Organizational Report Generation System

The organizational report generation system uses learning models to preprocess news information, determining relationships and sentiment, generating reliable reports with multiple perspectives by focusing on organizational context and analysis.

JP7588748B1Active Publication Date: 2024-11-22DENTSU INC +1

Patent Information

Application Number
JP2024094121
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2024-06-11
Publication Date
2024-11-22
Estimated Expiration
2044-06-11

AI Technical Summary

Technical Problem

Conventional systems using learning models like GPT struggle to generate reliable corporate reports that include multiple perspectives, often producing reports with untrue information.

Method used

An organizational report generation system that uses an existing learning model to input an organization's name, collects news information, preprocesses it to determine relationships and sentiment, and generates multiple report items based on industry and category analysis, including summaries and SWOT analyses.

Benefits of technology

The system generates reports with high reliability by using news information that is positively related to the organization, reflecting its vision, values, and environment, thereby enhancing the report's credibility.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007588748000001_ABST
    Figure 0007588748000001_ABST
Patent Text Reader

Abstract

An organizational report generation system is provided that can easily generate a corporate report including various perspectives by using an existing learning model and can increase the reliability of the corporate report. SOLUTION: When the name of a target company for which a report is to be generated is input, an organizational report generation system 1 collects news information about the target company, and uses an existing learning model to estimate from the collected news information whether or not there is a relationship between the news information and the target company, whether or not the news information is positive, and the industry and business category of the target company. Then, using an existing learning model, a plurality of report items are generated from the news information estimated to have a relationship with the target company, the estimation result of whether or not the news information is positive, and the estimation result of the industry and business category of the target company, and these are compiled to generate a report about the target company (corporate report).
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an organizational report generation system that generates a report on an organization using an existing learning model. [Background technology]

[0002] Conventionally, a system has been proposed that delivers a company report containing company information to a user via a communication line (see, for example, Patent Document 1). In the conventional system, a company ID set by a user is acquired and stored in the system, the company ID and company information are acquired and stored in a database, and trigger information including the company ID is acquired and stored in the database. Then, the company ID stored in the system for each user ID is acquired and compared with the trigger information stored in the database, company information is extracted from the database based on the compared company ID, and a company report is created for each user ID. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2002-324162 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the above conventional systems, there is no proposal for creating a corporate report using an existing learning model (e.g., GPT (Generative Pre-trained Transformer)), in particular for creating a corporate report that includes various perspectives (a corporate report that includes multiple report items). In particular, when creating a report using an existing learning model (e.g., GPT, etc.), there is a problem that the report generated will include false information, i.e., the reliability of the report will be low.

[0005] The present invention has been made in consideration of the above-mentioned problems, and aims to provide an organizational report generation system that can easily generate a report including various perspectives (a report regarding a target organization including multiple report items) using an existing learning model, while also increasing the reliability of the report. [Means for solving the problem]

[0006] The organizational report generation system of the present invention is an organizational report generation system that generates a report related to an organization, and the organizational report generation system includes an input unit into which an organization name of a target organization for which the report is to be generated is input, a news information collection unit that collects news information related to the target organization based on the organization name input to the input unit, a data preprocessing unit that uses an existing learning model generated by machine learning using predetermined learning data to estimate and output the presence or absence of a relationship between the news information and the target organization, whether the news information has positive or negative content, and the industry and industry category of the target organization identified from the news information, using an existing learning model, and a report item generation unit that uses as input news information estimated to have a relationship with the target organization, the estimation result of whether the news information has positive or negative content, and the estimation result of the industry and industry category of the target organization identified from the news information, and generates and outputs each of a plurality of report items that constitute a report on the target organization, and a report generation unit that compiles a plurality of report items output from the report item generation unit to generate a report on the target organization.

[0007] According to this configuration, when the organization name of the organization (target organization) for which a report is to be generated is input, a report including multiple report items related to the target organization is generated using an existing learning model (e.g., GPT, etc.) based on news information (news information related to the target organization) collected based on the input organization name. In this case, first, data preprocessing is performed using an existing learning model, and from the collected news information, the presence or absence of a relationship between the news information and the target organization, whether the news information is positive, and the industry and industry category of the target organization are estimated. Then, each of the multiple report items is generated using the existing learning model. At this time, since the news information estimated to have a relationship with the target organization is used, the reliability of the generated report items is high. In addition, multiple report items are generated from various perspectives using whether the news information is positive and the industry and industry category of the target organization. In this way, a report including various perspectives (a report related to the target organization including multiple report items) can be easily generated using an existing learning model, and the reliability of the report can be increased.

