, and "other processing" refers to, for example, standardization processing, deviation value calculation processing, etc., and is specifically described in detail in (Calculation of the Health Management Navigation Score Value) in <Configuration of Processing Means 30 / Score Calculation Means 34: Figure 6> in [Form for Implementing the Invention] described below.

[0021] Furthermore, as "processing for displaying information related to human capital management on the screen" by the "output means," screen display processing using the human capital management navi score is listed, and it is stated that at least one of these processes will be performed, but the intention is that the function for performing screen display processing using the human capital management navi score should be included in the output means, and it is not intended to exclude the function for performing screen display processing that does not use the human capital management navi score (for example, screen display processing using the overall evaluation value or the overall deviation score).

[0022] In the human capital management scoring system of the present invention, the perspective evaluation request means obtains perspective evaluation result data from a large-scale language model (LLM) that indicates whether each of multiple perspectives is met using one of two values, and the score calculation means uses the perspective evaluation result data to calculate a human capital management navigation score using TF-IDF, and the output means performs processing to display information about human capital management on a screen using the human capital management navigation score.

[0023] This allows a human capital management navigation score to be obtained by quantifying the evaluation of the free text data (task content data, measure implementation results data, and effectiveness verification results) freely written by each company, and information related to human capital management can be displayed on the screen using this human capital management navigation score. A correlation has been confirmed between the obtained human capital management navigation score and the overall evaluation value calculated by the evaluator of the questionnaire data, confirming the usefulness of the human capital management navigation score. In the case of the embodiment described below (when nine perspectives are set for the Ministry of Economy, Trade and Industry's health management questionnaire and normalization processing is performed by dividing by the maximum value), the correlation coefficient is approximately 0.39. Furthermore, the correlation has been confirmed visually using graphs such as those shown in Figure 21 described below. However, the graph shown in Figure 21 of this application described below is merely an image of the screen display for functional explanation purposes and is not a strict tracing of a graph output on an actual device using actual accurate numerical values.

[0024] Furthermore, even if the evaluation result score is simply calculated as the sum or average value for each combined text data using the perspective evaluation result data (a binary value indicating whether each of the multiple perspectives is met or not) (a value indicating how many of the multiple perspectives are met, or a value proportional to that), a correlation with the overall evaluation value can be seen, but the evaluation result score is a numerical value that does not take into account the difficulty of achieving each of the multiple perspectives.In contrast, in this invention, the human capital management navigation score is calculated using the TF-IDF value, so it is a numerical value that takes the difficulty into account, making the evaluation result score even more useful.

[0025] By using the human capital management navigation score obtained in this way, it becomes possible to recommend measures that are deemed effective to users who wish to improve their evaluation of human capital management, thereby achieving the above-mentioned objective.

[0026] <Configuration that uses large-scale language models (LLMs) to extract multiple perspectives for evaluation>

[0027] In addition, in the human capital management scoring system mentioned above, a viewpoint extraction means for preparing a large number of combinations of two combined text data selected from the combined text data of each company created by the combining means, and for each of these many combinations, creating a set of two combined text data and viewpoint extraction request data including a request for an answer as to which of the two combined text data is superior in terms of the initiative content expressed in the combined text data and why that conclusion has been reached, and inputting this into a large-scale language model, and storing the text data output from the large-scale language model indicating why that conclusion has been reached in a viewpoint extraction result storage means as viewpoint data indicating from what perspective the comparison was made and whether the combinations were superior or inferior; The viewpoint evaluation request means is: It is desirable that the configuration be such that multiple perspectives set using perspective data stored in a perspective extraction result storage means are used as multiple perspectives for evaluating each company's efforts regarding human capital management.

[0028] In this manner, when a large-scale language model (LLM) is used to extract multiple perspectives for evaluation, the perspective extraction means performs a so-called pairwise evaluation to have the large-scale language model (LLM) extract perspectives, and then the perspective evaluation request means has the large-scale language model (LLM) evaluate the content of the initiative expressed in the combined text data from multiple perspectives set using the extracted perspectives, making it possible to evaluate the combined text data in accordance with the thinking characteristics (decision-making processing algorithm) of the large-scale language model (LLM).

[0029] <Configuration that generates a recommendation candidate set from the top k policy implementation result clusters, and selects the policy implementation result data with the highest human capital management navigation score from each policy implementation result cluster as the policy to be recommended>

[0030] Furthermore, in the human capital management scoring system mentioned above, a company response vectorization means for vectorizing the problem content data and the policy implementation result data of each company stored in the questionnaire data storage means, and storing the obtained problem content vectors and policy implementation result vectors in the company response vector storage means in association with the company response identification information; a policy implementation result clustering means for performing clustering using the policy implementation result vectors of each company stored in the company response vector storage means, assigning a policy implementation result cluster number for identifying the classified policy implementation result cluster to each company's policy implementation result vector, and storing the assigned policy implementation result cluster number in the company response vector storage means in association with the company response identification information; a recommendation candidate set generation means for calculating a similarity between a problem content vector of a company specified by a user who receives a policy recommendation service and each of problem content vectors of a plurality of other companies, sorting the company response identification information of the other plurality of companies in descending order of the calculated problem similarity, and selecting company response identification information in descending order of the problem similarity until the number of policy implementation result clusters of the policy implementation result cluster numbers corresponding to the sorted company response identification information becomes equal to the number k of required policies specified by the user or the predetermined number k of required policies, thereby generating a recommendation candidate set consisting of policy implementation result data belonging to the top k policy implementation result clusters; and a recommendation candidate selection means for selecting, as a recommendation candidate, the policy implementation result data with the highest human capital management navigation score from among the top k policy implementation result clusters selected by the recommendation candidate set generation means; The output means is It is desirable that the system be configured to execute a process of displaying the policy implementation result data selected by the policy selection means for recommendation as a policy to be recommended on a screen.

[0031] In this way, by generating a set of recommendation candidates from the top k policy implementation result clusters and selecting the policy implementation result data with the highest human capital management navigation score within each policy implementation result cluster as the policy to be recommended, it becomes possible to recommend a variety of policy implementation result data.

[0032] That is, the policy implementation result clustering means performs clustering using the policy implementation result vector to create policy implementation result clusters, and then the recommendation candidate set generation means first sorts the company response identification information by the similarity of the issues using the issue content vector (thus sorting the policy implementation result data, policy implementation result vector, and policy implementation result cluster number corresponding to the company response identification information).Next, the top k policy implementation result cluster numbers (thus sorting the top k policy implementation result clusters) are selected to generate a recommendation candidate set, and further, the recommended policy selection means selects the policy implementation result data with the highest human capital management navigation score from each policy implementation result cluster as the policy to be recommended, and the output means displays this on the screen, making it possible to recommend a variety of policy implementation result data.

[0033] For this reason, simply sorting the company response identification information by the similarity of the issues and outputting the policy implementation result data corresponding to the top company response identification information may result in a large amount of similar policy implementation result data being output, but by selecting the top k policy implementation result clusters and selecting and outputting the policy implementation result data with the highest human capital management navigation score within each policy implementation result cluster, it is possible to prevent a situation in which a large amount of similar policy implementation result data (i.e., policy implementation result data belonging to the same policy implementation result cluster) is output, and to output a diverse range of policy implementation result data.

[0034] <Press release screen display configuration>

[0035] In addition, when a recommendation candidate set is generated from the top k policy implementation result clusters described above, and the policy implementation result data with the highest human capital management navigation score in each policy implementation result cluster is selected as the policy to be recommended, a press release acquisition means for acquiring data of a plurality of press releases from a press release distribution system using keywords stored in the keyword storage means and storing the data in the press release storage means; a title vectorization means for vectorizing the title data of each of the press releases acquired by the press release acquisition means, and storing the obtained title vectors in a press release storage means in association with the acquired press releases; a press release selection means for calculating a similarity between each title vector stored in the press release storage means and a policy implementation result vector stored in the company response vector storage means for the policy implementation result data selected by the recommendation target policy selection means, and selecting a predetermined number of press releases or a number of press releases designated by a user that have a high similarity between the calculated title and policy; The output means is It is desirable that the system be configured to display on the screen the implementation result data of the measures that are the target of recommendation, and to execute a process to display on the screen the data of a predetermined number of press releases selected by the press release selection means or a number of press releases specified by the user.

[0036] When press releases are displayed on the screen in this manner, it becomes possible to present more information about human capital management to the user.

[0037] <Configuration with a measure name extraction means and a measure tag creation means>

[0038] Furthermore, in the case where a recommendation candidate set is generated from the top k policy implementation result clusters described above, and the policy implementation result data with the highest human capital management navigation score in each policy implementation result cluster is selected as the policy to be recommended, a policy name extraction means for creating policy extraction request data for each company, the policy implementation result data being stored in the questionnaire data storage means, and the policy implementation result data including a request to extract health management or other human capital management policies from the text described in the policy implementation result data, and inputting the created policy extraction request data into a large-scale language model, and storing the text data indicating the extracted policy or policies output from the large-scale language model as policy name data in association with company response identification information in a policy extraction result storage means; and a policy tag creation means for vectorizing each of the policy name data extracted by the policy name extraction means, and performing clustering using the obtained policy name vectors, and for each of the plurality of policy name clusters formed by classification, creating a set of policy name data belonging to the same policy name cluster and policy tag naming request data including a request to name this set, and inputting the created set into a large-scale language model, and assigning the text data indicating the name of the same policy name cluster output from the large-scale language model as a policy tag to the policy name data belonging to the same policy name cluster, and storing the assigned policy tag in a policy extraction result storage means in association with company response identification information. The output means is It is desirable that the system be configured to display on the screen the policy implementation result data for the policy to be recommended, and to execute a process to display on the screen the policy tag stored in the policy extraction result storage means in association with the company response identification information corresponding to the policy implementation result data displayed on the screen.

[0039] In this configuration equipped with a policy name extraction means and a policy tag creation means, since the number of policies (policy name data) extracted from the policy implementation result data using the large-scale language model (LLM) is too large, clustering is performed to create policy name clusters, and then names are assigned to each policy name cluster using the large-scale language model (LLM). These names can be displayed on the screen as policy tags together with the policy implementation result data that is the recommended policy. Therefore, by looking at the policy tags, which are fewer in total number than the policy name data, i.e., by looking at the policy tags, which are a more cohesive unit than the policy name data, it is possible to understand the content of the policy implementation result data that is the recommended policy. Furthermore, by displaying the policy tags on the screen, it is possible to use the policy tags to perform various processes such as searches.

[0040] <Configuration with a measure tagging and related measure search means>

[0041] In addition, in the case where the above-mentioned policy name extraction means and policy tag creation means are provided, a policy tag search means for searching related policies by policy tag, which uses a policy tag designated by a user to extract company response identification information associated with the designated policy tag and stored in a policy extraction result storage means and / or policy implementation result data associated with the company response identification information and stored in a questionnaire data storage means, and selects a predetermined number or a number designated by the user from the extracted company response identification information and / or policy implementation result data in descending order of the human capital management navigation score associated with the extracted company response identification information and stored in a score storage means; The output means is It is desirable that the screen be configured to display the company name and / or the policy implementation result data corresponding to the company response identification information selected by the policy tagging related policy search means.

[0042] In this way, when the system is configured to include a measure tagging related measure searching means, the related measure implementation result data can be searched for and displayed on the screen using the measure tag.

[0043] <Program invention>

[0044] The program of the present invention is for causing a computer to function as the above-described human capital management scoring system.

[0045] The above program or a portion thereof may be recorded on a recording medium such as a magneto-optical disk (MO), compact disk (CD), digital versatile disk (DVD), flexible disk (FD), magnetic tape, read-only memory (ROM), electrically erasable and programmable read-only memory (EEPROM), flash memory, random access memory (RAM), hard disk drive (HDD), solid-state drive (SSD), or flash disk, and may be transmitted using a transmission medium such as a wired network (e.g., local area network (LAN), metropolitan area network (MAN), wide area network (WAN), the Internet, an intranet, or an extranet), a wireless communication network, or a combination thereof, or may be carried on a carrier wave. Furthermore, the above program may be a portion of another program, or may be recorded on a recording medium together with a separate program. [Effects of the Invention]

[0046] As described above, according to the present invention, perspective evaluation result data is obtained from a large-scale language model (LLM), indicating whether or not each of multiple perspectives is met using one of two values.Furthermore, this perspective evaluation result data is used to calculate a human capital management navigation score using TF-IDF, and information related to human capital management can be displayed on the screen using this human capital management navigation score, thereby having the effect of recommending measures that are thought to be effective to users who wish to improve their evaluation of human capital management. [Brief explanation of the drawings]

[0047] [Figure 1] FIG. 1 is an overall configuration diagram of a health management scoring system according to one embodiment of the human capital management scoring system of the present invention. [Figure 2] FIG. 10 is a diagram illustrating an example of health management survey data according to the embodiment. [Figure 3] FIG. 10 is a diagram illustrating the relationship between various data derived from the response data of each company entered in the health management survey form according to the embodiment. [Figure 4] FIG. 10 is an explanatory diagram of a process for extracting viewpoints for evaluating each company's efforts regarding health management according to the embodiment. [Figure 5] FIG. 10 is an explanatory diagram of the process of evaluating each company's efforts regarding health management from multiple (nine) perspectives set in the embodiment. [Figure 6] FIG. 10 is an explanatory diagram of the calculation process of the health management navigation score (human capital management navigation score) using TF-IDF according to the embodiment. [Figure 7] 5A and 5B are explanatory diagrams of the vectorization process and the policy implementation result cluster creation process according to the embodiment. [Figure 8] 10A and 10B are explanatory diagrams of the process of extracting action names and creating action tags according to the embodiment. [Figure 9] FIG. 4 is an explanatory diagram of a process for generating a recommendation candidate set according to the embodiment. [Figure 10] 10 is an explanatory diagram of the process of selecting a measure to be recommended from a recommendation candidate set and a press release to be displayed on the screen in the embodiment. FIG. [Figure 11]FIG. 10 is a flowchart showing the flow of preparation processing before service provision by the health management scoring system of the embodiment. [Figure 12] FIG. 10 is a flowchart showing the processing flow when a service is provided to a user by the health management scoring system of the embodiment. [Figure 13] FIG. 3 is an explanatory diagram of the overall screen display process by the output unit of the embodiment. [Figure 14] FIG. 10 is a diagram illustrating an example of an upper portion of a company profile screen in the embodiment. [Figure 15] FIG. 10 is a diagram illustrating a lower portion of a company profile screen according to the embodiment. [Figure 16] 10 is a diagram illustrating an example of a condition setting section and an upper part of a "search by company name" tab on the measure recommendation screen according to the embodiment. FIG. [Figure 17] FIG. 10 is a diagram illustrating an example of a lower part of a "Search by Company Name" tab on the policy recommendation screen according to the embodiment. [Figure 18] FIG. 10 is a diagram illustrating an example of a “Search from input issue” tab on the measure recommendation screen according to the embodiment. [Figure 19] FIG. 10 is a view showing an example of a "Measure Tag List" tab on the measure recommendation screen according to the embodiment. [Figure 20] FIG. 10 is a view showing an example of a similar company search screen in the embodiment. [Figure 21] FIG. 10 is a diagram illustrating an example of a health management navigation score screen according to the embodiment. [Figure 22] FIG. 10 is a diagram showing an example of a screen displaying search results for related measures based on a measure tag according to the embodiment. [Figure 23] FIG. 10 is a view showing an example of an analysis result details display screen according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0048] An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 illustrates the overall configuration of a health and productivity management scoring system 10, which is an embodiment of the human capital management scoring system. FIG. 2 illustrates an example of data entry on a health and productivity management questionnaire, and FIG. 3 illustrates the relationships between various data derived from each company's response data entered on the health and productivity management questionnaire. Also, FIG. 4 is an explanatory diagram of the process of extracting perspectives for evaluating each company's health and productivity management initiatives. FIG. 5 is an explanatory diagram of the process of evaluating each company's health and productivity management initiatives from multiple (nine) perspectives. FIG. 6 is an explanatory diagram of the process of calculating a health and productivity management navi score (human capital management navi score) using TF-IDF. FIG. 7 is an explanatory diagram of the process of vectorization and the process of creating a policy implementation result cluster. FIG. 8 is an explanatory diagram of the process of extracting policy names and creating policy tags. FIG. 9 is an explanatory diagram of the process of generating a recommendation candidate set. FIG. 10 is an explanatory diagram of the process of selecting a policy to be recommended from the recommendation candidate set and the process of selecting a press release to be displayed on the screen. 11 shows a flowchart of the preparation process before the health and productivity management scoring system provides a service, and FIG. 12 shows a flowchart of the process when the health and productivity management scoring system provides a service to a user. Also, FIG. 13 is an overall explanatory diagram of the screen display process by output means 45, FIG. 14 shows an example of the upper part of company profile screen 200, FIG. 15 shows an example of the lower part of company profile screen 200, FIG. 16 shows an example of the condition setting section 310 and the upper part of "Search by company name" tab 301 of measure recommendation screen 300, FIG. 17 shows an example of the lower part of "Search by company name" tab 301 of measure recommendation screen 300, and FIG. 20 shows an example of a similar company search screen 400, FIG. 21 shows an example of a health management navigation score screen 500, FIG. 22 shows an example of a policy tag search related policy search result display screen 600, and FIG. 23 shows an example of an analysis result detailed display screen 700.

