Classification system and method for classifying classification object

Abstract

This classification system comprises an acquisition unit that acquires classification object information, a classification unit that classifies the classification object information in accordance with a classification criterion, and a display device. The classification unit includes: a language processing model (430); and a retrieval-augmented generation (RAG) data storage unit (420) in which a data source for causing the language processing model to execute language processing is stored on the basis of RAG. The RAG data storage unit (420) includes a classification reference data storage unit (422) in which a data source relating to the classification criterion is stored in a format conforming to the RAG. The data source relating to the classification criterion includes a plurality of pieces of classification information established according to the classification criterion. The classification unit determines classification information corresponding to the classification object information using the language processing model (430) and the data source relating to the classification criterion and assumes the configuration of a display device in a state in which the classification object information and the classification information are associated with one another.

Description

Classification systems and methods for classifying objects

[0001] The present disclosure relates to a classification system and a method for classifying objects.

[0002] Crowdsourcing of human resources is being utilized in companies, etc. In this type of crowdsourcing, recruiters who want to outsource work to outsiders need to find human resources with the most suitable capabilities for the work.

[0003] Patent document 1 describes a method for comparing information about the personnel required by a development project with information about company members to determine whether they match, and extracting members from the entire organization who match the required personnel as a search result.

[0004] Japanese Patent Application Laid-Open No. 2003-44642

[0005] In order to match business cases with human resources, it is conceivable to classify business cases using some kind of classification information. For example, in the case of technical business cases, classification information such as IPC classifications and F-terms related to patents may be used to classify business cases. By classifying business cases according to classification information and managing the business cases in association with the classification information, users can appropriately search for the business cases they need.

[0006] Recently, there has been a trend toward utilizing large-scale language models, which are rapidly advancing, for text analysis. Therefore, it is conceivable to apply large-scale language models to business case classification. However, large-scale language models have the characteristic that it is difficult to obtain accurate answers from them because they infer information during the process of generating answers. In particular, it is extremely difficult to process an appropriate correspondence between one of the vast number of business cases and one of the vast number of classification information. For this reason, there is a problem in that it is difficult to appropriately classify business cases simply by utilizing a general large-scale language model.

[0007] This problem does not only arise when the classification target is business cases. Recruiters may decide who to hire by referring to the resumes of applicants for business cases. Work ability information such as skills described in resumes is an important factor in deciding who to hire. There are a huge number of types of work ability information. Therefore, it is conceivable to apply a large-scale language model to classify such work ability information, just as in the case of business cases. Just as when the classification target is business cases, the above problem also arises when the classification target is work ability information.

[0008] The present disclosure has been made to solve the above-mentioned problems, and its purpose is to appropriately classify classification target information.

[0009] A classification system relating to a first aspect of the present disclosure is a classification system that classifies information to be classified that is used by a user, and includes an acquisition unit that acquires the information to be classified, a classification unit that classifies the information to be classified in accordance with classification criteria, and a display device, wherein the classification unit includes a language processing model and a RAG data storage unit that stores a data source for causing the language processing model to perform language processing based on RAG (Retrieval-Augmented Generation), the RAG data storage unit includes a classification criteria data storage unit that stores a data source related to the classification criteria in a format that complies with RAG, and the data source related to the classification criteria includes a plurality of classification information defined by the classification criteria, and the classification unit uses the language processing model and the data source related to the classification criteria to determine classification information corresponding to the information to be classified, and displays the information to be classified on the display device in a state where it is associated with the classification information.

[0010] A method according to a second aspect of the present disclosure is a method for classifying information to be classified that is used by a user, the method causing a computer to execute steps of acquiring the information to be classified and classifying the information to be classified in accordance with a classification criterion, the computer being configured to access a Retrieval-Augmented Generation (RAG) data storage unit, the RAG data storage unit including a classification criterion data storage unit in which a data source related to the classification criterion is stored in a format conforming to the RAG, the data source related to the classification criterion including a plurality of pieces of classification information defined by the classification criterion, the classifying step including steps of determining classification information corresponding to the information to be classified using a language processing model and the data source related to the classification criterion, and displaying the information to be classified on a display device in a state in which the information is associated with the classification information.

[0011] According to the present disclosure, it is possible to appropriately classify classification target information.