[0008] In addition, in the organizational report generation system of the present invention, the input unit receives homepage information of the target organization in addition to the organization name of the target organization, the news information collection unit collects news information about the target organization based on the homepage information of the target organization, the data pre-processing unit does not estimate whether the news information collected from the homepage of the target organization has positive or negative content, and the report item generation unit uses the existing learning model to input news information collected from the homepage of the target organization that is estimated to have a relationship with the target organization, and the estimated results of the industry and business category of the target organization identified from the news information collected from the homepage of the target organization, and generates and outputs a news summary reflecting the vision and values ​​of the target organization as one of the multiple report items.

[0009] According to this configuration, when the homepage information of the target organization is input in addition to the organization name of the target organization, a report on the target organization including multiple report items is generated using an existing learning model (e.g., GPT, etc.) based on news information (news information on the target organization) collected based on the input homepage information. In this case, in the data preprocessing using the existing learning model, it is not estimated whether the news information is positive or not. Then, when multiple report items are generated using the existing learning model, a news summary reflecting the vision and values ​​of the target organization is generated as one of the multiple report items based on the news information collected from the homepage of the target organization that is estimated to have a relationship with the target organization, and the estimated results of the industry and business category of the target organization identified from the news information collected from the homepage of the target organization. In this way, a report including the perspective (report item) of "news summary reflecting the vision and values ​​of the target organization" can be generated as a report on the target organization.

[0010] In addition, in the organizational report generation system of the present invention, the report item generation unit may use the existing learning model to input news information that is estimated to have a relationship with the target organization, an estimation result as to whether the news information has positive or negative content, and an estimation result of the industry and business category of the target organization identified from the news information, and generate and output a news summary about the environment in which the target organization is located as one of the multiple report items.

[0011] According to this configuration, when multiple report items are generated using an existing learning model (e.g., GPT, etc.), a news summary about the environment in which the target organization is located is generated as one of the multiple report items based on news information estimated to have a relationship with the target organization, an estimation result of whether the news information has a positive content, and an estimation result of the industry and business category of the target organization identified from the news information. In this way, a report including the perspective (report item) of "news summary about the environment in which the target organization is located" can be generated as a report on the target organization.

[0012] In addition, in the organizational report generation system of the present invention, the report item generation unit may use the existing learning model to input news information that is estimated to have a relationship with the target organization, the estimation result of whether the news information has positive or negative content, and the estimation result of the industry and business category of the target organization identified from the news information, and generate and output the analysis results of a SWOT analysis of the target organization as one of the multiple report items.

[0013] According to this configuration, when multiple report items are generated using an existing learning model (e.g., GPT, etc.), the analysis result of the SWOT analysis of the target organization is generated as one of the multiple report items based on the news information estimated to have a relationship with the target organization, the estimation result of whether the news information has a positive content, and the estimation result of the industry and business category of the target organization identified from the news information. In this way, a report including the perspective (report item) of "analysis result of SWOT analysis of the target organization" can be generated as a report on the target organization.

[0014] In addition, in the organizational report generation system of the present invention, the report item generation unit may use the existing learning model to input the analysis results of a SWOT analysis of the target organization, and generate and output a PR message created based on the analysis results of an SO analysis, ST analysis, or WO analysis of the target organization as one of the multiple report items.

[0015] According to this configuration, when multiple report items are generated using an existing learning model (e.g., GPT, etc.), a PR message that combines the SO analysis, ST analysis, and WO analysis obtained from the SWOT analysis of the target organization is generated as one of the multiple report items based on the analysis results of the SWOT analysis of the target organization. In this way, a report can be generated that includes the perspective (report item) of "a PR message created based on the analysis results of the SO analysis, ST analysis, or WO analysis of the target organization" as a report on the target organization.

[0016] The method of the present invention is a method executed by an organizational report generation system that generates a report on an organization, and the method includes an input step in which the name of a target organization for which the report is to be generated is input; a news information collection step in which news information on the target organization is collected based on the organization name and the organization name input in the input step; a data preprocessing step in which, using an existing learning model generated by machine learning using predetermined learning data, the news information collected in the news information collection step is used as input to estimate and output the presence or absence of a relationship between the news information and the target organization, whether the news information has positive or negative content, and the industry and industry category of the target organization identified from the news information; a report item generation step in which, using the existing learning model, the news information estimated to have a relationship with the target organization, the estimation result of whether the news information has positive or negative content, and the estimation result of the industry and industry category of the target organization identified from the news information are used as input to generate and output each of a plurality of report items constituting a report on the target organization; and a report generation step in which the plurality of report items output in the report item generation step are compiled to generate a report on the target organization.

[0017] In this method, as in the above system, when the name of an organization (target organization) for which a report is to be generated is input, a report including multiple report items is generated for the target organization using an existing learning model (e.g., GPT, etc.) based on news information (news information related to the target organization) collected based on the input organization name. In this case, data preprocessing is first performed using an existing learning model, and the presence or absence of a relationship between the news information and the target organization, whether the news information is positive, and the industry and industry category of the target organization are estimated from the collected news information. Then, each of the multiple report items is generated using an existing learning model. At this time, since news information estimated to have a relationship with the target organization is used, the reliability of the generated report items is high. In addition, multiple report items are generated from various perspectives using whether the news information is positive and the industry and industry category of the target organization. In this way, a report including various perspectives (a report related to the target organization including multiple report items) can be easily generated using an existing learning model, and the reliability of the report can be increased.