[0049] <Overall structure of the Health Management Scoring System 10: Figure 1>

[0050] 1, health management scoring system 10 includes scoring server 20 that executes various processes to present information about health management to users and stores various data required to execute the various processes, and a large-scale language model (LLM)-based service provision system 70, a vectorization processing system 71, and a press release provision system 80 that are connected to scoring server 20 via network 1. Also connected to scoring server 20 via network 1 are an administrator terminal 90 operated by a system administrator and a user terminal 91 operated by a person in charge at a user company.

[0051] Here, network 1 is an external network mainly consisting of the Internet, but it can also be a combination of the Internet and an internal network such as a LAN or an intranet, and it does not matter whether it is wired or wireless, or even a combination of wired and wireless; in short, it is sufficient if it can transmit information at a certain speed between multiple points (regardless of the distance).

[0052] The scoring server 20 is composed of one or more computers and is equipped with a processing means 30 that executes various processes to present various information related to health management to users, and a storage means 50 that stores various data necessary for executing the various processes.

[0053] The processing means 30 is composed of a combining means 31, a viewpoint extraction means 32, a viewpoint evaluation request means 33, a score calculation means 34, a company response vectorization means 35, a policy implementation result clustering means 36, a recommendation candidate set generation means 37, a recommendation target policy selection means 38, a press release acquisition means 39, a title vectorization means 40, a press release selection means 41, a policy name extraction means 42, a policy tag creation means 43, a policy tag lookup related policy search means 44, and an output means 45.

[0054] Here, each of the means 31 to 45 constituting the processing means 30 is realized by a central processing unit (CPU) provided inside the scoring server 20, one or more programs that define the operating procedures of this CPU, and working memory such as main memory and cache memory. The details of each of these means 31 to 45 will be described later.

[0055] The storage means 50 is configured to include a survey form data storage means 51, a company information storage means 52, a viewpoint extraction result storage means 53, a set viewpoint storage means 54, a viewpoint evaluation result storage means 55, a score storage means 56, a company response vector storage means 57, a keyword storage means 58, a press release storage means 59, a policy extraction result storage means 60, and a policy name / policy tag correspondence storage means 61.

[0056] Here, for example, nonvolatile memories such as hard disk drives (HDDs) and solid state drives (SSDs) can be used as hardware for each of the storage means 51 to 61 constituting the storage means 50. Details of the data retention format of each of these storage means 51 to 61 (such as the table configuration of a database and the data recording format in a file) will be described later.

[0057] The large-scale language model (LLM) service provision system 70 is a computer system that provides cloud API (Application Programming Interface) services. In this embodiment, ChatGPT (Chat Generative Pre-trained Transformer, Azure OpenAI ChatCompletion API GPT-4) is used, but the present invention is not limited to this. For example, OpenAI's GPT-3.5, Google's Palm and Palm2, Amazon Web Services' (AWS) Titan, Meta Platforms' (Meta) Llama, etc. may also be used.

[0058] The vectorization processing system 71 is an external service provision system using a cloud API connected via the network 1, and is a computer system that performs vectorization processing of text data. In this embodiment, as an example, the Azure OpenAI Embeddings API service provision system (Azure OpenAI Embeddings API text-embedding-ada-002 version 2) can be used, but is not limited to this. The embedding method may also be Doc2Vec, BERT, Transformer, etc. The vector has a fixed length, but the number of dimensions is arbitrary; in this embodiment, for example, it is 1536 dimensions.

[0059] The press release providing system 80 is an external service providing system using a cloud API connected via network 1, and is a computer system that performs a process to output press release data (including title data) related to a keyword when that keyword is input.

[0060] The administrator terminal 90 and the user terminal 91 are configured by computers and equipped with display means such as an LCD display and input means such as a mouse and keyboard. The administrator terminal 90 and the user terminal 91 may be portable devices such as laptops, tablets, smartphones, etc.

[0061] <Health and Productivity Management Survey Data Structure: Figure 2>

[0062] In Figure 2, the health and productivity management survey data released by the survey body (Ministry of Economy, Trade and Industry) contains the response data of each company to the health and productivity management survey. This response data includes the theme number indicating the theme selected by each company for its response from 10 theme themes prepared by the survey body (in the example of Figure 2, theme number 4 is selected), as well as the task content data, which is free text written (input) by each company, the implementation result data of the measures, and the result data of the effectiveness verification. This response data of each company is stored in the survey data storage means 51 (see Figure 3) together with the overall evaluation value calculated by the survey body for the response data and data indicating the evaluation results such as various deviation scores.

[0063] <Company response identification information: Figure 3>

[0064] As shown in the survey data storage means 51 in Fig. 3, the response data of each company and data indicating the evaluation results for that response data are stored in association with company response identification information. As already described in [Means for Solving the Problems], the "company response identification information" in the present application is defined to include cases where one company can only enter (input) one piece of response data in the survey data, but the health and productivity management survey data of this embodiment shown in Fig. 2 allows one company to enter (input) multiple pieces of response data (two), so the "company response identification information" in this embodiment is information that can identify the company and also identify the response data.

[0065] In this embodiment, as shown in FIG. 3, for ease of explanation, the "company response identification information" is defined as an integrated combination of company identification information and response identification information. That is, the first half of the company response identification information functions as company identification information, and the second half functions as identification information for response data within the company. For example, in "00052300" in FIG. 3, the first half "000523" indicates the company, and the second half (last two digits) "00" indicates the first response data within the company "000523." The second response data within the company "000523" is, for example, "00052301" by adding "01" to "000523." Sub-numbers such as "000523-00" and "000523-01" may also be used. Furthermore, the company name may be used as the company identification information, and the company response identification information for the first and second response data of ABC Co., Ltd. may be, for example, "ABC Co., Ltd. (1)" and "ABC Co., Ltd. (2)."

[0066] Furthermore, a combination of company identification information and separately managed response identification information may be used as company response identification information. For example, company identification information such as "000523" and internal company response data identification information such as "00" or "01" may be stored in different columns of the same record, and the combination of these may function as company response identification information. Furthermore, the response data identification information may not be internal company response data identification information, but may be identification information that manages all company response data by serial number. For example, company identification information such as "000523" or "ABC Corporation" may be combined with internal company response identification information such as "002341" or "002342."

[0067] <Configuration of processing means 30 / combining means 31: Figure 3>

[0068] The combining means 31 executes a process of creating combined text data by combining the issue content data, the policy implementation result data, and the effectiveness verification result data stored in the survey table data storage means 51 (see FIG. 3). Note that the issue content data and the policy implementation result data may be combined excluding the effectiveness verification result data, but it is preferable to combine the issue content data and the policy implementation result data as well, since this may have an impact such as reducing the number of evaluation perspectives (9 in this embodiment), which will be described later.

[0069] At this time, the combining means 31 performs the combining process by removing line feed codes. The obtained combined text data may be stored in the main memory, but may also be saved in a non-volatile memory (a combined text data storage means not shown).

[0070] <Configuration of processing means 30 / perspective extraction means 32: Fig. 4>

[0071] The viewpoint extraction means 32 prepares a large number of combinations of two combined text data selected from the combined text data of each company created by the combining means 31, creates viewpoint extraction request data (including the two combined text data and a request to answer which of these two combined text data is superior in terms of the initiative content expressed in that combined text data, and why that conclusion has been reached) for each of these many combinations, inputs the created viewpoint extraction request data into a large scale language model (LLM), and executes a process of storing the text data output from the large scale language model (LLM) indicating why that conclusion was reached in the viewpoint extraction result storage means 53 as viewpoint data indicating from what viewpoint the comparison was made and superiority / inferiority was determined.

[0072] Specifically, the viewpoint extraction means 32 selects two combined text data and creates viewpoint extraction request data, as shown in Fig. 4. The example in Fig. 4 shows a state in which viewpoint extraction request data has been created using combined text data created by the combining means 31 for company response identification information = 00026400 and combined text data created by the combining means 31 for company response identification information = 00073400. Then, the viewpoint extraction means 32 transmits the created viewpoint extraction request data to the service providing system 70 using a large-scale language model (LLM) via the network 1.

[0073] When selecting two combined text data to be combined, the viewpoint extraction means 32 may select them from a large number of combined text data already created by the combining means 31, but instead of this order, the viewpoint extraction means 32 may select two pieces of company response identification information to be combined, and then the combining means 31 may create two combined text data corresponding to those two pieces of company response identification information.In this order, there is no need to store a large number of combined text data, and it is only necessary to store the pairs of company response identification information used for the combination.

[0074] Next, the viewpoint extraction means 32 receives LLM response data transmitted from the large scale language model (LLM)-based service provision system 70 via the network 1, and stores text data included in the LLM response data, which indicates why the conclusion was reached, as viewpoint data in association with the pair of company response identification information used in the combination, in the viewpoint extraction result storage means 53. The example in Fig. 4 shows a state in which the pair of company response identification information 00026400 and 00073400 and the viewpoint data "Because quantitative evaluation is being performed" are associated and stored in the viewpoint extraction result storage means 53.

[0075] The viewpoint extraction means 32 then repeats the above process for a large number of pairs of company response identification information (combinations of two combined text data). In this embodiment, there are health management questionnaires for more than 2,000 companies, and each company can enter two pieces of response data. Therefore, there are a total of approximately 4,200 pieces of response data (company response identification information). Since there are 10 types of problem themes, 40 pairs were randomly selected for each theme and the viewpoint extraction process was performed. Therefore, the total number of pairs used was 400 pairs (40 pairs x 10 themes). In other words, viewpoint extraction request data was created for each of the 400 pairs, and a response from the large-scale language model (LLM) was obtained for each of the 400 pairs.

[0076] Furthermore, the viewpoint extraction means 32 uses a large number (400 in this embodiment) of viewpoint data stored in the viewpoint extraction result storage means 53 to set multiple viewpoints (9 in this embodiment) to be used in the processing of the viewpoint evaluation request means 33 described below, and performs a process of storing the set multiple (9) viewpoint data in the set viewpoint storage means 54 in association with viewpoint numbers (1 to 9 in this embodiment), or a process of supporting the work for this setting.

[0077] In this embodiment, there are 400 extracted (collected) viewpoint data (viewpoint data stored in the viewpoint extraction result storage means 53), but naturally there are overlaps, so a setting operation was performed to eliminate overlaps, meaning that the set viewpoint data is now 9. This number of 400 is a number that allows a person to check the content with concentration; in other words, the total number of pairs was set to 400 so that such a number would be achieved. Therefore, in this embodiment, the viewpoint extraction means 32 outputs (screen display or print) all (400) viewpoint data stored in the viewpoint extraction result storage means 53 to the administrator terminal 90. Then, the system administrator (system developer) visually checks the output result and determines that there are multiple (9) pieces of viewpoint data, and also determines that even if the total number of pairs is increased, no more (10th and subsequent) pieces of viewpoint data will appear, and performs input work to set multiple (9) pieces of viewpoint data. The viewpoint extraction means 32 accepts the setting input and performs processing to associate the set multiple (9) pieces of viewpoint data with viewpoint numbers (1 to 9 in this embodiment) and store them in the set viewpoint storage means 54. At this time, if the system administrator (developer) determines that there are approximately multiple (9) pieces of viewpoint data among the large number (400) of extracted viewpoint data, he or she performs work to combine viewpoint data with similar expressions into one expression, and then performs setting input to the set viewpoint storage means 54.

[0078] Furthermore, when the number of viewpoint data extracted and stored in the viewpoint extraction result storage means 53 is very large, the viewpoint extraction means 32 may vectorize each of the viewpoint data, and create multiple (for example, nine) viewpoint clusters by clustering using the obtained viewpoint vectors, and create viewpoint expression aggregation request data (including multiple viewpoint data belonging to the same viewpoint cluster and a request statement requesting that the expressions of these multiple viewpoint data be combined into one expression) for each of the multiple (nine) viewpoint clusters created, and input this into a large-scale language model (LLM), and store the LLM answer data output from the large-scale language model (LLM) as set viewpoint data (viewpoint data with expressions aggregated) in the set viewpoint storage means 54 in association with a viewpoint number (in this embodiment, 1 to 9). This process is similar to the process performed by the policy tag creation means 43 described later (see Figure 8), i.e., vectorizing policy name data to create a policy name vector, creating multiple policy name clusters by clustering using the policy name vector, requesting a large-scale language model (LLM) to name the policy name cluster using a set of policy name data belonging to the policy name cluster, and using the LLM response data as the policy tag.

[0079] It is also possible to omit the viewpoint extraction process by the viewpoint extraction means 32 and store viewpoint data created by the system administrator (developer) as set viewpoint data in the set viewpoint storage means 54 in association with a viewpoint number (1 to 9 in this embodiment). In other words, if a plurality of viewpoint data are stored in the set viewpoint storage means 54 before the processing of the viewpoint evaluation request means 33 described later is performed, the processing of the viewpoint evaluation request means 33 described later can be performed, and therefore the viewpoint data may be viewpoint data created by human thought. However, it is preferable to store viewpoint data (expression-aggregated viewpoint data) set using viewpoint data obtained in the viewpoint extraction process by the viewpoint extraction means 32 in the set viewpoint storage means 54, because this allows viewpoint data to be set in accordance with the thinking algorithm of a large-scale language model (LLM) and improves the reliability of the processing of the viewpoint evaluation request means 33 described later.

[0080] <Configuration of processing means 30 / perspective evaluation request means 33: FIG. 5>

[0081] The perspective evaluation request means 33 creates perspective evaluation request data (including the combined text data and a request statement requesting a binary response as to whether or not the content of the efforts expressed in the combined text data of each company satisfies each of the multiple perspectives) for each of the combined text data of each company in order to determine whether or not the content expressed in the combined text data of each company created by the combining means 31 satisfies each of the multiple perspectives for evaluating each company's efforts regarding human capital management, inputs the created perspective evaluation request data into a large-scale language model (LLM), and executes a process of storing the perspective evaluation result data output from the large-scale language model (LLM) in a perspective evaluation result storage means 55 in association with the company response identification information.