[0012] 1 is a block diagram showing the overall configuration of a classification system according to a first embodiment. FIG. 2 is a block diagram showing an overview of the configurations of a server and a user device. FIG. 3 is a block diagram explaining the configuration related to RAG (Retrieval-Augmented Generation) in the server. FIG. 4 is a diagram explaining the concepts of a classification method and a matching method by the classification system. FIG. 5 is a block diagram showing the configuration of a company database. FIG. 6 is a block diagram showing the configuration of a user database. FIG. 7 is a block diagram showing the configuration of a project database. FIG. 8 is a block diagram showing the configuration of a matching database. FIG. 9 is a block diagram showing the configurations of a classification unit, an AI construction unit, a data import unit, an acquisition unit, a matching unit, and a result display unit. FIG. 10 is a flowchart showing the processing procedure of the acquisition unit. FIG. 11 is a flowchart showing the processing procedure of the data import unit. FIG. 12 is a flowchart showing the processing procedure of the AI ​​construction unit. FIG. 13 is a flowchart showing the processing procedure of a chat unit (summary). FIG. 14 is a flowchart showing the processing procedure of a chat unit (classification). FIG. 15 is a flowchart showing the processing procedure of a classification storage unit. FIG. 16 is a flowchart showing the processing procedure of a matching unit. FIG. 17 is a diagram explaining the concepts of a classification method and a matching method according to a first modification. FIG. 18 is a block diagram showing the concepts of a classification method and a matching method according to a second modification. FIG. 1 is a diagram showing an example of a community database. FIG. 2 is a diagram showing an example of a case database. FIG. 3 is a diagram showing an example of a work history database. FIG. 4 is a diagram showing an example of a dialogue between a generation server and a user. FIG. 5 is a diagram showing an example of a dialogue between a generation server and a user (with dictionary registration). FIG. 6 is a diagram showing an example of a dialogue between a generation server and a user (with dictionary registration). FIG. 7 is a diagram for explaining the function of the generation server from the perspective of a divergence phase and a convergence phase. FIG. 8 is a diagram showing the functional configuration of the generation server.1 is a flowchart showing the processing steps of the data import unit. 2 is a flowchart showing the processing steps of the information acquisition unit. 3 is a flowchart showing the processing steps of the dictionary information registration unit. 4 is a flowchart showing the processing steps of the dictionary information display unit. 5 is a flowchart showing the processing steps of the information organization unit. 6 is a flowchart showing the processing steps of the resume registration unit. 7 is a timing chart showing the processing steps of the sharing server, recruiter device, applicant device, and generation server related to resumes. 8 is a diagram showing the functional configuration of a generation server relating to a modified example. 9 is a diagram showing another example of a dialogue conducted between the generation server and a user. 10 is a diagram showing another example of a dialogue conducted between the generation server and a user. 11 is a diagram showing the functional configuration of a generation server including an IPC classification unit. 12 is a flowchart showing the processing steps of the IPC classification unit. 13 is a diagram showing an example of a resume including an IPC. 14 is a diagram showing an example of a resume including an IPC. 15 is a diagram showing an example of a resume including an IPC. 16 is a diagram showing an example of a resume including an IPC. 17 is a diagram showing an example of a resume including an IPC. 18 is a diagram showing an example of an RAG technology being utilized in the functional configuration of the generation server.

[0013] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. While multiple embodiments will be described below, it is anticipated from the beginning that the configurations described in each embodiment will be appropriately combined. Note that identical or corresponding parts in the drawings will be designated by the same reference numerals, and their description will not be repeated.

[0014] Embodiment 1. Embodiment 1 will be described using Figures 1 to 19. In Embodiment 1, the classification system 1 will be described using business cases and profile information as examples of classification targets. Here, the profile information includes the name, age, career history, qualifications, skills, etc. of a user who is expected to perform work related to the business case.

[0015] As described below, the classification system 1 has a function of classifying business cases and profile information using classification information such as IPC. More specifically, the classification system 1 identifies (collects) classification information related to each of the business cases and the profile information. The classification system 1 stores the classification information corresponding to the business cases in a database in association with the business cases. The classification system 1 stores the classification information corresponding to the profile information in a database in association with the profile information. It is expected that the information stored in the database will be used, for example, for business matching.

[0016] 1 is a block diagram showing an overview of a classification system 1 according to a first embodiment. The classification system 1 is used, for example, in crowdsourcing between companies. Crowdsourcing is generally a process of soliciting contributions from an unspecified number of people to obtain needed services, ideas, or content.

[0017] The general configuration of the classification system 1 will be described with reference to Fig. 1. The classification system 1 comprises a server 400, recruiter devices 200A, 200B, 200C, . . . , and applicant devices 300A, 300B, 300C, .

[0018] Server 400 provides a matching service to many companies that matches business orders and receipts between companies. Figure 1 shows company A, company B, company C, etc. as examples of companies that use the matching service. Company A, company B, company C, etc. are registered as company members of classification system 1. Employees of company A, company B, company C, etc. who use classification system 1 are also individually registered as members of classification system 1.

[0019] Hereinafter, work for which contractors are being recruited in classification system 1 may be referred to as a "work case" or "recruited case," a person who provides a work case may be referred to as a "recruiter," and a person who applies to receive a work case may be referred to as an "applicant."