[0018] The program of the present invention is a program executed in an organizational report generation system that generates a report related to an organization, and the program causes the organizational report generation system to execute an input process in which the name of a target organization for which the report is to be generated is input; a news information collection process in which news information related to the target organization is collected based on the organization name input in the input process; a data preprocessing process in which, using an existing learning model generated by machine learning using predetermined learning data, the news information collected in the news information collection process is used as input to estimate and output the presence or absence of a relationship between the news information and the target organization, whether the news information has positive or negative content, and the industry and industry category of the target organization identified from the news information; a report item generation process in which, using the existing learning model, the news information estimated to have a relationship with the target organization, the estimation result of whether the news information has positive or negative content, and the estimation result of the industry and industry category of the target organization identified from the news information are used as input to generate and output each of a plurality of report items that constitute a report related to the target organization; and a report generation process in which the plurality of report items output in the report item generation process are compiled to generate a report related to the target organization.

[0019] As with the above system, this program also generates a report on the target organization including multiple report items using an existing learning model (e.g., GPT) based on news information (news information on the target organization) collected based on the input organization name. In this case, data preprocessing is first performed using an existing learning model, and the presence or absence of a relationship between the news information and the target organization, whether the news information is positive, and the industry and industry category of the target organization are estimated from the collected news information. Then, each of the multiple report items is generated using an existing learning model. At this time, since news information estimated to have a relationship with the target organization is used, the reliability of the generated report items is high. In addition, multiple report items are generated from various perspectives using whether the news information is positive and the industry and industry category of the target organization. In this way, a report including various perspectives (a report on the target organization including multiple report items) can be easily generated using an existing learning model, and the reliability of the report can be increased. Effect of the Invention

[0020] According to the present invention, a report including various perspectives can be easily generated using an existing learning model, and the reliability of the report can be improved. [Brief description of the drawings]

[0021] [Figure 1] 1 is a block diagram showing a configuration of an organizational report generating system according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram showing an example of news information collected in the embodiment of the present invention. [Diagram 3] FIG. 11 is a diagram showing an example of a result of data pre-processing according to the embodiment of the present invention. [Figure 4] FIG. 2 is a diagram showing an example of a news summary (company internal) generated in an embodiment of the present invention. [Diagram 5] FIG. 2 is a diagram showing an example of a news summary (outside a company) generated in an embodiment of the present invention. [Figure 6] FIG. 2 is a diagram showing an example of a result of a SWOT analysis according to an embodiment of the present invention. [Figure 7] FIG. 2 is a diagram showing an example of a PR message created from the analysis results of an SO analysis in an embodiment of the present invention. [Figure 8] 11 is a diagram showing an example of a PR message created from the analysis results of an ST analysis in the embodiment of the present invention. FIG. [Figure 9] FIG. 2 is a diagram showing an example of a PR message created from the analysis results of a WO analysis in an embodiment of the present invention. [Figure 10] FIG. 2 is a diagram showing an example of noteworthy news (corporate internal news) generated in an embodiment of the present invention. [Figure 11] FIG. 2 is a diagram showing an example of noteworthy news (outside a company) generated in an embodiment of the present invention. [Figure 12] FIG. 2 is a diagram showing an example of a mind map generated in an embodiment of the present invention. [Figure 13] FIG. 2 is a diagram showing an example of a time series graph generated in an embodiment of the present invention. [Figure 14] FIG. 4 is a sequence diagram illustrating the operation of the organizational report generating system according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0022] An organizational report generating system according to an embodiment of the present invention will be described below with reference to the drawings. In this embodiment, an organizational report generating system used for generating reports on companies (corporate reports) will be illustrated. The organizational report generating system according to this embodiment has a function of generating a corporate report using an existing learning model. This function is realized by a program stored in the memory area of ​​the organizational report generating system.

[0023] The configuration of an organizational report generating system according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the organizational report generating system according to the present embodiment. As shown in FIG. 1, an organizational report generating system 1 is connected to a user device 2 via a network N such as the Internet. In this embodiment, the organizational report generating system 1 is configured, for example, by a cloud server device. The user device 2 is configured, for example, by a business computer device, and includes an input unit 3 such as a keyboard and a mouse, and a display unit 4 such as a display.