[0082] Specifically, in this embodiment, as shown in Fig. 5, the perspective evaluation request means 33 creates perspective evaluation request data (including a request statement requesting a binary response as to whether the combined text data and the initiatives expressed in the combined text data satisfy any one of the multiple (nine) perspectives) for each of the combined text data of each company created by the combining means 31 and for each of the multiple (nine) perspectives, and transmits the created perspective evaluation request data to the large-scale language model (LLM)-based service provision system 70 via the network 1. The example of Fig. 5 shows a state in which the perspective evaluation request data was created using the combined text data for company response identification information = 00115200 (Company C) = "There are many employees on leave due to mental illness. Mental health training was conducted. The number of people suffering from depression has decreased" and the perspective data for perspective number = 2, "Is quantitative evaluation being performed?"

[0083] Next, the perspective evaluation request means 33 receives LLM response data (perspective evaluation result data indicating whether or not a perspective is satisfied using either of two values, 1 (satisfied) or 0 (not satisfied)) transmitted via the network 1 from the large-scale language model (LLM)-based service provision system 70, and performs processing to associate the perspective evaluation result data with the company response identification information and the perspective number and store it in the perspective evaluation result storage means 55. The example in Fig. 5 shows a state in which perspective evaluation result data of "1", which indicates that the perspective of perspective number = 2 is satisfied, is stored in the perspective evaluation result storage means 55 in association with company response identification information = 00115200 (Company C) and perspective number = 2.

[0084] The viewpoint evaluation request means 33 then repeatedly executes this process for all combined text data (all company response identification information) and all viewpoint numbers. Therefore, if there are, for example, 4200 items of combined text data and nine viewpoints (viewpoint numbers = 1 to 9), 4200 items x 9 viewpoints = 37800 viewpoint evaluation request data are created, and 37800 viewpoint evaluation result data (either 1 or 0) are obtained.

[0085] At this time, the viewpoint evaluation request means 33 creates viewpoint evaluation request data using a plurality (nine) of viewpoint data stored in the set viewpoint storage means 54, as shown in Fig. 5. Therefore, if the set viewpoint storage means 54 stores a plurality (nine) of viewpoint data (expression-aggregated viewpoint data) set using viewpoint data extracted by the viewpoint extraction means 32 described above, the viewpoint evaluation request data will be created using viewpoint data that reflects the processing by the viewpoint extraction means 32. Note that in the automatic creation process of viewpoint evaluation request data, the viewpoint evaluation request data may be created by changing the endings of the viewpoint data (for example, by deleting "ka.", "whether or not.", "ka?", etc.), or the viewpoint evaluation request data may be created using the viewpoint data as is.

[0086] In addition, in this embodiment, the perspective evaluation request means 33 creates one perspective evaluation request data for each of the combined text data of each company and one for each of the multiple (nine) perspectives, but perspective evaluation request data may be created for several perspectives together rather than one by one. For example, the respondent may be asked whether or not each of the multiple (nine) perspectives is satisfied, or the multiple (nine) perspectives may be divided into several (for example, into three perspective numbers = 1 to 3, 4 to 6, 7 to 9) and asked to answer whether or not each perspective is satisfied. However, in order to prevent the perspective evaluation request data from becoming too long or to prevent misunderstandings of the request text due to a large-scale language model (LLM), it is preferable to create perspective evaluation request data for each perspective. Therefore, although the claims of the present application state that "a request for a binary response as to whether or not each of the multiple viewpoints is met," this does not only refer to the case where a response is made for all of the multiple viewpoints (9 viewpoints) at once, but also includes the case where a response is made for each of the multiple viewpoints (9 viewpoints) one by one, as in this embodiment, or the case where a response is made for each of the multiple viewpoints (9 viewpoints) divided into several viewpoints.

[0087] Furthermore, in this embodiment, the "binary value" is 1 (satisfied) or 0 (not satisfied), but is not limited to this and may be, for example, 10 (satisfied) or 0 (not satisfied), or 100 (satisfied) or 0 (not satisfied), which are equivalent because the value is adjusted (normalized, etc.) by the processing of the score calculation means 34 described below.

[0088] Furthermore, when perspective evaluation request data is created by the perspective evaluation request means 33, if the combined text data used in the processing by the perspective extraction means 32 described above remains in the main memory or is stored in non-volatile memory, the perspective evaluation request data may be created using that combined text data, but each time perspective evaluation request data is created by the perspective evaluation request means 33, the combined text data may also be created by the combining means 31 using the response data (issue content data, policy implementation result data, and effect verification result data) stored in the questionnaire data storage means 51.

[0089] <Configuration of processing means 30 / score calculation means 34: FIG. 6>

[0090] The score calculation means 34 uses the binary values ​​(in this embodiment, 1 or 0) that constitute the perspective evaluation result data for the combined text data of each company stored in the perspective evaluation result storage means 55 to calculate TF-IDF (Term Frequency-Inverse Document Frequency) by treating each of the multiple (nine) perspectives as a word and treating the combined text data of each company as a document, and then sums (or averages) the TF-IDF values ​​for each of the multiple (nine) perspectives obtained for each combined text data to obtain a total value (or average value), and then performs a Yeo-Johnson transformation on these total values ​​(or average value), and then normalizes these values ​​using the Yeo-Johnson transformed value of the total values ​​(or average value) for all combined text data, and performs a process of storing the normalized value as a health management navigation score (human capital management navigation score) in the score storage means 56 in association with the company response identification information.

[0091] TF-IDF is generally calculated by multiplying TF (Term Frequency), which indicates the frequency of occurrence of a word in a document, by the value of IDF (Inverse Document Frequency), which indicates the rarity of the word. Therefore, as described above, since TF-IDF is calculated by regarding each of the multiple (nine in this embodiment) perspectives as a word and the combined text data of each company as a document, the TF-IDF value in this application is a numerical value that takes into account the difficulty of achieving each of the multiple perspectives.

[0092] Specifically, the score calculation means 34 performs TF-IDF calculation processing as shown in Fig. 6. However, the viewpoint evaluation result storage means 55 shown at the top of Fig. 6 is the same as the viewpoint evaluation result storage means 55 shown at the bottom of Fig. 5. However, in Fig. 6, the number of viewpoints and the number of combined text data (number of company response identification information) used in the calculation are reduced in order to explain the contents of the TF-IDF calculation processing by performing specific calculations using specific numerical examples. That is, in Fig. 6, the number of viewpoints is actually nine in this embodiment, but is reduced to three, and the number of combined text data actually exists about 4,200, but the calculation will be explained assuming that only three items exist.

[0093] (Calculation of TF value) For TF, in this embodiment, the value (value of 1 or 0) of the viewpoint evaluation result data shown in the matrix stored in the viewpoint evaluation result storage means 55 is used as it is. Therefore, since the value of TF is the value of the matrix stored in the viewpoint evaluation result storage means 55, ultimately, speaking only in the case of this embodiment, the value of TF-IDF is obtained by multiplying the value (value of 1 or 0) of the viewpoint evaluation result data shown in the matrix by the value of IDF, which will be described later.

[0094] Furthermore, the TF used to calculate the TF-IDF in the present invention (the TF-IDF described in the claims) is not limited to the case where the value of the viewpoint evaluation result data (a value of 1 or 0) is used as the TF value as is, as in this embodiment.

[0095] For example, a standard TF is the value obtained by dividing the number of occurrences of word t in document d by the total number of words in sentence d (the total number of words when counting multiple occurrences of the same word separately). In this case, the denominator is, in other words, the total number of occurrences of each word in sentence d, and is usually expressed using Σ. When this standard TF is adopted, the value obtained is the number of occurrences of perspective t in the combined text data d (obviously 1 or 0) divided by the total number of occurrences of each perspective in the combined text data d (obviously 1 or 0). In this case, the denominator is ultimately the number of perspectives satisfied by the combined text data d. A specific calculation for the example in Figure 6 is as follows.

[0096] The standard TF value for the combined text data of company response identification information=00127100 (Company H) is [0 / 1, 1 / 1, 0 / 1]=[0, 1, 0]. The standard TF value for the combined text data of company response identification information=00127200 (Company J) is [1 / 3, 1 / 3, 1 / 3]=[0.33, 0.33, 0.33]. The standard TF value for the combined text data of company response identification information=00127300 (company K) is [1 / 2, 1 / 2, 0 / 2]=[0.5, 0.5, 0].

[0097] When calculating the standard TF value as described above, if the denominator becomes zero (when combined text data exists that does not satisfy all the viewpoints), it is possible to adjust the denominator by adding "+1", for example.

[0098] Alternatively, the TF may be expressed as a logarithmic scale frequency, and is calculated using the following formula: TF = log(1 + number of occurrences of word t in document d)

[0099] So, by substitution, we get: TF = log{1 + the number of occurrences of viewpoint t in the combined text data d (1 or 0)}

[0100] Furthermore, the TF may be calculated by other methods, for example, by the following calculation formula:

[0101] TF = 0.5 + 0.5 × (number of occurrences of word t in document d / number of occurrences of the most frequently occurring word in document d)

[0102] Therefore, after substitution, we get the following: In this case too, if the denominator becomes zero (if there is combined text data that does not satisfy all the points), it can be adjusted by adding "+1" to the denominator, for example.

[0103] TF = 0.5 + 0.5 × {number of occurrences of viewpoint t in combined text data d (1 or 0) / number of occurrences of viewpoint with the highest number of occurrences in combined text data d (1 or 0)}

[0104] (Calculating the IDF value) In this embodiment, the IDF is calculated using the scikit-learn library, so ``+1'' is added to the outside of the log term in the IDF calculation formula, and ``+1'' is also added to the denominator and numerator inside the log term, resulting in the following calculation formula.

[0105] IDF of term t = log{(1 + total number of documents) / (1 + number of documents containing term t)}+1

[0106] So, by substitution, we get: IDF of viewpoint t = log{(1 + total number of combined text data) / (1 + number of combined text data that satisfy viewpoint t)}+1

[0107] In the example of Figure 6, if we calculate the IDF value for each perspective (perspective number = 1, 2, 3), we get the following, as shown near the center of Figure 6.

[0108] IDF(perspective number = 1) = log{(1 + 3) / (1 + 2)} + 1 = 1.29 IDF(perspective number = 2) = log{(1 + 3) / (1 + 3)} + 1 = 1.00 IDF(perspective number = 3) = log{(1 + 3) / (1 + 1)} + 1 = 1.69

[0109] Regarding IDF, in this embodiment, the IDF value is calculated using the default calculation formula of the scikit-learn library described above, but it is not limited to the calculation formula described above. For example, it is possible to omit adjusting the "+1" in the denominator or numerator inside the log term, or adjusting the "+1" outside the log term.

[0110] (Calculating the TF-IDF value) The TF-IDF value in this embodiment is calculated by multiplying the value of each element (which is also the TF value) in the matrix stored in the viewpoint evaluation result storage means 55 shown at the top of Fig. 6 by the above-mentioned IDF value. That is, the TF-IDF value as shown in the matrix near the center of Fig. 6 is obtained by multiplying the value of each element in the column (first column) of viewpoint number = 1 stored in the viewpoint evaluation result storage means 55 by 1.29, multiplying the value of each element in the column (second column) of viewpoint number = 2 by 1.00, and multiplying the value of each element in the column (third column) of viewpoint number = 3 by 1.69.

[0111] (Calculation of Health and Productivity Management Navigation Score) Next, the score calculation means 34 calculates the TF-IDF value, and then sums (or may average) the TF-IDF values ​​for each of the multiple (nine) perspectives obtained for each combined text data to calculate a total value (or may average). This total value (or may average) becomes the health and productivity management navigation score. However, this is a value before processing such as normalization.

[0112] In the example of Figure 6, for the combined text data of company response identification information = 00127100 (Company H), the sum of the TF-IDF values ​​of viewpoint numbers = 1, 2, and 3 is 0.00 + 1.00 + 0.00 = 1.00, so this total value of 1.00 is stored in the score storage means 56 in association with company response identification information = 00127100 as the health management navigation score (however, the value before processing such as normalization).

[0113] Furthermore, for the combined text data of company response identification information = 00127200 (Company J), the sum of the TF-IDF values ​​for viewpoint numbers = 1, 2, and 3 is 1.29 + 1.00 + 1.69 = 3.98, and this total value of 3.98 is stored in the score storage means 56 in association with company response identification information = 00127200 as the health management navigation score (however, the state before processing such as normalization).

[0114] Furthermore, for the combined text data of company response identification information = 00127300 (Company K), the sum of the TF-IDF values ​​of viewpoint numbers = 1, 2, and 3 is 1.29 + 1.00 + 0.00 = 2.29, so this total value of 2.29 is stored in the score storage means 56 in association with company response identification information = 00127300 as the health management navigation score (however, the state before processing such as normalization).

[0115] The score calculation means 34 then performs a Yeo-Johnson transform (converts to a Gaussian distribution to make it easier to handle as a machine learning feature) on the sum (or average) of the TF-IDF values ​​for each combined text data obtained, and then normalizes the value after the Yeo-Johnson transform of the sum (or average) (hereinafter referred to as the health and productivity management navigation score before normalization) so that it falls within the range of 0 to 100. This normalization process involves extracting the maximum value from the health and productivity management navigation scores before normalization, dividing the health and productivity management navigation score for each combined text data by the maximum value, and multiplying the result by 100 to obtain a final normalized health and productivity management navigation score. The score calculation means 34 then stores the obtained final health and productivity management navigation score in the score storage means 56 in association with the company response identification information.

[0116] In this embodiment, normalization is performed by dividing by the maximum value (normalization that results in a proportional relationship between the pre-processing value and the post-processing value). However, this is not limiting. For example, the maximum and minimum values ​​may be extracted from the pre-normalization health management navigation score, and normalization may be performed such that the maximum value is 100 and the minimum value is 0 (normalization that results in a non-proportional relationship between the pre-processing value and the post-processing value). In this case, the final normalized health management navigation score is calculated by subtracting the minimum value from the pre-normalization health management navigation score, dividing the result by the result of subtracting the minimum value from the maximum value, and multiplying the resulting value by 100. Furthermore, the maximum value used in normalization may be the maximum value of a base year rather than the maximum value of each year, and normalization may be performed such that the maximum value of the base year is 100. Furthermore, these normalization processes may be performed after excluding outliers.

[0117] Furthermore, instead of normalization, standardization may be performed to set the average for each year to 0 and the variance for each year to 1, and a deviation value for each year may be calculated. Processing of the Health and Productivity Management Navigation Score (the Health and Productivity Management Navigation Score in the form of the total or average value of the TF-IDF values) such as normalization, standardization, and deviation value calculation may be performed by any method as long as it makes it easier for the user to understand the magnitude and meaning of the value. Therefore, for example, processing may be performed to set the median for each year to 50, or to set the median for the base year to 50, etc.

[0118] <Configuration of processing means 30 / company response vectorization means 35: Figure 7>

[0119] As shown in Figure 7, the company response vectorization means 35 vectorizes each company's issue content data, policy implementation result data, and effectiveness verification result data stored in the questionnaire data storage means 51, and performs a process of storing the obtained issue content vector, policy implementation result vector, and effectiveness verification result vector in the company response vector storage means 57 in association with the company response identification information.

[0120] Although not shown in Figure 7, the company response vectorization means 35 also vectorizes the combined text data created by the combining means 31, and also stores the resulting combined text vector in the company response vector storage means 57 in association with the company response identification information.

[0121] Specifically, the company response vectorization means 35 transmits the issue content data, the policy implementation result data, the effect verification result data, and the combined text data to the vectorization processing system 71 via the network 1. Then, the company response vectorization means 35 receives the issue content vector, the policy implementation result vector, the effect verification result vector, and the combined text vector transmitted from the vectorization processing system 71 via the network 1, and stores them in the company response vector storage means 57 in association with the company response identification information. In this embodiment, all vectors are 1536-dimensional vectors.