[0020] In classification system 1, for example, an applicant from company A can receive an order for work from company A. Therefore, classification system 1 also allows an applicant who belongs to a different department Y of company A to receive an order for work from department X of company A.

[0021] An applicant who accepts a job offer is referred to as the "contractor," and a recruiter who places an order with a contractor is referred to as the "orderer." However, in the following, the term "contractor" may be used to refer to the "applicant," and the term "orderer" may be used to refer to the "contractor."

[0022] A database 120 necessary for the matching service is constructed in the server 400. The database 120 includes various databases in which information necessary for providing the matching service is registered. In the database 120, information on business cases is registered in a state categorized by classification information.

[0023] In this disclosure, "registering" information in a database may be referred to as "storing" information in a database.

[0024] The recruiter device 200A is operated by an administrator of company A. The recruiter device 200B is operated by an administrator of company B. The recruiter device 200C is operated by an administrator of company C. Hereinafter, the recruiter devices 200A, 200B, 200C, etc. may be collectively referred to as "recruiter devices 200."

[0025] Applicant device 300A is operated by an applicant from company A. Applicant device 300B is operated by an applicant from company B. Applicant device 300C is operated by an applicant from company C. Hereinafter, applicant devices 300A, 300B, 300C, etc. may be collectively referred to as "applicant devices 300."

[0026] 1 shows two applicants for each company, but the number of applicants is not limited to this. There may be more applicants for each company, or a company may have only one applicant. The server 400 may also accept applicants who are not affiliated with a company, such as freelancers.

[0027] In the first embodiment, the managers of companies A, B, C, etc. are assumed to play the role of recruiters. Therefore, hereinafter, the managers of each company may be referred to as "recruiters." Recruiters can also act as applicants for jobs being recruited by other recruiters. In this case, the recruiter device 200 functions as the applicant device 300. In the first embodiment, when a company manager acts as a recruiter, the device that the manager uses to use the matching service is referred to as the recruiter device 200.

[0028] Company A may have one or more administrators. When company A has administrators, each administrator may be given a recruiter device 200, or one recruiter device 200 may be shared by multiple administrators. The same applies to companies B, C, etc.

[0029] The server 400 and the recruiter device 200 are configured to be able to communicate with each other via the Internet 50, which is an example of a communication network. The server 400 and the applicant device 300 are configured to be able to communicate with each other via the Internet 50.

[0030] The recruiter device 200 accepts various operations by the recruiter. For example, the recruiter device 200 accepts an operation to input a recruitment request (a requested task) and an operation to search for members of the matching service.

[0031] The recruiter device 200 communicates with the server 400 in response to each operation on the recruiter device 200. The server 400 registers a recruitment request (requested work) in the database 120 in response to an operation to input the recruitment request, and provides member information to the recruiter device 200 in response to an operation to search for a member.

[0032] The applicant device 300 accepts various operations by the applicant. For example, the applicant device 300 accepts an operation to search for a job posting, an operation to apply for a job posting, and the like.

[0033] The applicant device 300 communicates with the server 400 in response to each operation on the applicant device 300. The server 400 provides the applicant device 300 with appropriate job postings in response to an operation to search for job postings, and issues a notice of acceptance or rejection to the applicant device 300 in response to an operation to apply for the job posting.

[0034] Members who use classification system 1 access server 400 as recruiters or applicants. Hereinafter, members of classification system 1 may be referred to as "users." Also, below, recruiter device 200 and applicant device 300 may be collectively referred to as "user device 500."

[0035] The server 400 requests sign-in, which involves inputting a user ID and password, when accepting access from the user device 500. The server 400 identifies the user by the user ID notified at the time of sign-in.

[0036] The server 400 provides the user with an interface for interacting with AI (Artificial Intelligence). The user interacts with the AI ​​when classifying business cases using classification information. The AI ​​interacts with the business cases and determines classification information corresponding to the business cases. The server 400 registers the business cases in the database 120 in association with the classification information. The database 120 is an example of a business case storage unit.

[0037] The server 400 searches the database 120 for business cases that match the user's skills and provides them to the user. Fig. 1 shows an example in which a list of business cases that match the user's skills is displayed on a screen 550 of the user device 500. As shown in Fig. 1, the screen 550 may display the user's skills in addition to the list of business cases. In addition to business case information, profile information for identifying the user's skills is registered in the database 120 of the server 400.

[0038] As described above, the classification system 1 functions as a matching system that matches users with business cases.