[0024] The organizational report generation system 1 includes an input unit 5, a news information collection unit 6, a data preprocessing unit 7, a report item generation unit 8, a report generation unit 9, an output unit 10, and a storage unit 11. The input unit 5 receives the company name (the company name of the target company for which the report is to be generated, such as "Company A") inputted through the input unit 3 of the user device 2. In addition to the company name of the target company, homepage information of the target company (such as "URL of Company A's homepage") may also be inputted into the input unit 5. The period for which news information is to be collected (such as "past month" or "past year") may also be inputted into the input unit 5.

[0025] The news information gathering unit 6 has a function of collecting news information about the target company based on the company name input to the input unit 5. When the homepage information and period of the target company are input to the input unit 5 in addition to the company name of the target company, the news information gathering unit 6 can collect news information about the target company based on the homepage information and period. For example, the news information gathering unit 6 collects news information about the target company by crawling. In addition, the news information gathering unit 6 may collect news information about the target company using multiple crawlers (for example, a proprietary crawler that crawls highly reliable news sites and an external crawler provided by an external business operator).

[0026] Fig. 2 is a diagram showing an example of news information (main text) collected by the news information collecting unit 6. Fig. 2 shows an example in which a company name "Company A", "Company A's homepage information (URL)", and a period of "past one year" are input. Such various news information is collected by crawling. Note that publicly known technology can be used for crawling.

[0027] The data preprocessing unit 7 has a function of performing data preprocessing using an existing learning model. Specifically, the data preprocessing using the existing learning model is performed by inputting the news information collected by the news information collecting unit 6, estimating and outputting whether or not there is a relationship between the news information and the target company, whether the news information is positive or negative, and the industry and business category of the target company identified from the news information. In this case, the data preprocessing unit 7 does not estimate whether the news information collected from the homepage of the target company is positive or negative. The existing learning model is a learning model (such as "GPT") generated by machine learning using predetermined learning data. Any method such as deep learning using a neural network is used for machine learning.

[0028] In this embodiment, the data preprocessing unit 7 generates a prompt for data preprocessing that "infers, based on the input news information and the company name of the target company, whether the news information has a relationship with the target company, whether the news information has a positive or negative content, and the industry and business category of the target company identified from the news information," and inputs the prompt into an existing learning model. As a result, the data preprocessing unit 7 can obtain the result of data preprocessing as an output from the existing learning model.

[0029] FIG. 3 is a diagram showing an example of the result of data preprocessing (estimated result by an existing learning model). In the example of FIG. 3, the estimated result of the presence or absence of a relationship between the news information and the target company is written as "Yes" or "No" in the "Relationship" column. In addition, in the example of FIG. 3, the estimated result of whether the news information is positive or negative is written as "Positive" or "Negative" in the "Tone" column. Furthermore, in the example of FIG. 3, the estimated results of the industry and business category of the target company identified from the news information are written as "Investment" and "Financial performance" in the "Industry" and "Business" columns, respectively. As shown in FIG. 3, the result of data preprocessing may be given the "Title (title of the news information)", "URL (source of the news information)", "Classification (whether the source of the news information is internal or external to the company)", "Date (publication date of the news information)", etc. of the news information.

[0030] The report item generation unit 8 has a function of generating and outputting each of a plurality of report items constituting a report on a target company (company report) from the result of data preprocessing using an existing learning model (for example, "GPT"). Specifically, the process of generating and outputting each of a plurality of report items is executed using an existing learning model, using as input news information estimated to have a relationship with the target company, an estimation result as to whether the news information has a positive or negative content, and an estimation result of the industry and business category of the target company identified from the news information. The report item generation unit 8 can also perform a process of generating and outputting other report items using a known technology. Details of the functions of the report item generation unit 8 will be described later.

[0031] The report generation unit 9 has a function of aggregating multiple report items output from the report item generation unit 8 to generate a report on the target company (company report). When aggregating multiple report items, known techniques can be used. The output unit 10 has a function of outputting the company report generated by the report generation unit 9. The company report generated by the report generation unit 9 is sent from the output unit 10 to the user device 2 and displayed on the display unit 4 of the user device 2. The storage unit 11 is composed of a large-capacity memory, etc., and stores data and programs necessary for generating the company report.

[0032] Here, we will explain in detail the functions of the report item generation unit 8. The report item generation unit 8 includes, as functional blocks, a news summary generation unit 80, a SWOT analysis unit 81, a PR message generation unit 82, a featured news generation unit 83, a mind map generation unit 84, and a time series graph generation unit 85.

[0033] The news summary generation unit 80 has the function of using an existing learning model to input news information collected from the target company's homepage that is estimated to have a relationship with the target company, and the estimated results of the target company's industry and business category identified from the news information collected from the target company's homepage, and to generate and output a news summary (internal company) that reflects the target company's vision and values ​​as one of multiple report items.

[0034] In this embodiment, the news summary generating unit 80 generates a prompt for generating a news summary (inside the company) that "generates a news summary (inside the company) that reflects the vision and values ​​of the target company based on news information that is estimated to have a relationship with the target company and the estimated results of the industry and business category of the target company," and inputs the prompt into an existing learning model. As a result, the news summary generating unit 80 can obtain a "news summary (inside the company)" as shown in FIG. 4 as an output from the existing learning model.