[0122] <Configuration of processing means 30 / policy implementation result clustering means 36: Fig. 7>

[0123] The policy implementation result clustering means 36 performs clustering using the policy implementation result vectors of each company stored in the company response vector storage means 57 (see Figure 7), assigns a policy implementation result cluster number to each company's policy implementation result vector to identify the classified policy implementation result cluster, and stores the assigned policy implementation result cluster number in the company response vector storage means 57 (see Figure 7) in association with the company response identification information.

[0124] Specifically, in this embodiment, the policy implementation result clustering means 36 performs clustering using BERTopic, as an example. BERTopic is a type of sentence clustering algorithm that converts text data into high-dimensional vectors, performs dimensional reduction (dimensional compression), and clusters the resulting low-dimensional vectors using an algorithm such as Hdbscan. The process of converting text data into high-dimensional vectors (1,536 dimensions in this embodiment) is performed by the aforementioned company response vectorization means 35, and therefore the vectorization processing system 71 (in this embodiment, Azure OpenAI Embeddings API text-embedding-ada-002 version 2) is used. In this embodiment, the low-dimensional vectors created by BERTopic are, as an example, five-dimensional vectors.

[0125] In addition, in order to create policy implementation result clusters, clustering is performed using policy implementation result vectors. However, classifying policy implementation result vectors also means classifying policy implementation result data corresponding to the policy implementation result vectors, and further classifying company response identification information corresponding to the policy implementation result data. Therefore, in the explanation of this application, these terms are used as equivalent expressions.

[0126] This clustering is performed across the issue themes (10 types). The number of policy implementation result clusters is determined by the system administrator (developer), and in this embodiment is, for example, about 300, and therefore the number of policy implementation result cluster numbers is also the same, about 300. Therefore, if there are, for example, about 4,200 policy implementation result vectors and corresponding policy implementation result data and company response identification information, these 4,200 or so will be classified into, for example, about 300.

[0127] The same policy implementation result cluster number is assigned to policy implementation result vectors and their corresponding policy implementation result data and company response identification information that belong to the same policy implementation result cluster. In the present description, as shown in FIG. 7, clustering is performed using policy implementation result vectors. Therefore, the policy implementation result vectors are first considered to be the direct objects of classification, and the policy implementation result cluster numbers are stored in association with company response identification information in the company response vector storage means 57. However, since the purpose of clustering is to cluster policies, the policy implementation result data may be considered to be classified, and the policy implementation result cluster numbers may be stored in association with company response identification information together with the policy implementation result data in the survey data storage means 51 (see FIG. 3). This storage format is equivalent. Furthermore, a policy implementation result cluster belonging information storage means (not shown) may be provided that stores the correspondence between company response identification information and policy implementation result cluster numbers (i.e., belonging information indicating which policy implementation result cluster the company response identification information belongs to). This is also equivalent. Therefore, in the present invention (claim description), it is stated that "the assigned policy implementation result cluster number is associated with the company response identification information and stored in the company response vector storage means," but this description includes the equivalent storage forms described above.

[0128] <Configuration of the processing means 30 / recommendation candidate set generation means 37: Fig. 9>

[0129] The recommendation candidate set generation means 37 performs different processing depending on whether the user receiving the policy recommendation service inputs a query on the screen displayed by the output means 45, such as a company (usually, the user who wants to improve their company's health management efforts will specify their own company, but any company can be specified) or an issue (usually, the user will specify an issue their company is facing, but any issue can be specified).

[0130] (Processing when the received query is for a company specified by the user) The recommendation candidate set generation means 37 receives the company response identification information as a query from the output means 45 when the output means 45 receives, as a query, a company (company identification information) related to the user's specification who will receive the policy recommendation service (see company name input section 321 in Figure 16), and further, when the company related to the received user specification has entered multiple (two) response data (including task content data) in the health management level survey form, and receives information related to the user's task selection (company response identification information corresponding to the selected task content data) (see task selection section 322 in Figure 16).

[0131] Then, the recommendation candidate set generation means 37 calculates the similarity (task similarity) between the task content vector of the company specified by the user received from the output means 45 (task content vector corresponding to the received company response identification information) and each of the task content vectors of the other multiple companies (task content vectors corresponding to company response identification information other than the received company response identification information), sorts the company response identification information for the other multiple companies in order of the calculated task similarity, and selects company response identification information in order of the highest task similarity until the number of policy implementation result clusters for the policy implementation result cluster numbers corresponding to the sorted company response identification information becomes the same as the number of required policies k specified by the user (see number of policies input section 311 in Figure 16) or the predetermined number of required policies k, thereby performing a process of generating a recommendation candidate set consisting of policy implementation result data belonging to the top k policy implementation result clusters.

[0132] Specifically, as shown in Figure 9, if the query received from the output means 45 is a recommendation of measures for the problem content data = "There are many employees who leave their jobs due to depression" (company response identification information i = 00003300) of the company specified by the user (usually the company itself, which will be referred to as Company X), the recommendation candidate set generation means 37 performs the following processing.

[0133] 9, the recommendation candidate set generation means 37 first calculates the similarity (task similarity) between the task content vector corresponding to the company response identification information i=00003300 of the company (Company X) specified by the user, which is stored in the company response vector storage means 57 (see FIG. 7), and the task content vector corresponding to the company response identification information i=00115200 (Company C), 00115300 (Company D), 00115400 (Company E), 00115500 (Company F), 00115600 (Company G), ... of all other companies (Company C, Company D, Company E, Company F, Company G, ...). In this embodiment, cosine similarity is calculated as an example. The calculated task similarity is stored in the main memory in association with the company response identification information.

[0134] Next, the recommendation candidate set generation means 37 rearranges (sorts) the company response identification information for the other multiple companies (Company C, Company D, Company E, Company F, Company G, ...) in descending order of the calculated similarity of the problem. The sorting result is shown near the center of Figure 9. After sorting, the data itself indicating the similarity values ​​of the problem (0.94, 0.90, 0.70, 0.60, ...) are arranged from the top in descending order of value, and the corresponding company response identification information is also stored in the main memory in that order (Company C, Company G, Company D, Company E, ...).

[0135] Also shown near the center of Figure 9 is the issue content data and policy implementation result data stored in the questionnaire data storage means 51 (see Figure 3) in association with the sorted company response identification information in the order of the sorted company response identification information. However, the issue content data and policy implementation result data do not necessarily have to be stored in the main memory at this stage, and may be obtained from the questionnaire data storage means 51 using the company response identification information in subsequent processing. Furthermore, the policy implementation result cluster numbers obtained from the company response vector storage means 57 (see Figure 7) using this company response identification information are also arranged in the order of the sorted company response identification information and stored in the main memory.

[0136] Note that the issue content vectors used to calculate the issue similarity, the issue content data, the policy implementation result data, and the corresponding company response identification information have a one-to-one correspondence, so sorting the company response identification information in descending order of issue similarity is equivalent to sorting the issue content vectors, issue content data, and policy implementation result data in that order. Also, the company response identification information and the policy implementation result cluster numbers do not have a one-to-one correspondence; for example, there is a relationship between approximately 4,200 pieces of company response identification information and 300 policy implementation result cluster numbers. However, since one policy implementation result cluster number is determined when a company response identification information is determined, sorting the company response identification information will result in the policy implementation result cluster numbers being correspondingly sorted.

[0137] Therefore, the purpose of sorting when preparing a policy recommendation by the recommendation candidate set generation means 37 is to sort the policies (policy implementation result data) in order of similarity of the issues (issue content vectors, i.e., issue content data), but as described above, this sorting is equivalent to sorting (sorting) of company response identification information. For this reason, the present invention (claim description) states that "the company response identification information for multiple other companies is sorted in order of the highest similarity of the calculated issues," but this description includes equivalent sorting.

[0138] Then, the recommendation candidate set generation means 37 executes a process of generating a recommendation candidate set using the number of necessary measures k (see the number of measures input unit 311 in FIG. 16) specified by the user received from the output means 45 and the arranged policy implementation result cluster numbers (the policy implementation result cluster numbers arranged corresponding to the company response identification information sorted in descending order of the similarity of the issues). Note that in this embodiment, the number of necessary measures k is specified by the user, but it may also be a fixed number of necessary measures k that is predetermined by the system.

[0139] That is, when the company response identification information sorted in descending order of problem similarity is arranged in the main memory, the policy implementation result cluster numbers corresponding to these company response identification information are also arranged in the main memory in the same order, so the recommendation candidate set generation means 37 selects k policy implementation result cluster numbers in descending order (in descending order of problem similarity) and selects the company response identification information corresponding to them. That is, the recommendation candidate set generation means 37 selects company response identification information corresponding to the numbers that appear in ascending order from the top until the kth policy implementation result cluster number appears. In other words, the recommendation candidate set generation means 37 selects company response identification information in descending order of problem similarity until the number of policy implementation result clusters becomes the same as the required number k of policies. At this time, for the kth policy implementation result cluster number (the last number), it selects all of the company response identification information corresponding to it as long as they appear consecutively from the first appearance.

[0140] Then, a set of policy implementation result data (or various data including policy implementation result data) corresponding to the selected company response identification information is set as a recommendation candidate set. If the corresponding policy implementation result data is not stored in the main memory, a set of policy implementation result data acquired from the survey form data storage means 21 using the selected company response identification information is set as a recommendation candidate set. This makes it possible to generate a recommendation candidate set consisting of policy implementation result data belonging to the top k policy implementation result clusters.

[0141] In this case, if the same number appears multiple times for the first to (k-1)th policy implementation result cluster numbers before the kth policy implementation result cluster number appears, all of the company response identification information corresponding to those policy implementation result cluster numbers are selected. Therefore, if the required number of policies is k=5, even if the same number appears two or more times for the first to fourth policy implementation result cluster numbers before the fifth policy implementation result cluster number appears, all of the company response identification information corresponding to them are selected, and a recommendation candidate set is generated.

[0142] Therefore, the first number to appear (the policy implementation result cluster number for the project with the highest issue similarity) is naturally the first policy implementation result cluster number, and although it is not known whether the next number to appear will be the same as the first policy implementation result cluster number or the second policy implementation result cluster number, even if it is the same as the first policy implementation result cluster number, the company response identification information corresponding to that number is selected to generate a recommendation candidate set. For example, if the first number appears four times, the second number appears three times, the third number appears two times, and the fourth number appears five times before the fifth policy implementation result cluster number (the last number) first appears, the company response identification information corresponding to these 4 + 3 + 2 + 5 = 14 occurrences is selected. Furthermore, if the fifth number (the last number) appears three times consecutively, including the first appearance, the company response identification information corresponding to these three occurrences is selected, so a total of 14 + 3 = 17 company response identification information is selected to generate a recommendation candidate set.

[0143] In the example of Figure 9, when [1] the number of required measures k = 1, only the measure implementation result cluster number = 3 (the first number) is selected, and the various data (issue content data, measure implementation result data, etc.) linked to the company response identification information i = 00115200, 00115600 (Company C, Company G) corresponding to the measure implementation result cluster number = 3 becomes the recommendation candidate set. Note that when k = 1, the first measure implementation result cluster number is also the last number, so as long as the last number = 3 appears consecutively (in the example of Figure 9, it appears twice consecutively), all of the company response identification information corresponding to it are selected, and the various data linked to them (issue content data, measure implementation result data, etc.) become the recommendation candidate set.

[0144] [2] When the number of required measures k = 2, the policy implementation result cluster numbers = 3 (first number) and 26 (second number) are selected, and the various data (issue content data, policy implementation result data, etc.) linked to the company response identification information i = 00115200, 001156000, 0115300 (Company C, Company G, Company D) corresponding to policy implementation result cluster numbers = 3 and 26 become the set of recommendation candidates.

[0145] [3] When the number of required measures k = 3, the policy implementation result cluster numbers = 3 (first number), 26 (second number), and 188 (third number) are selected, and the various data (issue content data, policy implementation result data, etc.) linked to the company response identification information i = 00115200, 001156000, 0115300, 00115400 (Company C, Company G, Company D, Company E) corresponding to policy implementation result cluster numbers = 3, 26, and 188 become the set of recommendation candidates.

[0146] As described above, processing is performed using the policy implementation result cluster number, i.e., processing is performed in units of policy implementation result clusters, and the required number of policies is k rather than the required number of clusters.This is because the recommended policy selection means 38, which will be described later, selects only one policy implementation result data belonging to each of the top k selected policy implementation result clusters as the policy to be recommended.

[0147] (Processing when the received query is an issue entered by a user) When the output means 45 receives a task entered by the user as a query (see the health management task input section 351 in Figure 18), the recommendation candidate set generation means 37 receives health management task input data (text data) as a query from the output means 45.

[0148] The recommendation candidate set generation means 37 then vectorizes the health management issue input data received from the output means 45 and calculates the similarity (issue similarity) between the obtained health management issue input vector and the issue content vectors of all companies stored in the company response vector storage means 57 (see FIG. 7). In this embodiment, cosine similarity is calculated as an example. The calculated issue similarity is stored in the main memory in association with the company response identification information.

[0149] Here, the process of vectorizing the health and productivity management issue input data is the same as the process of creating issue content vectors by the aforementioned company response vectorization means 35. This is because, because cosine similarity is calculated, the health and productivity management issue input vectors created must have the same number of dimensions and vectorization algorithm as the issue content vectors. Therefore, if Azure OpenAI Embeddings API text-embedding-ada-002 version 2 is used as the vectorization processing system 71 to create the issue content vectors, it is also used to create the health and productivity management issue input vectors.

[0150] Furthermore, the processing after calculating the similarity of the assignment is the same as that described above (the processing when the received query is for a company designated by the user). However, in the processing described above (the processing when the received query is for a company designated by the user), the company designated by the user is input, so the company response identification information for other companies is sorted in descending order of assignment similarity, whereas in this processing (the processing when the received query is for an assignment entered by the user), the assignment is input instead of the company designated by the user, so the company response identification information for all companies is sorted in descending order of assignment similarity.

[0151] In the policy extraction result storage means 60 (see FIG. 8), policy name data and policy tags created from policy implementation result data are stored in association with company response identification information, but it is also possible to create in advance from policy content data using the same method as for creating the policy name data and policy tags, and store them in association with company response identification information in the policy extraction result storage means (not shown). In other words, the policy name data, policy tag, and policy extraction result storage means (not shown) correspond to the policy name data, policy tag, and policy extraction result storage means 60 (see FIG. 8), respectively. Then, the recommendation candidate set generation means 37 determines whether or not there is any task name data or task tag stored in the task extraction result storage means (not shown) that matches the health and productivity management task input data entered by the user. If there is a match, the recommendation candidate set generation means 37 extracts company response identification information corresponding to the matching task name data or task tag, acquires the policy implementation result data and various data (including the overall evaluation or other evaluation values) corresponding to the extracted company response identification information from the survey form data storage means 51 (see FIG. 3), acquires the health and productivity management navigation score stored in the score storage means 56 (see FIG. 6), and further acquires the policy implementation result class stored in the company response vector storage means 57 (see FIG. 7). The extracted company response identification information is then obtained, and sorted in descending order of highest Health Management Navi score or overall rating or other evaluation value. After sorting, the company response identification information when the first policy implementation result cluster number first appears is selected, followed by the company response identification information when the second policy implementation result cluster number first appears, ..., and the company response identification information when the k-th policy implementation result cluster number first appears is selected from the top until the same number of policy implementation result cluster numbers as the required number k of policies appear. A recommendation candidate set may be generated using the k company response identification information obtained, based on the policy implementation result data and various data corresponding to those company response identification information.