[0035] In addition, the news summary generation unit 80 has the function of using an existing learning model to input news information that is estimated to have a relationship with the target company, the estimation result of whether the news information has positive or negative content, and the estimation result of the industry and business category of the target company identified from the news information, and generating and outputting a news summary (outside the company) about the environment in which the target company operates as one of multiple report items.

[0036] In this embodiment, the news summary generating unit 80 generates a prompt for generating a news summary (outside the company) that "generates a news summary (outside the company) about the environment in which the target company is located, based on news information that is estimated to have a relationship with the target company, the estimation result of whether the news information is positive, and the estimation result of the industry and business category of the target company," and inputs the prompt into an existing learning model. As a result, the news summary generating unit 80 can obtain a "news summary (outside the company)" as shown in FIG. 5 as an output from the existing learning model.

[0037] The SWOT analysis unit 81 has the function of using an existing learning model to input news information that is estimated to have a relationship with the target company, the estimated result of whether the news information has positive or negative content, and the estimated result of the industry and business category of the target company identified from the news information, and generates and outputs the analysis results of a SWOT analysis of the target company as one of multiple report items.

[0038] In this embodiment, the SWOT analysis unit 81 generates a prompt for SWOT analysis that "generates the analysis result of SWOT analysis of the target company based on the news information estimated to have a relationship with the target company, the estimation result of whether the news information has positive content, and the estimation result of the industry and business category of the target company," and inputs the prompt into the existing learning model. As a result, the SWOT analysis unit 81 can obtain the "analysis result of SWOT analysis" as shown in FIG. 6 as an output from the existing learning model.

[0039] The PR message generation unit 82 has the function of using an existing learning model to input the results of a SWOT analysis of the target company, and generating and outputting a PR message based on the results of either an SO analysis, ST analysis, or WO analysis of the target company as one of multiple report items.

[0040] In this embodiment, the PR message generating unit 82 generates a prompt for generating a PR message (SO analysis pattern) that "generates a PR message based on the analysis results of the target company's SO analysis, based on the analysis results of the target company's SWOT analysis," and inputs the prompt into an existing learning model. As a result, the PR message generating unit 82 can obtain a "PR message (SO analysis pattern)" as shown in FIG. 7 as an output from the existing learning model.

[0041] In addition, in this embodiment, the PR message generating unit 82 generates a prompt for generating a PR message (ST analysis pattern) that "generates a PR message based on the analysis results of the ST analysis of the target company based on the analysis results of the SWOT analysis of the target company" and inputs the prompt into the existing learning model. As a result, the PR message generating unit 82 can obtain a "PR message (ST analysis pattern)" as shown in FIG. 8 as an output from the existing learning model.

[0042] In this embodiment, the PR message generating unit 82 generates a prompt for generating a PR message (WO analysis pattern) that "generates a PR message based on the analysis results of the target company's WO analysis, based on the analysis results of the target company's SWOT analysis," and inputs the prompt into the existing learning model. As a result, the PR message generating unit 82 can obtain a "PR message (WO analysis pattern)" as shown in FIG. 9 as an output from the existing learning model.

[0043] The featured news generation unit 83 has a function of assigning weight values ​​to news information that has been subjected to data pre-processing by performing a predetermined weight calculation process, and generating and outputting news information with a high weight value as one of multiple report items as "noteworthy news." In the weight calculation process, a "quality score (0-100)" and a "quantity score (0-100)" are assigned to the news information. Then, news information with a high weight value (for example, the top five news information with the highest "quantity scores") is output as "noteworthy news." Note that the weight calculation process can use known technology.

[0044] In this embodiment, the featured news generation unit 83 can generate "noteworthy news (internal company)" as shown in Fig. 10 based on news information to which "internal company" has been assigned as a "classification" in data preprocessing. Also, the featured news generation unit 83 can generate "noteworthy news (external company)" as shown in Fig. 11 based on news information to which "external company" has been assigned as a "classification" in data preprocessing.

[0045] The mind map generating unit 84 has a function of generating a mind map for the target company based on the news information that has been subjected to data preprocessing. The mind map can be generated using known technology.

[0046] In this embodiment, the mind map generating unit 84 can generate a "mind map (company internal)" as shown in Fig. 12(a) based on news information to which "company internal" has been assigned as a "classification" in data preprocessing. Also, the mind map generating unit 84 can generate a "mind map (company external)" as shown in Fig. 12(b) based on news information to which "company external" has been assigned as a "classification" in data preprocessing.

[0047] The time series graph generating unit 85 has a function of generating a time series graph of news information related to a target company based on the news information that has been subjected to data preprocessing. The time series graph is, for example, a graph in which the horizontal axis represents the date of the news information and the vertical axis represents the number of news information (number of news items). A publicly known technology can be used to generate the time series graph.