[0152] <Configuration of the processing means 30 / recommended measure selection means 38: FIG. 10>

[0153] The recommendation target policy selection means 38 selects the policy implementation result data (which may be considered as various data including policy implementation result data) with the highest health management navigation score or overall rating or other evaluation value (see sorting criteria selection unit 312 in Figure 16) from each of the top k policy implementation result clusters selected by the recommendation candidate set generation means 37 as the recommendation target policy, and executes a process of passing the selected policy implementation result data to the output means 45.

[0154] Here, the health and productivity management navigation score, overall rating, or other rating value is a criterion selected by the user in the sorting criterion selection unit 312 in Fig. 16. Note that the sorting criterion in Fig. 16 is a criterion for determining recommended measures within the same policy implementation result cluster after sorting by issue similarity, and is an option for determining what type of rating value (health and productivity management navigation score, overall rating, etc.) to use to select the highest value within the same policy implementation result cluster.

[0155] Specifically, the recommendation target policy selection means 38 receives from the recommendation candidate set generation means 37 multiple pieces of company response identification information corresponding to multiple pieces of policy implementation result data constituting the recommendation candidate set and the corresponding top k policy implementation result cluster numbers, and selects the company response identification information (company response identification information belonging to the same policy implementation result cluster) assigned the same policy implementation result cluster number with the highest health and productivity management navigation score, overall rating, or other evaluation value. At this time, the health and productivity management navigation score is obtained from the score storage means 56 (see FIG. 6) using the company response identification information, and the overall rating or other evaluation value is obtained from the survey form data storage means 51 using the company response identification information. Then, this selection process within the same cluster is performed for all of the top k policy implementation result clusters (policy implementation result cluster numbers), and the policy implementation result data and various data corresponding to the obtained k company response identification information are selected as recommendation targets and passed to the output means 45.

[0156] The sorting results shown in Figure 10 are the same as those shown in Figure 9 described above, but columns for the Health and Productivity Management Navigation Score and Overall Evaluation are added. In the example of Figure 10, the policy implementation result cluster with policy implementation result cluster number = 3 (the first number) contains two pieces of company response identification information, but if the Health and Productivity Management Navigation Score is selected as the criterion for determining the recommended policy, the company response identification information i = 00115200 (Company C) has the highest Health and Productivity Management Navigation Score in this policy implementation result cluster, so the policy implementation result data and various data corresponding to 00115200 (Company C) are selected as the policy to be recommended and passed to output means 45.

[0157] In addition, the policy implementation result cluster with policy implementation result cluster number = 26 (the second number) contains only one company response identification information i = 00115300 (Company D), which is the company response identification information with the highest health management navigation score, and the policy implementation result data and various data corresponding to this 00115300 (Company D) are selected as the policy to be recommended and passed to output means 45.

[0158] Similarly, the policy implementation result cluster with policy implementation result cluster number = 188 (the third number) contains only one company response identification information i = 00115400 (Company E), which is the company response identification information with the highest health management navigation score, and the policy implementation result data and various data corresponding to this 00115400 (Company E) are selected as the policy to be recommended and passed to output means 45.

[0159] <Configuration of processing means 30 / press release acquisition means 39:: Figure 10>

[0160] 10, the press release acquisition means 39 sequentially transmits a plurality of keywords (many keywords related to health and productivity management) stored in the keyword storage means 58 to the press release distribution system 80 via the network 1, receives press release data, its title data, and date data transmitted from the press release distribution system 80 via the network 1, and stores the received press release data, its title data, and date data in the press release storage means 59. The process of registering keywords in the keyword storage means 58 is performed in advance by the system administrator (developer).

[0161] <Configuration of the processing means 30 / title vectorization means 40: FIG. 10>

[0162] As shown in Figure 10, the title vectorization means 40 vectorizes the title data of each press release acquired by the press release acquisition means 39, and performs a process of storing the obtained title vector in the press release storage means 59 in association with the data of the acquired press release.

[0163] Here, the process of vectorizing the title data is the same as the process of creating the policy implementation result vector by the company response vectorization means 35 described above. Because the cosine similarity with the policy implementation result vector is calculated by the press release selection means 41 described below, the created title vector must have the same number of dimensions and vectorization algorithm as the policy implementation result vector. Therefore, if Azure OpenAI Embeddings API text-embedding-ada-002 version 2 is used as the vectorization processing system 71 to create the policy implementation result vector, it is also used to create the title vector.

[0164] <Configuration of the processing means 30 / press release selection means 41: Figure 10>

[0165] As shown in Figure 10, the press release selection means 41 calculates the similarity (in this embodiment, cosine similarity) between each title vector stored in the press release storage means 59 and the policy implementation result vector stored in the company response vector storage means 57 (see Figure 7) for the policy implementation result data (and the corresponding company response identification information) selected by the recommended policy selection means 38, selects a predetermined number (p items) or a number (p items) specified by the user that have a high similarity between the calculated title and policy, obtains the data, title data, and date data of the selected press release from the press release storage means 59, and passes them to the output means 45 for screen display (see press release display section 331C in Figure 17).

[0166] <Configuration of processing means 30 / measure name extraction means 42: FIG. 8>

[0167] In order to extract health management-related measures from the content expressed in the policy implementation result data of each company stored in the questionnaire data storage means 51 (see Figure 3), the policy name extraction means 42 creates policy extraction request data (including the policy implementation result data and a request text requesting that health management-related measures be extracted from the text contained in the policy implementation result data) for each company's policy implementation result data, inputs the created policy extraction request data into a large-scale language model (LLM), and executes a process of storing the LLM response data (text data indicating the extracted policy or policies) output from the large-scale language model (LLM) as policy name data in association with company response identification information in the policy extraction result storage means 60 (see Figure 8).

[0168] Specifically, the policy name extraction means 42 uses any one of the policy implementation result data acquired from the questionnaire data storage means 51 (see FIG. 3) to create one policy extraction request data as shown in FIG. 8, transmits the created one policy extraction request data to the large-scale language model (LLM)-based service provision system 70 via the network 1, receives LLM response data transmitted from the service provision system 70 via the network 1, and stores text data indicating one or more policies included in the received LLM response data as policy name data in the policy extraction result storage means 60 (see FIG. 8) in association with the company response identification information. In the example of FIG. 8, two policy name data, "Walking competition, distribution of pedometers" (pedometer is a registered trademark), are created (extracted) using the policy implementation result data "For the purpose of promoting exercise habits and communication among employees..." for company response identification information = 00132200 (Company L). This process is then repeated for all policy implementation result data (all company response identification information).

[0169] <Configuration of processing means 30 / measure tag creation means 43: Fig. 8>

[0170] The policy tag creation means 43 vectorizes each of the policy name data extracted by the policy name extraction means 42, performs clustering using the obtained policy name vectors, and further creates policy tag naming request data (including a set of policy name data belonging to the same policy name cluster and a request to name this set) for each of the multiple policy name clusters formed by classification, inputs the created policy tag naming request data into a large-scale language model (LLM), assigns the LLM response data (text data indicating the name of the policy name cluster) output from the large-scale language model (LLM) as a policy tag to each policy name data belonging to the policy name cluster, and executes a process of associating the assigned policy tag with company response identification information and storing it in the policy extraction result storage means 60 (see Figure 8).

[0171] Specifically, the policy tag creation means 43 first vectorizes each of the policy name data extracted by the policy name extraction means 42 and stored in the policy extraction result storage means 60 (see FIG. 8 ) to create a policy name vector. This vectorization process is the same as the process of creating a policy implementation result vector by the company response vectorization means 35 described above. Therefore, the policy tag creation means 43 transmits policy name data one by one to the vectorization processing system 71 via the network 1, receives the policy name vector transmitted from the vectorization processing system 71 via the network 1, and stores the received policy name vector together with the corresponding policy name data in the policy name / policy tag correspondence storage means 61 (see FIG. 8 ). This process is performed for all policy name data. Note that this process is performed for all policy name data, including cases where multiple policy name data correspond to one company response identification information (such as 00132200 (Company L) and 00132600 (Company Q) in the example of FIG. 8 ). In this embodiment, Azure OpenAI Embeddings API text-embedding-ada-002 version 2 is used as the vectorization processing system 71.

[0172] Next, the policy tag creation means 43 performs clustering using all policy name vectors created from all policy name data. In this embodiment, this clustering process is performed using a method similar to that used by the policy implementation result clustering means 36 described above, i.e., using BERTopic. Therefore, using BERTopic, the high-dimensional policy name vectors (1,536 dimensions in this embodiment) stored in the policy name-policy tag correspondence storage means 61 (see FIG. 8) are dimension-removed (dimensionally compressed) to create low-dimensional policy name vectors (e.g., five-dimensional vectors), and clustering is performed using the resulting low-dimensional policy name vectors. Then, policy name cluster numbers for identifying the multiple policy name clusters thus classified are stored in the policy name-policy tag correspondence storage means 61 (see FIG. 8) in association with the policy name data and policy name vectors belonging to each policy name cluster.

[0173] In this case, the number of policy name clusters formed by clustering is determined by the system administrator (developer), and is, for example, about 300. The total number of policy name data is, for example, about 10,000, and from the 10,000 policy name data, about 300 policy name clusters and policy name cluster numbers are created, and as described below, the same number of policy tags, about 300, are created.

[0174] Next, the policy tag creation means 43 creates policy tag naming request data (including a set of policy name data belonging to the same policy name cluster and a request to name this set) as shown in Fig. 8 for each of the multiple policy name clusters that have been classified. Since the same policy name cluster number is assigned to policy name data that belong to the same policy name cluster, the policy tag naming request data is created using a set of policy name data with the same policy name cluster number stored in the policy name / policy tag correspondence storage means 61 (see Fig. 8).

[0175] In the example of Figure 8, the three policy name data, "Walking Competition," "Walking Rally," and "Walking Event," are assigned the same policy name cluster number = 41, so the collection of these three policy name data forms one policy name cluster.

[0176] Then, the policy tag creation means 43 transmits the created policy tag naming request data to a service provision system 70 using a large-scale language model (LLM) via the network 1, receives the LLM response data (text data indicating the name of the policy name cluster) transmitted from the service provision system 70 via the network 1, assigns the received text data as a policy tag to each policy name data belonging to the policy name cluster, and stores the assigned policy tag in the policy extraction result storage means 60 (see Figure 8) in association with the company response identification information.

[0177] In the example of Figure 8, for a set of three policy name data, "walking competition," "walking rally," and "walking event," LLM response data (policy tag) "walking event" was obtained, so this "walking event" is stored in the policy name / policy tag correspondence storage means 61 (see Figure 8) in correspondence with "walking competition," "walking rally," and "walking event."

[0178] Furthermore, the correspondence between company response identification information and policy name data may not be a one-to-one relationship but may be a one-to-many relationship, and since the policy extraction result storage means 60 (see FIG. 8) stores one or more policy name data in association with one company response identification information, this policy extraction result storage means 60 also stores one or more policy tags assigned to the policy name data in a state where they are associated with one or more policy name data. Therefore, for example, in the policy extraction result storage means 60 shown in FIG. 8, for company response identification information = 00132200 (Company L), the policy tag = "walking event" is stored in association with policy name data = "walking competition," and the policy tag = "step count management" is stored in association with policy name data = "pedometer distribution" (pedometer is a registered trademark).

[0179] <Configuration of the processing means 30 / measure tagging related measure search means 44>

[0180] The measure tag lookup related measure search means 44 receives the measure tag designated by the user from the output means 45 (see the evaluation result display section 331B in FIG. 17 and the measure tag list display section 380 in FIG. 19), and uses the received measure tag to extract company response identification information associated with the designated measure tag and stored in the measure extraction result storage means 60 (see FIG. 8) and / or measure implementation result data associated with the company response identification information and stored in the survey form data storage means 51 (see FIG. 3), and extracts the extracted company response identification information and / or measure implementation result data from the extracted company response identification information and / or measure implementation result data. The system selects a predetermined number (u items) or a number (u items) specified by the user of company response identification information and / or policy implementation result data in descending order of the health management navigation score associated with the identification information and stored in the score storage means 56 (see Figure 6), or in descending order of the overall rating or other evaluation value stored in the questionnaire data storage means 51 (see Figure 3), and executes a process of passing the selected u company response identification information and / or policy implementation result data to the output means 45 for display on the screen (see Figure 22) as a search result for related policies using policy tags.

[0181] <Configuration of Processing Means 30 / Output Means 45>

[0182] Output means 45 executes a process of displaying information about health and productivity management on the screen of user terminal 91 using the health and productivity management navigation score, overall rating, or other evaluation values ​​(various deviation values, etc.). Here, the health and productivity management navigation score is stored in score storage means 56 (see FIG. 6), and the overall rating or other evaluation values ​​(various deviation values, etc.) are stored in survey form data storage means 51 (see FIG. 3).

[0183] More specifically, the output means 45 performs the following processes: (1) displaying on a screen at least one of the company name (stored in the company information storage means 52), task content data, policy implementation result data, effectiveness verification result data, and combined text data corresponding to the company response identification information, along with the health management navi score or overall rating or other evaluation value (various deviation values, etc.); (2) displaying on a screen at least one of the company name, task content data, policy implementation result data, effectiveness verification result data, and combined text data corresponding to the company response identification information according to the results of a selection, extraction, sorting, or grouping process using the health management navi score or overall rating or other evaluation value (various deviation values, etc.); and (3) displaying on a screen the distribution of each company's health management navi score or overall rating or other evaluation value (various deviation values, etc.), or the changes over time in each company's health management navi score, overall rating, or other evaluation value (various deviation values, etc.).

[0184] (Explanation of Figure 13) 13 shows various screens displayed by the output means 45 and the transition status of those screens. First, a top screen 100 is displayed by the output means 45, and from this top screen 100, transitions can be made to a company profile screen 200 (see FIGS. 14 and 15), a policy recommendation screen 300 (see FIGS. 16 to 19), a similar company search screen 400 (see FIG. 20), and a health and productivity management navigation score screen 500 (see FIG. 21). In addition, transitions can be made from the policy recommendation screen 300 to a policy tag-related policy search result display screen 600 (see FIG. 22), an analysis result detail display screen 700 (see FIG. 23), and a press release display screen 800.

[0185] In Figure 13, the company profile screen 200 (see Figures 14 and 15) has a company information display section 210, a White 500 display section 220, a health management brand display section 230, an annual trend display section 240, a benchmark comparison display section 250, a Health Management Excellent Corporation Certification Standards achievement status display section 260, and a White 500 acquisition conditions display section 270.

[0186] The policy recommendation screen 300 (see Figures 16 to 19) has display sections for when the "Search by company name" tab 301, the "Search by input issue" tab 302, and the "Policy tag list" tab 303 are selected, as well as a condition setting section 310.

[0187] When the "Search by company name" tab 301 is selected, the display section (see FIGS. 16 and 17) includes a company issue extraction display section 320, a measure recommendation result display section 330, and a proposed measure summary display section 340. The measure recommendation result display section 330 includes a measure 1 display section 331 (including a related press release display section), a measure 2 display section 332 (including a related press release display section), ..., a measure k display section (including a related press release display section). k is the number of required measures specified by the user, and the proposed measure summary display section 340 is a summary of the k recommended measures.

[0188] When the "Search from input issue" tab 302 is selected, the display section (see FIG. 18) includes a company issue input section 350, a measure recommendation result display section 360, and a proposed measure summary display section 370. The measure recommendation result display section 360 includes a measure 1 display section 361 (including a related press release display section), ..., a measure k display section (including a related press release display section). k is the number of required measures specified by the user, and the proposed measure summary display section 370 is a summary of the k recommended measures.

[0189] When the "Measure tag list" tab 303 is selected, a measure tag list display section 380 is provided on the display section (see FIG. 19).

[0190] The similar company search screen 400 (see FIG. 20) is provided with a latest fiscal year data display section 410 and an age-related data display section 420.