[0048] In this embodiment, the time series graph generating unit 85 can generate a "time series graph (graph shown by white circles and a broken line in FIG. 13)" as shown in Fig. 13 based on news information to which "positive" has been assigned as the "tone" in data pre-processing. Also, the time series graph generating unit 85 can generate a "time series graph (graph shown by black circles and a broken line in FIG. 13)" as shown in Fig. 13 based on news information to which "negative" has been assigned as the "tone" in data pre-processing.

[0049] The operation of the organizational report generating system 1 configured as above will be described with reference to the sequence diagram of FIG.

[0050] When generating a company report using the organizational report generating system 1 of this embodiment, first, the user inputs, into the user device 2, the company name (e.g., "Company A") of the company for which the company report is to be generated, the homepage information of the company (e.g., "URL of Company A's homepage"), and the period for which news information is to be collected (e.g., "past year") (S1). The input company name, homepage information, and period information are transmitted from the user device 2 to the organizational report generating system 1 (S2).

[0051] In the organizational report generation system 1, when the company name, homepage information, and period information are input, news information (see FIG. 2) is collected based on the information (S3), and data preprocessing (see FIG. 3) is performed on the collected news information using an existing learning model (e.g., "GPT" etc.) (S4). Then, multiple report items are generated based on the results of the data preprocessing using the existing learning model (e.g., "GPT" etc.) (S5-S12).

[0052] Specifically, using an existing learning model, a "news summary (internal to the company)" (see Figure 4) reflecting the vision and values ​​of the target company is generated as one of multiple report items (S5). Also, using an existing learning model, a "news summary (external to the company)" (see Figure 5) about the environment in which the target company is located is generated as one of multiple report items (S6). Also, using an existing learning model, the "analysis results of SWOT analysis" (see Figure 6) of the target company is generated as one of multiple report items (S7). Also, using an existing learning model, a "PR message created based on the analysis results of SO analysis, ST analysis, and WO analysis of the target company" (see Figures 7 to 9) is generated as one of multiple report items (S8).

[0053] Furthermore, based on the news information to which "internal company" has been added in the data preprocessing, "notable news (internal company)" (see Figure 10) is generated as one of multiple report items (S9). Based on the news information to which "external company" has been added in the data preprocessing, "notable news (external company)" (see Figure 11) is generated as one of multiple report items (S10). Based on the news information to which the data preprocessing has been performed, a "mind map" (see Figure 12) regarding the target company is generated as one of multiple report items (S11). Based on the news information to which the data preprocessing has been performed, a "time series graph" (see Figure 13) of the news information regarding the target company is generated as one of multiple report items (S12).

[0054] Then, the multiple report items generated as described above (S5 to S12) are compiled to generate a report on the target company (company report) (S13). The generated company report is transmitted from the organizational report generation system 1 to the user device 2 (S14) and displayed on the display unit 4 of the user device 2 (S15).

[0055] According to the organizational report generation system 1 of this embodiment, when the company name of the company (target company) for which a report is to be generated is input, a report including multiple report items related to the target company is generated using an existing learning model (e.g., GPT, etc.) based on news information (news information related to the target company) collected based on the input company name (see FIG. 2). In this case, first, data preprocessing is performed using an existing learning model (see FIG. 3), and from the collected news information, it is estimated whether or not there is a relationship between the news information and the target company, whether or not the news information is positive, and the industry and business category of the target company. Then, each of the multiple report items is generated using an existing learning model (see FIG. 4 to FIG. 9).

[0056] At this time, news information that is estimated to have a relationship with the target company is used, so the reliability of the generated report items is high. In addition, multiple report items are generated from various perspectives using whether the news information is positive or not and the industry and business category of the target company. In this way, it is possible to easily generate a report that includes various perspectives (a report on the target company that includes multiple report items) using an existing learning model, and to increase the reliability of the report.

[0057] In addition, in this embodiment, when the homepage information of the target company is input in addition to the company name of the target company, a report on the target company including multiple report items is generated using an existing learning model (e.g., GPT) based on news information (news information related to the target company) collected based on the input homepage information.

[0058] In this case, data preprocessing using an existing learning model does not estimate whether the news information is positive or not. Then, when multiple report items are generated using an existing learning model, a news summary reflecting the vision and values ​​of the target company is generated as one of the multiple report items based on the news information collected from the target company's website that is estimated to have a relationship with the target company, and the estimated results of the target company's industry and business category identified from the news information collected from the target company's website. In this way, a report on the target company that includes a perspective (report item) of "news summary (company internal)" that reflects the target company's vision and values ​​can be generated (see Figure 4).