[0191] (Explanation of Figure 14) The company profile screen 200 in FIG. 14 is provided with a company name input section 201, where the user inputs the name of the company about which the user wants to refer to information.

[0192] The company information display section 210 displays the corporate name (company name), industry name, whether or not the company is listed, and the ranking of the overall evaluation for each year for the company specified by the user input in the company name input section 201. The corporate name (company name), industry name, and whether or not the company is listed are stored in the company information storage means 52, and the overall evaluation is stored in the survey form data storage means 51 (see FIG. 3).

[0193] The White 500 display section 220 and the Health and Productivity Management brand display section 230 display whether or not the company designated by the user has been certified for each fiscal year.

[0194] The annual trend display unit 240 is provided with a vertical axis selection unit 241 for selecting the content to be plotted (content of the vertical axis of the graph), and a graph display unit 242. The graph display unit 242 displays the annual trend of the content selected in the vertical axis selection unit 241 (in the example of FIG. 14, the Health and Productivity Management Navigation Score). The annual trends of the company, the industry average, and the overall average specified by the user input in the company name input unit 201 are displayed. The Health and Productivity Management Navigation Score for the company is stored in score storage means 56 (see FIG. 6). The industry average and the overall average may be calculated and displayed by output means 45 each time they are displayed, or pre-calculated values ​​may be stored in score storage means 56 (for example, by providing virtual subject identification information such as industry average identification information or overall average identification information corresponding to the company response identification information, and storing the average value of the Health and Productivity Management Navigation Score in association with this virtual subject identification information in the same format as the values ​​for individual companies), or may be stored in statistical information storage means (not shown). In addition, if there are multiple (two) response data for the company specified by the user entered in the company name input section 201, the health management navigation score with the highest numerical value for the company may be plotted, or the average health management navigation score within the company may be plotted.

[0195] (Explanation of Figure 15) 15, the benchmark comparison display section 250 is provided with an item selection section 251 for selecting an item to be plotted, a graph display section 252 for the item selected in the item selection section 251, a difference display section 253 for displaying the difference in the deviation value between the user-specified company and the industry average for that item, and an advice display section 254 for the user-specified company. Various deviation values ​​are stored in the questionnaire data storage means 51 (see FIG. 3), and the industry average value may be calculated and displayed by the output means 45 each time it is displayed, or a pre-calculated value may be stored in the questionnaire data storage means 51 (for example, by providing virtual subject identification information such as industry average identification information or overall average identification information corresponding to company response identification information, and storing the average values ​​of various deviation values ​​in association with this virtual subject identification information in the same format as the values ​​for individual companies), or may be stored in a statistical information storage means (not shown). 15 shows an overall deviation value, but if processing to calculate a deviation value for the health and productivity management navigation score has also been performed, the health and productivity management navigation score (deviation value) may be displayed instead of or together with the overall deviation value. Advice display unit 254 may display content automatically generated by a large-scale language model (LLM) using the display content of difference display unit 253 or graph display unit 252, or may display one selected from a plurality of pre-created fixed sentences based on the content of difference display unit 253 or graph display unit 252.

[0196] The Health and Productivity Management Excellent Corporation Certification Standards achievement status display unit 260 is provided with a display unit 261 indicating whether each certification standard has been achieved or not, and an advice display unit 262 for the company designated by the user regarding the achievement of each certification standard. The advice display unit 254 may use the display content of the display unit 261 to display content automatically generated by a large-scale language model (LLM), or may display a sentence selected based on the content of the display unit 261 from a plurality of fixed sentences prepared in advance.

[0197] A fixed sentence that has been created and prepared in advance is displayed in the acquisition condition display section 270 of the White 500. The sentence may be written in the program or may be stored externally in memory.

[0198] (Explanation of Figure 16) In the measure recommendation screen 300 of Figure 16, the condition setting unit 310 for extracting measures to be recommended (endorsed) includes a measure number input unit 311 that accepts input of the number of measures k required as specified by the user (the company or its representative that receives the measure recommendation service), a sorting criteria selection unit 312 that selects the criteria for determining recommended measures within the same measure implementation result cluster after sorting by issue similarity, a change rate filtering selection unit 313 that selects whether to enable filtering by the rate of change in overall deviation value for the recommended measures (a process of narrowing down the measures to those of companies whose rate of change in overall deviation value is equal to or greater than a set value, i.e., a process of narrowing down the measures to those of companies whose overall deviation value has increased compared to the previous year), a listing category selection unit 314 that selects whether to include listed companies and unlisted companies in the recommended measures, a same industry restriction selection unit 315 that selects whether to limit the recommended measures to those within the same industry, and a task theme selection unit 316 that individually selects whether to include each of 10 types of task themes in the recommended measures. The assignment theme selection section 316 is provided with a "Check all checkboxes" button and a "Check all checkboxes" button, and clicking these makes it easy to select the assignment theme you need.

[0199] In this embodiment, when the change rate filtering selection unit 313 is selected, the setting value (threshold) of the change rate is set to a fixed value predetermined by the system, but the user may be allowed to set the setting value (threshold) of the change rate as desired.

[0200] By selecting from the change rate filtering selection section 313, listing category selection section 314, same industry restriction selection section 315, and issue theme selection section 316, the user can narrow down the candidates for measures to be recommended, so that measures of companies in similar situations to the user's own company or measures of companies that have significantly increased the evaluation of their measures can be selected as candidates for measures to be recommended.

[0201] 16, the company problem extraction display section 320 includes a company name input section 321 that accepts input of a company (usually the company itself) designated by the user, a problem selection section 322 that selects which problem to recommend a measure for when the company entered in the company name input section 321 has entered multiple (two) pieces of response data (including problem content data) in the health and productivity management survey form, and a company information display section 323 that displays information about the company designated by the user (corporate name, industry name, listed status, problem theme, problem content data, and measure implementation result data). The corporate name, industry name, and listed status are stored in the company information storage means 52, and the problem theme (problem theme number), problem content data, and measure implementation result data are stored in the survey form data storage means 51 (see FIG. 3).

[0202] (Explanation of Figure 17) 17, when the "Search by company name" tab 301 is selected, information on k measures (k is the number of measures required as specified by the user) selected by the recommendation target measure selection means 38 (including response data of companies related to the measures, various evaluation information derived from the response data, and other related information) is displayed in the measure recommendation result display section 330. In this case, the information on the k measures is displayed separately for each measure in a measure 1 display section 331, a measure 1 display section 332, ..., a measure k display section, and in addition, information integrating the k measures is displayed in a proposed measure summary display section 340.

[0203] More specifically, the measure 1 display section 331 that displays the details of the recommended measure 1 includes a company response display section 331A, an evaluation result display section 331B, and a press release display section 331C.

[0204] The company response display section 331A displays the problem theme, problem content data, measure implementation result data, and effect verification result data for the recommended measure 1. This information is stored in the survey form data storage means 51 (see FIG. 3).

[0205] The evaluation result display section 331B displays the overall evaluation, health and productivity management navigation score, task similarity, extracted measures (measure name data), and measure tag. The overall evaluation is stored in the survey data storage means 51 (see FIG. 3), the health and productivity management navigation score is stored in the score storage means 56 (see FIG. 6), the task similarity is the similarity of the task content vectors calculated by the recommendation candidate set generation means 37 (the similarity of the tasks shown in FIG. 9), and the extracted measures (measure name data) and measure tag are stored in the measure extraction result storage means 60 (see FIG. 8).

[0206] In the row of the health management navigation score, an "analysis result details" selection section is displayed, and when the user clicks here, an analysis result details display screen 700 (see FIG. 23) is displayed.

[0207] The user can select a policy tag by clicking it, and when the user clicks, the policy tag lookup related policy search means 44 performs processing, and a policy tag lookup related policy search result display screen 600 (see FIG. 22) is displayed.

[0208] Press release display section 331C displays title data and date data for each press release selected by press release selection means 41. The user can select title data by clicking, and when clicked, press release display screen 800 is displayed, which displays the data content of the press release (including the title, date, and text).

[0209] The measure 2 display section 332, which displays the details of the recommended measure 2, is provided with a company response display section, an evaluation result display section, and a press release display section, similar to the measure 1 display section 331. Furthermore, this display is similar up to the measure k display section.

[0210] The proposed measure summary display section 340 displays the contents of the LLM response data (summary data) obtained as an output of a summary creation request data input to a large-scale language model (LLM) using the contents displayed in the measure 1 display section 331, measure 2 display section 332, ..., measure k display section, requesting that a summary be created by summarizing the contents of measures 1 to k. This summary is displayed when the "Generate Summary" button is clicked.

[0211] (Explanation of Figure 18) 18, the company issue input section 350 has a health management issue input section 351 where the user inputs an issue related to health management (usually an issue the company is facing or a similar or related issue). Therefore, in this "search from input issue" tab 302, instead of inputting a company specified by the user, the user inputs an issue and measures by each company for an issue that is the same as or similar to the issue are recommended.

[0212] When the "Search by input issue" tab 302 is selected on the measure recommendation screen 300 in Fig. 18, the measure recommendation result display section 360 is provided with a measure 1 display section 361, a measure 2 display section 362, ..., a measure k display section. These are exactly the same as the display form of the measure recommendation result display section 330 when the "Search by company name" tab 301 is selected on the measure recommendation screen 300 in Fig. 17 described above. Therefore, for example, the measure 1 display section 361 is provided with a company response display section 361A, an evaluation result display section, and a press release display section.

[0213] Note that, as in the case where the "Search by company name" tab 301 described above is selected, information on k measures (k is the number of measures required as specified by the user) selected by the recommendation target measure selection means 38 is displayed (including response data of companies related to the measures and various evaluation information and other related information derived from the response data), but the content of the selection process of the recommendation target measures by the recommendation target measure selection means 38 is different. This point has already been described in detail in the explanation of the recommendation target measure selection means 38, so a detailed explanation will be omitted here.

[0214] In addition, the proposed measures summary display section 370 is the same as when the aforementioned "Search by company name" tab 301 is selected, and is displayed using the same processing as the proposed measures summary display section 340.

[0215] (Explanation of Figure 19) When the "Policy Tag List" tab 303 is selected on the policy recommendation screen 300 in Fig. 19, all policy tags and the number of companies implementing the policies to which those policy tags are attached are displayed in the policy tag list display section 380. The user can select these policy tags by clicking them, and when clicked, processing is performed by the policy tag lookup related policy search means 44, and a policy tag lookup related policy search result display screen 600 (see Fig. 22) is displayed. The policy tags are stored in the policy extraction result storage means 60 (see Fig. 8).

[0216] (Explanation of Figure 20) The similar company search screen 400 in FIG. 20 is provided with a company name input section 401 for inputting a company (usually the user's own company) designated by the user.

[0217] The latest year data display section 410 displays the results of a comparison of the latest year's scores (two of the following scores: aspect deviation score, detailed item deviation score, issue theme deviation score, overall deviation score, health management navigation score, etc.) with the company specified by the user entered in the company name input section 401 (Company Y in the example of Figure 20) and other companies (companies other than Company Y) that have a high similarity.

[0218] Therefore, the latest fiscal year data display section 410 is provided with a vertical axis selection section 411 for selecting the type of score to be used on the vertical axis, a horizontal axis selection section 412 for selecting the type of score to be used on the horizontal axis, and a distribution display section 413 for showing the distribution of two types of scores for the company specified by the user (Company Y) and similar companies (companies other than Company Y) in terms of visual distance. Note that three axes may be selected and a three-dimensional display may be displayed so that three types of scores can be compared, and in this case, the three-dimensional display may be made freely rotatable on the screen.

[0219] In this embodiment, the output means 45 uses the root mean square (RMS) as an example to calculate the similarity between the two scores selected by the vertical axis selection unit 411 and the horizontal axis selection unit 412. That is, the RMS value is calculated by adding the squared value of the difference between the two companies in terms of the scores on the vertical axis (the difference between Company Y and companies other than Company Y) to the squared value of the difference between the two companies in terms of the scores on the horizontal axis (the difference between Company Y and companies other than Company Y), dividing the result by 2 and taking the root of the resulting value. Furthermore, in the case of a three-axis three-dimensional display, the RMS value is calculated by adding the squared value of the difference between the two companies in terms of the scores on the first axis (the difference between Company Y and companies other than Company Y), the squared value of the difference between the two companies in terms of the scores on the second axis (the difference between Company Y and companies other than Company Y), and the squared value of the difference between the two companies in terms of the scores on the third axis (the difference between Company Y and companies other than Company Y), dividing the result by 3 and taking the root of the resulting value. If the distance indicated by this RMS value is small, it indicates that the similarity between the two companies is high, so the companies with the smallest RMS values ​​(a predetermined number of companies selected in ascending order) are determined as similar companies to Company Y and plotted in the distribution display section 413. Note that similar companies to be plotted may also be selected based on distances other than RMS.

[0220] Furthermore, the types of scores used to determine whether a company is a similar company may be different from the types of scores on the vertical and horizontal axes when plotting in distribution display section 413. For example, there may be five types of scores used to determine whether a company is a similar company (i.e., five types of scores for calculating the RMS), and two of the five types of scores may be used as the types of scores on the vertical and horizontal axes when plotting in distribution display section 413. Furthermore, one or both of the types of scores on the vertical and horizontal axes when plotting in distribution display section 413 may not necessarily be included in the types of scores for calculating the RMS. For example, an RMS may be calculated using three fixed types of scores to determine similar companies, and two arbitrary types of scores (which may not be included in the three fixed types) may be plotted and displayed in distribution display section 413. In order to display these flexibly, in addition to a vertical axis selection section 411 and a horizontal axis selection section 412 (in the case of a three-dimensional display, three axis selection sections for selecting the first to third axes) for selecting the vertical and horizontal axes when plotting in the distribution display section 413, a score selection section for determining similar companies may be provided for selecting any number of score types for determining similar companies.

[0221] The longitudinal data display section 420 is provided with an item selection section 421 for selecting the content to be plotted (the type of score for which the change over time is to be displayed) for the company (Company Y in the example of Figure 20) specified by the user entered in the company name input section 401, and a longitudinal change graph display section 422 for displaying a longitudinal change graph for the company (Company Y) specified by the user and similar companies (companies other than Company Y) for the item (type of score) selected in this item selection section 421.

[0222] (Explanation of Figure 21) The health management navigation score screen 500 in Figure 21 is provided with a correlation display target score selection section 501 for selecting a target score (score on the vertical axis) for displaying the correlation with the health management navigation score (score on the horizontal axis), a selection section 502 for selecting filtering by issue theme and inputting the issue theme, a selection section 503 for selecting filtering by measure tag and inputting the measure tag, and a correlation display section 510 for displaying the correlation between the health management navigation score and the target score selected in the correlation display target score selection section 501.

[0223] The correlation display section 510 divides the health management navigation score, which takes values ​​from 0 to 100, into 11 ranges, and displays the median (second quartile, 50% value) 511, first quartile (25% value) 512, third quartile (75% value) 513, minimum value 514, and maximum value 515 of the target score (in the example of Figure 21, the overall deviation value) in each range.

[0224] (Explanation of Figure 22) The policy tagging related policy search result display screen 600 in Figure 22 is provided with a display section 601 for the policy tag specified by the user used in the search (in the example of Figure 22, recommendation for medical examination / increase in medical examination rate), a display section 602 for the current sorting criteria for this page (in the example of Figure 22, health management navigation score), a sorting criteria selection section 603 for selecting (changing) the sorting criteria for this page only, and a related policy display section 610 for displaying the policies (company names and response data of companies related to the policies) extracted by the policy tagging related policy search means 44 using the policy tag specified by the user displayed in display section 601.

[0225] The related measure display section 610 displays the company name, assignment theme, assignment content data, assignment implementation result data, and effectiveness verification result data corresponding to the company response identification information obtained by searching using the policy tag, as well as the sorting criteria displayed in the display section 602 or the value of the sorting criteria subsequently selected (changed) in the sorting criteria selection section 603. Therefore, in the example of FIG. 22 , the policies (company names and response data of companies related to the policies) extracted by the policy tagging related policy search means 44 are displayed in descending order of the value of the health management navigation score, which is the current sorting criteria for this page. Furthermore, if a different sorting criteria (e.g., overall evaluation) is selected in the sorting criteria selection section 603, sorting is performed according to the changed sorting criteria (e.g., overall evaluation). Note that changing the sorting criteria on the policy tagging related policy search result display screen 600 applies only to this page; changing the overall sorting criteria is performed in the sorting criteria selection section 312 of the condition setting section 310 in FIG. 16 .

[0226] (Explanation of Figure 23) The analysis result details display screen 700 in FIG. 23 displays perspective evaluation result data (however, binary data of 1 or 0 is converted to O and X) for each perspective (9 perspectives in this embodiment) for the company specified by the user (usually the company itself) and the companies of recommended measures 1 to k (k is the number of necessary measures specified by the user), as well as the number of satisfied perspectives (how many of the 9 perspectives were satisfied), the health and productivity management navigation score, and the overall evaluation. The perspective evaluation result data is stored in the perspective evaluation result storage means 55 (see FIG. 5). The number of satisfied perspectives is displayed by adding up the perspective evaluation result data using the output means 45.

[0227] <Configuration of storage means 50 / survey table data storage means 51: Figure 3>

[0228] As shown in Figure 3, the survey data storage means 51 stores each company's response data, which include the issue theme number, issue content data, policy implementation result data, and effectiveness verification result data, as well as data showing the evaluation results for the response data (the overall evaluation value calculated by the evaluation body, various deviation values, for example, the overall deviation value, aspect 1 management philosophy / policy deviation value, aspect 2 organizational structure deviation value, aspect 3 system / policy implementation deviation value, aspect 4 evaluation / improvement deviation value, etc.), in association with company response identification information.

[0229] Other examples of the deviation values ​​stored in the survey form data storage means 51 include, for example, detailed item 1_1: clarification and internal penetration deviation value, detailed item 1_2: information disclosure and dissemination to other companies deviation value, detailed item 2_1: management involvement deviation value, detailed item 2_2: implementation system deviation value, detailed item 2_3: employee penetration deviation value, detailed item 3_1: goal setting, utilization of health checkups and medical examinations, etc. deviation value, detailed item 3_2: building a foundation for practicing health management deviation value, detailed item 3_3: health guidance deviation value, detailed item 3_4: lifestyle improvement deviation value, detailed item 3_5: other measures deviation value, detailed item 4_1: health checkups and stress checks deviation value, detailed item 4_2: working hours and leave of absence deviation value, detailed item 4_3: issue unit and overall measure effectiveness verification and improvement deviation value, etc.

[0230] In addition, there are also standard scores for Issue 1: preventing the occurrence of disease for all employees regardless of their health condition; Issue 2: preventing the worsening of diseases for those at high risk of lifestyle-related diseases and other illnesses; Issue 3: preventing the occurrence, early detection, and response to stress-related diseases such as mental health problems; Issue 4: preventing declines in employee productivity and preventing accidents; Issue 5: responding to health-related issues specific to women, maintaining and improving women's health; Issue 6: returning to work after leave of absence, balancing work and treatment; Issue 7: optimizing working hours, achieving a work-life balance, and ensuring time for personal life; Issue 8: promoting communication between employees; Issue 9: preventing infectious diseases among employees (such as influenza); and Issue 10: reducing the smoking rate among employees.

[0231] <Configuration of Storage Means 50 / Company Information Storage Means 52>

[0232] The company information storage means 52 stores the company name (in this application, referred to as company name, corporation name, company name, etc., including notation on drawings), industry name (industry identification information), whether or not the company is listed, whether or not it has been certified as a White 500 company, whether or not it has been certified as a health management stock, whether or not it has been certified as a health management excellent company, etc., in association with company identification information (which in this embodiment coincides with the first half of the company response identification information).

[0233] <Configuration of storage means 50 / perspective extraction result storage means 53: Fig. 4>

[0234] As shown in FIG. 4, the viewpoint extraction result storage means 53 stores the extracted viewpoint data in association with the pair of company response identification information used in the combination at the time of extraction.

[0235] <Configuration of storage means 50 / setting viewpoint storage means 54: Figs. 4 and 5>

[0236] As shown in Figures 4 and 5, the set viewpoint storage means 54 stores multiple (in this embodiment, nine) set viewpoint data (expression-aggregated viewpoint data) in association with viewpoint numbers (in this embodiment, 1 to 9).

[0237] <Configuration of storage means 50 / perspective evaluation result storage means 55: Figs. 5 and 6>

[0238] The viewpoint evaluation result storage means 55 stores viewpoint evaluation result data indicated by either of the two values ​​1 (satisfied) or 0 (not satisfied) in association with the company response identification information and viewpoint number.

[0239] <Configuration of storage means 50 / score storage means 56: Figure 6>

[0240] As shown in FIG. 6, the score storage means 56 stores the health management navigation score (human capital management navigation score) in association with the company response identification information.

[0241] <Configuration of storage means 50 / company response vector storage means 57: Figure 7>

[0242] As shown in Figure 7, the company response vector storage means 57 stores the issue content vector, the policy implementation result vector, the effectiveness verification result vector, as well as the combined text vector (not shown), and the policy implementation result cluster number in association with the company response identification information.

[0243] <Configuration of storage means 50 / keyword storage means 58: Fig. 10>

[0244] The keyword storage means 58 stores a plurality of keywords (numerous keywords related to health and productivity management). Examples of keywords to be stored include health and productivity management, collaborative health, presenteeism, absenteeism, human resources, health insurance, data health, healthcare, health tech, the 36 Agreement, EAP, action plan, incentive, walking event, exercise, health committee, engagement, online seminar, smoking cessation, no overtime day, recommendation to see a doctor, stress check, self-care, and flextime system.

[0245] <Configuration of storage means 50 / press release storage means 59: Figure 10>

[0246] The press release storage means 59 stores press release data acquired from the press release providing system 80, title data, date data, and title vectors created by the title vectorization means 40 in association with each other.

[0247] <Configuration of storage means 50 / measure extraction result storage means 60: Fig. 8>

[0248] As shown in FIG. 8, the measure extraction result storage means 60 stores company response identification information, one or more measure name data, and one or more measure tags in association with each other.

[0249] <Configuration of storage means 50 / measure name-measure tag correspondence storage means 61: FIG. 8>

[0250] The measure name-measure tag correspondence storage means 61 stores measure name data, measure name vectors, measure name cluster numbers, and measure tags in association with each other, as shown in Fig. 8. Therefore, it stores information on the attribution of each measure name data to a measure name cluster.

[0251] <Flow of preparation process before service provision by Health Management Scoring System 10: Figure 11>

[0252] 11, health management scoring system 10 performs the following process as preparation processing before providing a service to a user. This preparation processing is started by the system administrator (developer) operating administrator terminal 90, and the processing progresses.

[0253] First, the combining means 31 combines the issue content data, the policy implementation result data, and the effect verification result data included in the health management level survey form data (see FIG. 2) stored in the survey form data storage means 51 (see FIG. 3) to create combined text data (step S1). The details of this process have already been described in detail in the description of the combining means 31, so a detailed description will be omitted here.

[0254] Next, the viewpoint extraction means 32 uses the combined text data to perform pairwise evaluation using a large-scale language model (LLM) to extract viewpoints for evaluation, stores the extraction results in the viewpoint extraction result storage means 53 (see FIG. 4), and further uses the extracted viewpoint data to set multiple (nine) viewpoint data to be used in the evaluation, which are stored in the set viewpoint storage means 54 (see FIG. 4) (step S2). The details of this process have already been described in detail in the explanation of the viewpoint extraction means 32, so a detailed explanation will be omitted here.

[0255] Next, the viewpoint evaluation request means 33 uses the combined text data and a plurality (nine) of viewpoint data stored in the set viewpoint storage means 54 (see FIGS. 4 and 5) to evaluate whether the written content of the combined text data satisfies each viewpoint using a large-scale language model (LLM), and stores the obtained viewpoint evaluation result data (binary values ​​of 1 or 0) in the viewpoint evaluation result storage means 55 (see FIG. 5) (step S3). The details of this process have already been described in detail in the explanation of the viewpoint evaluation request means 33, so a detailed explanation will be omitted here.

[0256] Then, the score calculation means 34 calculates the TF-IDF using the perspective evaluation result data (binary values ​​of 1 or 0) stored in the perspective evaluation result storage means 55 (see FIGS. 5 and 6), and further calculates the total value, normalizes, and so on to calculate the health management navigation score, which is stored in the score storage means 56 (see FIG. 6) (step S4). The details of this process have already been described in detail in the explanation of the score calculation means 34, so a detailed explanation will be omitted here.

[0257] Thereafter, the company response vectorization means 35 vectorizes the issue content data, policy implementation result data, and effect verification result data stored in the questionnaire data storage means 51 (see FIG. 3) to create an issue content vector, a policy implementation result vector, and an effect verification result vector, which are then stored in the company response vector storage means 57 (see FIG. 7) (step S5). The details of this process have already been described in detail in the explanation of the company response vectorization means 35, so a detailed explanation will be omitted here.

[0258] Next, the policy implementation result clustering means 36 performs clustering using the policy implementation result vectors of each company stored in the company response vector storage means 57 (see FIG. 7), assigns a policy implementation result cluster number to each policy implementation result vector in accordance with the clustering results, and stores them in the company response vector storage means 57 (step S6). The details of this process have already been described in detail in the explanation of the policy implementation result clustering means 36, so a detailed explanation will be omitted here.

[0259] Furthermore, the policy name extraction means 42 uses the policy implementation result data stored in the questionnaire data storage means 51 (see FIG. 3) to extract policies using a large-scale language model (LLM), and stores the obtained policy name data in the policy extraction result storage means 60 (see FIG. 8) (step S7). The details of this process have already been described in detail in the explanation of the policy name extraction means 42, so a detailed explanation will be omitted here.

[0260] Next, the policy tag creation means 43 vectorizes the policy name data stored in the policy extraction result storage means 60 (see FIG. 8) to create a policy name vector, and stores the created policy name vector together with the corresponding policy name data in the policy name-policy tag correspondence storage means 61 (see FIG. 8). Furthermore, clustering is performed using all the obtained policy name vectors, and policy tags are created using a large-scale language model (LLM) using a set of policy name data belonging to the classified and formed policy name clusters, and are stored in the policy name-policy tag correspondence storage means 61 (see FIG. 8) and the policy extraction result storage means 60 (see FIG. 8) (step S8). The details of this process have already been described in detail in the explanation of the policy tag creation means 43, so a detailed explanation will be omitted here.

[0261] Furthermore, the press release acquisition means 39 uses the keywords stored in the keyword storage means 58 to acquire press release data and its title data and date data from the press release distribution system 80, and stores them in the press release storage means 59 (step S9). The details of this process have already been described in detail in the explanation of the press release acquisition means 39, so a detailed explanation will be omitted here.

[0262] Next, the title vectorization means 40 vectorizes the title data stored in the press release storage means 59 to create a title vector, and stores the created title vector in the press release storage means 59 (step S10). The details of this process have already been described in detail in the description of the title vectorization means 40, so a detailed description will be omitted here.

[0263] <Processing flow when health management scoring system 10 provides services to users: Figure 12>

[0264] 12, the health and productivity management scoring system 10 uses the output means 45 to display information about health and productivity management on the screen of the user terminal 91. The transitions of the screens displayed by the output means 45 are diverse, as shown in FIG. 13, so here we will explain the processing for displaying the main screens.

[0265] First, the output means 45 accepts input of the number of required measures k specified by the user (the company or its representative that will receive the recommended measures service) in the measure number input section 311 of the condition setting section 310 of the measure recommendation screen 300 of Figure 16, and also accepts input of the company (usually the company itself) specified by the user in the company name input section 321 of the company issue extraction display section 320 of the measure recommendation screen 300 of Figure 16 (step S21).

[0266] Next, the recommendation candidate set generation means 37 calculates the similarity (the similarity of the tasks, which is cosine similarity in this embodiment) between the task content vector of the company specified by the user, which is stored in the company response vector storage means 57 (see FIG. 7), and each of the task content vectors of the other multiple companies, sorts the company response identification information for the other multiple companies in descending order of the calculated task similarity, and selects the company response identification information in descending order of similarity until the number of policy implementation result clusters of the policy implementation result cluster numbers corresponding to the sorted company response identification information becomes the same as the number k of required measures specified by the user, thereby generating a recommendation candidate set consisting of policy implementation result data (stored in the survey form data storage means 51 (see FIG. 3)) belonging to the top k policy implementation result clusters. The details of this process have already been described in detail in the description of the recommendation candidate set generation means 37, so a detailed description will not be repeated here.

[0267] Next, the recommendation target policy selection means 38 selects, as a recommendation target policy, the policy implementation result data with the highest health management navigation score, overall evaluation, or other evaluation value from among the top k policy implementation result clusters constituting the recommendation candidate set generated by the recommendation candidate set generation means 37 (step S23). The details of this process have already been described in detail in the description of the recommendation target policy selection means 38, so a detailed description will be omitted here.

[0268] Furthermore, the press release selection means 41 calculates the similarity (the similarity between title and measure, which in this embodiment is cosine similarity) between each of the title vectors of the multiple press releases stored in the press release storage means 59 and the measure implementation result vector (stored in the company response vector storage means 57) for the measure implementation result data selected by the recommendation target measure selection means 38, and selects a predetermined number (p) of press releases or a number (p) specified by the user in descending order of the calculated similarity between title and measure (step S24). The details of this process have already been described in detail in the explanation of the press release selection means 41, so a detailed explanation will not be given here.

[0269] Then, the output means 45 displays on the screen in the measure recommendation result display section 330 of the measure recommendation screen 300 of Figure 17 the top k measure implementation result data selected as measures to be recommended by the measure selection means 38 to be recommended, as well as the measure tag corresponding to the measure implementation result data (stored in the measure extraction result storage means 60), and further displays on the screen information of the press release selected by the press release selection means 41 (title data, date data, and body data displayed after clicking the title) (step S25).

[0270] 17 and the measure tag list display section 380 of the measure recommendation screen 300 of FIG. 19 can be selected by clicking, and this selection operation is accepted by the output means 45. Then, the measure tag lookup / related measure search means 44 receives the measure tag designated by the user from the output means 45, extracts company response identification information associated with the received measure tag and stored in the measure extraction result storage means 60 (see FIG. 8) and / or the measure implementation result data associated with this company response identification information and / or the measure implementation result data stored in the questionnaire data storage means 51 (see FIG. 3), and selects a predetermined number (u items) or a user-specified number (u items) of company response identification information and / or the measure implementation result data in descending order of health and productivity management navigation score, overall evaluation, or other evaluation value (step S26).

[0271] Then, the output means 45 displays on the policy tagging related policy search result display screen 600 various data such as the company name (stored in the company information storage means 52) corresponding to the company response identification information extracted by the policy tagging related policy search means 44 and policy implementation result data (step S27).

[0272] <Effects of this embodiment>

[0273] This embodiment has the following effects: In other words, in the health and productivity management scoring system 10, the perspective evaluation request means 33 obtains, from the large-scale language model (LLM), perspective evaluation result data that indicates whether each of multiple (for example, nine in this embodiment) perspectives is satisfied using a binary value (1 or 0 in this embodiment), and further, the score calculation means 34 calculates a health and productivity management navigation score using TF-IDF using the perspective evaluation result data, and the output means 45 can display information related to health and productivity management on a screen using the health and productivity management navigation score.