[0059] In addition, in this embodiment, when multiple report items are generated using an existing learning model (e.g., GPT, etc.), a news summary about the environment in which the target company is located is generated as one of the multiple report items based on the news information estimated to have a relationship with the target company, the estimation result of whether the news information is positive, and the estimation result of the industry and business category of the target company identified from the news information. In this way, a report including a perspective (report item) of "news summary (outside the company)" about the environment in which the target company is located can be generated as a report on the target company (see FIG. 5).

[0060] In addition, in this embodiment, when multiple report items are generated using an existing learning model (e.g., GPT, etc.), the analysis result of the SWOT analysis of the target company is generated as one of the multiple report items based on the news information estimated to have a relationship with the target company, the estimation result of whether the news information is positive, and the estimation result of the industry and business category of the target company identified from the news information. In this way, a report including the perspective (report item) of the "analysis result of SWOT analysis" of the target company can be generated as a report on the target company (see FIG. 6).

[0061] In addition, in this embodiment, when multiple report items are generated using an existing learning model (for example, GPT, etc.), a PR message including any one of SO analysis, ST analysis, or WO analysis obtained from the SWOT analysis of the target company is generated as one of the multiple report items based on the analysis result of the SWOT analysis of the target company. In this way, a report including the viewpoint (report item) of "PR message created based on any one of the analysis results of SO analysis, ST analysis, or WO analysis of the target company" can be generated as a report on the target company (see Figs. 7 to 9).

[0062] According to the organizational report generation system 1 of the present embodiment described above, news related to the target company can be obtained in real time from a reliable source, and various analyses related to corporate public relations can be performed. Therefore, detailed and specific original prompts based on the obtained news information make it possible to propose SWOT analysis and PR activities suitable for the company / organization.

[0063] In this embodiment, by providing a news information gathering unit 6 (crawl function), it is possible to gather all news related to a company within a specified period. The news information gathering unit 6 is capable of gathering press releases and news from company homepages (internal crawl function) and also capable of gathering global news that mentions the company (external crawl function).

[0064] In addition, in this embodiment, by providing a data preprocessing unit 6 (data preprocessing function), it is possible to present each news article to an existing learning model (such as "GPT"), cleanse the data, and obtain various outputs for generating a SWOT analysis. Here, the various outputs include "relationships," "tone," "industry and category classification," "hot news," and "news summary."

[0065] In this case, "Relevance" allows to determine if the news is relevant to the organization. "Tone" allows to analyze the sentiment of each news. "Industry and Category Classification" allows to identify if the retrieved news is in the relevant industry and category of the company. "Hot News" allows to evaluate all news with a unique score and select the most noteworthy ones for the company or organization. The unique score includes a Quality Score (a trusted news source score) and a Quantity Score (a score that indicates the influence of each news item on the target company, calculated from news distribution and industry and category data). "News Summary (Company Internal)" allows to create a summary of the organization's vision and values ​​using hot news and advanced prompts. "News Summary (Company External)" allows to generate a summary of external news related to the environment surrounding the company using information with high reliability.

[0066] In addition, in this embodiment, the mind map generation unit 84 (visualization function) can build a mind map from the company's related industries and categories. In addition, the time series graph generation unit 85 (visualization function) can visualize the relationship between the number of news items related to the company and time. In addition, the SWOT analysis unit 81 (SWOT analysis function based on categorized data) can perform a SWOT analysis of the company using categorized data to address the problem of token restrictions. Furthermore, the report generation unit 9 (report output function) can output various analysis results in a specified report layout.

[0067] Although the embodiments of the present invention have been described above by way of example, the scope of the present invention is not limited to these, and changes and modifications can be made according to the purpose within the scope of the claims.

[0068] For example, the above description has been given as an example of generating a report regarding a company, but the scope of the present invention is not limited to this, and reports can also be generated in a similar manner regarding organizations, unions, teams, and the like other than companies. [Industrial Applicability]

[0069] As described above, the organizational report generation system of the present invention has the effect of being able to easily generate reports including various perspectives using existing learning models and increasing the reliability of the reports, and is useful for generating reports on companies (corporate reports), etc. [Explanation of symbols]

[0070] 1 Organizational Report Generation System 2. User Device 3. Input section 4 Display 5 Input section 6. News Information Gathering Department 7 Data Preprocessing Section 8 Report Item Generation Section 9 Report Generation Section 10 Output section 11 Storage section 80 News Summary Generation Department 81 SWOT Analysis Department 82 PR Message Generation Unit 83 Featured News Generation Department 84 Mind Map Generator 85 Time series graph generator

Claims

1. An organizational report generation system that generates a report about an organization using an existing learning model generated by machine learning using predetermined learning data, comprising: The organizational report generation system includes: an input section for inputting the name of a target organization for which the report is to be generated; a news information gathering unit that gathers news information about the target organization based on the organization name inputted into the input unit; a data pre-processing unit that performs data pre-processing for generating the report using the existing learning model, the data pre-processing being a process of inputting news information collected by the news information collecting unit into the existing learning model, and estimating and outputting the presence or absence of a relationship between the news information and the target organization, whether the news information has positive or negative content, and the industry and business category of the target organization identified from the news information; a report item generation unit that performs a report item generation process using the existing learning model and a result of the data preprocessing, the report item generation process being a process of inputting news information that is estimated to have a relationship with the target organization, an estimation result as to whether the news information has positive or negative content, and an estimation result of the industry and business category of the target organization identified from the news information into the existing learning model, and generating and outputting each of a plurality of report items that constitute a report on the target organization; a report generation unit that compiles a plurality of report items output from the report item generation unit to generate a report regarding the target organization; An organizational report generation system comprising:

2. In addition to the organization name of the target organization, homepage information of the target organization is input to the input unit, The news information gathering unit gathers news information about the target organization based on homepage information of the target organization; The data preprocessing unit does not infer whether the news information collected from the homepage of the target organization is positive or negative, The report item generating unit includes: The organizational report generation system of claim 1, which uses the existing learning model to input news information collected from the target organization's homepage that is estimated to have a relationship with the target organization, and the estimated results of the target organization's industry and business category identified from the news information collected from the target organization's homepage, and generates and outputs a news summary reflecting the target organization's vision and values ​​as one of the multiple report items.

3. The report item generating unit includes: The organizational report generation system of claim 1, which uses the existing learning model to input news information estimated to have a relationship with the target organization, an estimation result as to whether the news information has positive or negative content, and an estimation result of the industry and business category of the target organization identified from the news information, and generates and outputs a news summary about the environment in which the target organization is located as one of the multiple report items.

4. The report item generating unit includes: The organizational report generation system of claim 1, which uses the existing learning model to input news information estimated to have a relationship with the target organization, an estimation result as to whether the news information has positive or negative content, and an estimation result of the industry and business category of the target organization identified from the news information, and generates and outputs the analysis results of a SWOT analysis of the target organization as one of the multiple report items.

5. The report item generating unit includes:

5. The organizational report generation system of claim 4, wherein the existing learning model is used to input the analysis results of a SWOT analysis of the target organization, and a PR message created based on the analysis results of an SO analysis, ST analysis, or WO analysis of the target organization is generated and output as one of the multiple report items.

6. A method executed in an organizational report generation system for generating a report on an organization using an existing learning model generated by machine learning using predetermined learning data, comprising: The method comprises: an input step for inputting the name of an organization for which the report is to be generated; a news information gathering step of gathering news information about the target organization based on the organization name and the organization name input in the input step; a data preprocessing step of performing data preprocessing for generating the report using the existing learning model, the data preprocessing being a process of inputting news information collected in the news information collection step into the existing learning model, and estimating and outputting the presence or absence of a relationship between the news information and the target organization, whether the news information has positive or negative content, and the industry and business category of the target organization identified from the news information; a report item generation unit that performs a report item generation process using the existing learning model and a result of the data preprocessing, the report item generation process being a process of inputting news information that is estimated to have a relationship with the target organization, an estimation result as to whether the news information has positive or negative content, and an estimation result of the industry and business category of the target organization identified from the news information into the existing learning model, and generating and outputting each of a plurality of report items that constitute a report on the target organization; a report generating step of generating a report on the target organization by compiling a plurality of report items output in the report item generating step; A method comprising:

7. A program executed in an organizational report generation system that generates a report on an organization using an existing learning model generated by machine learning using predetermined learning data, comprising: The program includes: An input process in which the name of the target organization for which the report is to be generated is input; a news information gathering process for gathering news information related to the target organization based on the organization name input in the input process; a data preprocessing step of performing data preprocessing for generating the report using the existing learning model, the data preprocessing step including inputting news information collected in the news information gathering step into the existing learning model, and estimating and outputting the presence or absence of a relationship between the news information and the target organization, whether the news information has positive or negative content, and the industry and business category of the target organization identified from the news information; a report item generation process using the existing learning model and utilizing a result of the data preprocessing, the report item generation process inputting, into the existing learning model, news information estimated to have a relationship with the target organization, an estimation result as to whether the news information has positive or negative content, and an estimation result of the industry and business category of the target organization identified from the news information, to generate and output each of a plurality of report items constituting a report on the target organization; a report generation process for generating a report on the target organization by collecting a plurality of report items outputted in the report item generation process; A program to execute.

Citation Information

Patent Citations

  • Information distribution system, storage medium and program thereof

    JP2002324162A

  • Latest popular information informing program, recording medium, device, and method

    JP2008234090A

  • Consulting support system, and consulting support method

    JP2021163122A

  • Evaluation support system and evaluation support program

    JP2023096804A

  • JPP7483170B

Cited By

  • system

    JP2026155688A

  • system

    JP2026155766A