[0274] Therefore, a Health and Productivity Management Navigation Score can be obtained by quantifying the evaluation of the free text data (task content data, measure implementation results data, and effectiveness verification results data) written by each company, and information related to health and productivity management can be displayed on the screen using this Health and Productivity Management Navigation Score. A correlation has been confirmed between the obtained Health and Productivity Management Navigation Score and the overall evaluation value calculated by the entity that evaluates the questionnaire data (in this case, the Ministry of Economy, Trade and Industry), confirming the usefulness of the Health and Productivity Management Navigation Score. When the Ministry of Economy, Trade and Industry's Health and Productivity Management Survey Form is set to nine perspectives and normalized by dividing by the maximum value, the correlation coefficient is approximately 0.39. The correlation can also be confirmed visually using the correlation display unit 510 in Figure 21, for example.

[0275] Furthermore, even if the evaluation result score is simply calculated as the sum or average value for each combined text data using the perspective evaluation result data (a binary value indicating whether each of the multiple perspectives is met or not) (a value indicating how many of the multiple perspectives are met, or a value proportional to that), a correlation with the overall evaluation value can be seen, but the evaluation result score is a numerical value that does not take into account the difficulty of achieving each of the multiple perspectives.In contrast, in this embodiment, the health management navigation score is calculated using the TF-IDF value, so it is a numerical value that takes the difficulty into account, making the evaluation result score even more useful.

[0276] By using the health and productivity management navigation score obtained in this way, measures that are thought to be effective can be recommended to users who wish to improve their health and productivity management evaluation.

[0277] Furthermore, in the health management scoring system 10, as shown in Fig. 4, the viewpoint extraction means 32 performs so-called pairwise evaluation to have the large-scale language model (LLM) extract viewpoints, and then, as shown in Fig. 5, the viewpoint evaluation request means 33 has the large-scale language model (LLM) evaluate the details of the efforts expressed in the combined text data from multiple viewpoints (9 viewpoints) set using the extracted viewpoints, so that the combined text data can be evaluated in accordance with the thinking characteristics (decision processing algorithm) of the large-scale language model (LLM). This can improve the reliability of the LLM response data.

[0278] Furthermore, the health and productivity management scoring system 10 includes a company response vectorization means 35, a policy implementation result clustering means 36, a recommendation candidate set generation means 37, and a recommendation target policy selection means 38, so that a recommendation candidate set is generated from the top k policy implementation result clusters, and the policy implementation result data with the highest health and productivity management navigation score in each policy implementation result cluster can be selected as the policy to be recommended. This makes it possible to recommend a variety of policy implementation result data.

[0279] That is, as shown in Figure 7, the policy implementation result clustering means 36 performs clustering using the policy implementation result vector to create policy implementation result clusters, and then, as shown in Figure 9, first, the recommendation candidate set generation means 37 sorts the company response identification information by task similarity using the task content vector (thus, this also sorts the policy implementation result data, policy implementation result vector, and policy implementation result cluster number corresponding to the company response identification information), then selects the top k policy implementation result cluster numbers (thus, the top k policy implementation result clusters) to generate a recommendation candidate set, and further, as shown in Figure 10, the recommendation target policy selection means 38 selects the policy implementation result data with the highest health management navigation score from each policy implementation result cluster as the policy to be recommended, and displays it on the screen using the output means 45, so that a variety of policy implementation result data can be recommended.

[0280] For this reason, simply sorting the company response identification information by the similarity of the issues and outputting the policy implementation result data corresponding to the top company response identification information may result in a large amount of similar policy implementation result data being output. However, by selecting the top k policy implementation result clusters and selecting and outputting the policy implementation result data with the highest health management navigation score within each policy implementation result cluster, it is possible to prevent a situation in which a large amount of similar policy implementation result data (i.e., policy implementation result data belonging to the same policy implementation result cluster) is output, and to output a diverse range of policy implementation result data.

[0281] 17, the health and productivity management scoring system 10 is equipped with a press release acquisition means 39, a title vectorization means 40, a press release selection means 41, and a press release storage means 59, and therefore, as shown in Fig. 17, the output means 45 can display on a screen the implementation result data of the measures that are the target of the recommendation, as well as the data of a predetermined number (p items) of press releases selected by the press release selection means 41 or a number (p items) designated by the user. This makes it possible to present more information on health and productivity management to the user.

[0282] Furthermore, since the health management scoring system 10 is equipped with a policy name extraction means 42 and a policy tag creation means 43, as shown in Fig. 8, since the number of policies (policy name data) extracted from the policy implementation result data using the large-scale language model (LLM) is too large, clustering is performed to create policy name clusters, and names are then assigned to each policy name cluster using the large-scale language model (LLM), and these names can be displayed on the screen as policy tags together with the policy implementation result data that is the policy to be recommended. Therefore, by looking at the policy tags, which have a smaller total number (for example, about 300) than the policy name data (for example, about 10,000), that is, by looking at the policy tags, which are a more cohesive unit than the policy name data, it is possible to understand the content of the policy implementation result data that is the policy to be recommended.

[0283] In addition, the health management scoring system 10 is equipped with a measure tag search means 44, so that as shown in Figure 22, it is possible to use measure tags to search for related measure implementation result data and display it on the screen.

[0284] <Transformation Form>

[0285] The present invention is not limited to the above-described embodiment, and modifications within the scope of the present invention are included in the present invention.

[0286] For example, in the above embodiment, the health management scoring system 10 is described as a server-client type system in which the scoring server 20, the administrator terminal 90, and the user terminal 91 are connected via a network 1, but this is not limited to this, and all or part of the system may be a standalone type system.

[0287] Specifically, among the various functions of scoring server 20, the functional portion relating to the preparatory processing before providing a service to a user (see Figure 11) is provided on administrator terminal 90, and among the various functions of scoring server 20, the functional portion relating to the processing when providing a service to a user (see Figure 12) is provided on user terminal 91, and data (health management navigation data, etc.) prepared in advance on administrator terminal 90 can be sent to user terminal 91 via network 1 or transferred to user terminal 91 using a recording medium such as a DVD or USB memory, and then the screen display processing of information related to health management can be performed by user terminal 91 alone (however, communication with external systems such as service provision system 70 using a large-scale language model (LLM) is performed).

[0288] Furthermore, only the functional parts of the various functions of scoring server 20 relating to the preparatory processing before providing the service to the user (see Figure 11) may be left on the server, and the functional parts of the various functions of scoring server 20 relating to the processing when providing the service to the user (see Figure 12) may be provided on user terminal 91, with the preparatory processing being performed by the server and administrator terminal 90, and the data obtained by the preparatory processing (health management navigation data, etc.) being sent to user terminal 91 via network 1 or transferred to user terminal 91 using a recording medium such as a DVD or USB memory, and then the screen display processing of information related to health management may be performed by user terminal 91 alone (however, communication with external systems such as service provision system 70 using a large-scale language model (LLM) is performed). [Industrial Applicability]

[0289] As described above, the human capital management scoring system and program of the present invention are suitable for use in analyzing data describing a company's health management initiatives and presenting measures that are thought to be effective in addressing the health management challenges each company faces, for example, with the aim of enabling employer companies of health insurance associations to obtain certification as outstanding health management corporations or to receive a health management brand recognition. [Explanation of symbols]

[0290] 10 Health and Productivity Management Scoring System 31 Coupling means 32 Perspective Extraction Method 33 Perspective evaluation request method 34 Score calculation method 35. Company response vectorization method 36 Clustering method for policy implementation results 37 Recommendation candidate set generation method 38 Method of selecting recommended measures 39 How to obtain press releases 40 Title Vectorization Method 41 Press Release Selection Method 42 Measure name extraction means 43 Measure tag creation method 44 Search method for policy tagging and related policies 45 Output Method 51 Survey data storage means 53. Storage means for viewpoint extraction results 54 Setting viewpoint storage means 55. Storage means for viewpoint evaluation results 56 Score storage means 57 Company response vector storage means 58 Keyword storage means 59 Press Release Storage Means 60 Measure extraction result storage means 61 Measure name / measure tag correspondence storage means

Claims

1. A computer-based human capital management scoring system that presents information on health management or other human capital management, a questionnaire data storage means for storing the health management survey data or other human capital management-related survey data obtained as responses from each company to a survey on health management or other human capital management, which are associated with company response identification information for identifying the company or, in cases where multiple responses from one company are permitted, for identifying the company and the response data, including issue content data consisting of text indicating the content of issues included in the health management survey data or other human capital management-related survey data, policy implementation result data consisting of text indicating the results of policy implementation, and effectiveness verification result data consisting of text indicating the results of effectiveness verification; a combining means for combining the problem content data, the policy implementation result data, and the effect verification result data stored in the questionnaire data storage means, or for combining the problem content data and the policy implementation result data to create combined text data; a perspective evaluation requesting means for generating perspective evaluation request data for each of the combined text data of each company, the perspective evaluation request data including a request for a binary answer as to whether the combined text data and the content of the efforts expressed in the combined text data satisfy each of the multiple perspectives, one for each of the combined text data of each company, in order to determine whether the content expressed in the combined text data of each company created by the combining means satisfies each of the multiple perspectives for evaluating each company's efforts related to human capital management, and inputting the generated perspective evaluation result data, which indicates whether each of the multiple perspectives is satisfied using one of the binary values ​​output from the large-scale language model, in association with company response identification information and stored in a perspective evaluation result storage means; score calculation means for calculating a term frequency-inverse document frequency (TF-IDF) by using two values ​​constituting the viewpoint evaluation result data for the combined text data of each company stored in the viewpoint evaluation result storage means, regarding each of the plurality of viewpoints as a word and regarding the combined text data of each company as a document, and for storing a total or average value obtained by summing or averaging the TF-IDF values ​​for each of the plurality of viewpoints obtained for each of the combined text data, or a value obtained by performing normalization processing or other processing processing on these total or average values, as a health management navi score or other human capital management navi score in association with company response identification information in a score storage means; an output means for executing at least one of the following processes as a process for displaying information related to the human capital management on a screen: a process for displaying on a screen at least one of the company name corresponding to company response identification information, the issue content data, the policy implementation result data, the effectiveness verification result data, and the combined text data, together with the human capital management navigation score stored in the score storage means; a process for displaying on a screen at least one of the company name corresponding to company response identification information, the issue content data, the policy implementation result data, the effectiveness verification result data, and the combined text data, according to the results of a selection, extraction, sorting, or grouping process using the human capital management navigation score; and a process for displaying on a screen the distribution of the human capital management navigation scores of each company or changes over time in the human capital management navigation score for each company; A human capital management scoring system comprising:

2. a viewpoint extraction means for preparing a large number of combinations of two pieces of combined text data selected from the combined text data of each company created by the combining means, creating one set for each of the large number of combinations, the two pieces of combined text data, and viewpoint extraction request data including a request for an answer as to which of the two pieces of combined text data is superior in terms of the initiative content expressed in the combined text data and why that conclusion has been reached, and inputting the created set of viewpoint extraction request data into the large-scale language model, and storing the text data output from the large-scale language model indicating why that conclusion has been reached in a viewpoint extraction result storage means as viewpoint data indicating from what perspective the comparison was made and superiority / inferiority was determined; The viewpoint evaluation request means As the plurality of viewpoints for evaluating each company's efforts regarding human capital management, a plurality of viewpoints set using the viewpoint data stored in the viewpoint extraction result storage means are used.

2. The human capital management scoring system of claim 1.

3. a company response vectorization means for vectorizing the problem content data and the policy implementation result data of each company stored in the questionnaire data storage means, and storing the obtained problem content vectors and policy implementation result vectors in a company response vector storage means in association with company response identification information; a policy implementation result clustering means for performing clustering using the policy implementation result vector of each company stored in the company response vector storage means, assigning a policy implementation result cluster number for identifying the classified policy implementation result cluster to the policy implementation result vector of each company, and storing the assigned policy implementation result cluster number in the company response vector storage means in association with company response identification information; a recommendation candidate set generation means for calculating a similarity between the issue content vector of the company specified by the user receiving the policy recommendation service and each of the issue content vectors of a plurality of other companies, sorting the company response identification information of the other plurality of companies in descending order of the calculated issue similarity, and selecting company response identification information in descending order of the issue similarity until the number of policy implementation result clusters of the policy implementation result cluster numbers corresponding to the sorted company response identification information becomes equal to the number k of required policies specified by the user or the predetermined number k of required policies, thereby generating a recommendation candidate set consisting of the policy implementation result data belonging to the top k policy implementation result clusters; a recommendation candidate set generation means for generating a recommendation candidate set and selecting, as a recommendation candidate set, the policy implementation result data having the highest human capital management navigation score from among the top k policy implementation result clusters selected by the recommendation candidate set generation means; The output means The system is configured to execute a process of displaying the policy implementation result data selected by the recommendation target policy selection means on a screen as a recommendation target policy.

2. The human capital management scoring system of claim 1.

4. a press release acquisition means for acquiring data of a plurality of press releases from a press release distribution system using keywords stored in the keyword storage means and storing the data in the press release storage means; a title vectorization means for vectorizing the title data of each of the press releases acquired by the press release acquisition means, and storing the obtained title vectors in the press release storage means in association with the acquired press releases; a press release selection means for calculating a similarity between each of the title vectors stored in the press release storage means and the policy implementation result vector stored in the company response vector storage means for the policy implementation result data selected by the recommendation target policy selection means, and selecting a predetermined number of press releases or a number of press releases designated by the user that have a high similarity between the calculated title and policy; The output means The system is configured to display on a screen the policy implementation result data of the policy to be recommended, and to execute a process of displaying on a screen the data of the press releases of a predetermined number selected by the press release selection means or a number specified by the user.

4. The human capital management scoring system of claim 3.

5. a policy name extraction means for creating policy extraction request data for each company, the policy implementation result data including a request to extract health management or other human capital management policies from the content expressed in the policy implementation result data using the policy implementation result data of each company stored in the questionnaire data storage means, and inputting the created policy extraction request data, the request requesting that the policy implementation result data and the policy implementation result data be extracted as policies related to health management or other human capital management policies, into the large-scale language model, and storing text data indicating the extracted one or more policies output from the large-scale language model as policy name data in a policy extraction result storage means in association with company response identification information; and a policy tag creation means for vectorizing each of the policy name data extracted by the policy name extraction means, and performing clustering using the obtained policy name vectors, and for each of the plurality of policy name clusters formed by classification, creating a set of the policy name data belonging to the same policy name cluster and policy tag naming request data including a request to name the set, and inputting the created set into the large-scale language model, and assigning text data indicating the name of the same policy name cluster output from the large-scale language model as a policy tag to the policy name data belonging to the same policy name cluster, and storing the assigned policy tag in the policy extraction result storage means in association with company response identification information. The output means The measure implementation result data that is the measure to be recommended is displayed on the screen, and the measure tag that is stored in the measure extraction result storage means in association with the company response identification information that corresponds to the measure implementation result data that is displayed on the screen is displayed on the screen.

4. The human capital management scoring system of claim 3.

6. a policy tag search means for searching for a related policy by using the policy tag designated by the user to extract company response identification information associated with the designated policy tag and stored in the policy extraction result storage means and / or the policy implementation result data associated with the company response identification information and stored in the survey data storage means, and selecting a predetermined number of company response identification information or a number of company response identification information designated by the user and / or the policy implementation result data from the extracted company response identification information and / or the policy implementation result data in descending order of the human capital management navigation score associated with the extracted company response identification information and stored in the score storage means; The output means The system is configured to display on the screen the company name and / or the policy implementation result data corresponding to the company response identification information selected by the policy tagging related policy search means.

6. The human capital management scoring system of claim 5.

7. A program for causing a computer to function as the human capital management scoring system according to any one of claims 1 to 6.

Images