[0039] 2 is a block diagram showing an outline of the configuration of the server 400 and the user device 500. A user accesses the server 400 using the user device 500. The user device 500 includes a processor 501, a memory 502, a communication interface 503, an input / output interface 504, a display 505, and an operation unit 506. The operation unit 506 includes a mouse, a keyboard, etc.

[0040] The memory 502 may include a random access memory (RAM), a read only memory (ROM), a flash memory, or any other suitable memory system. The memory 502 stores programs necessary for the arithmetic processing of the processor 501, temporary data calculated in the arithmetic processing, and the like.

[0041] The processor 501 connects to the Internet 50 via the communication interface 503 in accordance with a program stored in the memory 502. The processor 501 connects to the Internet 50 and communicates with the server 400.

[0042] The server 400 includes a microcomputer 410 and a storage 404. The microcomputer 410 is an example of a "computer." The storage 404 is configured with a hard disk drive, a solid state drive, etc. The storage 404 stores a large language model (LLM) 430 and a database 120.

[0043] The large-scale language model 430 is an example of a language processing model. The large-scale language model 430 is a language model (trained language model) that has been trained in advance by machine learning. A huge amount of training data is used to train the large-scale language model 430. The large-scale language model 430 is formed as an autoregressive model that uses a transformer such as a GPT (Generative Pre-trained Transformer). The large-scale language model 430 according to the first embodiment may be configured using a GPT (GPT-2, GPT-3, GPT-4), Bard, or the like.

[0044] The database 120 includes a company database 121, a user database 122, a job database 123, a matching database 124, and a RAG database 420. The company database 121 includes information on companies that use the classification system 1. The user database 122 includes information on users that use the classification system 1. The job database 123 includes information on business jobs. The matching database 124 includes information on business jobs that match users. The RAG database 420 includes information for improving the quality of answers generated by the large-scale language model 430.

[0045] The microcomputer 410 accesses the storage 404 and calls the large-scale language model 430. The microcomputer 410 accesses the storage 404 and executes a process of registering (storing) data in the database 120, a process of reading out the data registered in the database 120, and a process of updating the data registered in the database 120.

[0046] 3 is a block diagram for explaining the configuration related to RAG (Retrieval-Augmented Generation) in server 400. Here, the configuration related to RAG among the configurations in server 400 shown in FIG. 2 will be explained in detail.

[0047] The microcomputer 410 includes a processor 401, a memory 402, and a communication interface 403. The memory 402 includes a random access memory (RAM), a read only memory (ROM), a flash memory, or any other suitable memory system. The memory 402 stores programs necessary for the arithmetic processing of the processor 401, temporary data calculated in the arithmetic processing, and the like. The processor 401 operates based on the programs stored in the memory 402 and references the storage 404.

[0048] RAG is a technology that allows the large-scale language model 430 to generate answers based on a knowledge source containing the most recent and accurate information. Generally, large-scale language models may occasionally output inaccurate or misleading information, which is called hallucination. RAG can compensate for such imperfect behavior of the large-scale language model and improve the quality of the answers generated by the large-scale language model.

[0049] The RAG database 420 is, for example, a vector database. In addition to the data source of the classification information, the RAG database 420 defines instructions (behavior information) and embedded parameters (chunk size, etc.). The RAG database 420 is an example of a RAG data storage unit. The RAG database 420 includes multiple related information databases in which various data related to RAG is registered. FIG. 1 illustrates related information databases 421 to 423 as examples of the RAG database 420.

[0050] The related information databases 421 to 423 are examples of knowledge sources that contain the latest, accurate information. In the first embodiment, a method is proposed in which RAG technology is used to allow a large-scale language model 430 to appropriately classify various pieces of information.

[0051] In particular, in the first embodiment, the various information includes user profile information and job case information. The user profile information includes user skill information related to the job.

[0052] By viewing the contents of the job listings, users can determine whether they have the skills necessary to fill the job listings. However, it is time-consuming for users to view each of the numerous job listings and check whether they are suited to the job listings.

[0053] Therefore, in embodiment 1, we propose a classification system 1 that can classify profile information and business case information using classification information and present the most suitable business case to the user based on the classified profile information and business case information.

[0054] The related information database 421 has a large amount of business case information registered therein. The related information database 422 has a large amount of classification information registered therein for classifying technical information based on classification criteria. The related information database 423 has a large amount of profile information registered therein. The related information database 421 is an example of a business case data storage unit. The related information database 422 is an example of a classification criteria data storage unit. The related information database 423 is an example of a profile data storage unit. The data stored in the related information database 422 is an example of a data source related to the classification criteria. The related information databases 421 and 423 are examples of a classification target data storage unit.

[0055] 3 shows an example in which related information databases 421 to 423 are configured in storage 404. However, the storage in which related information database 421 is configured may be different from the storage in which related information database 422 is configured and the storage in which related information database 423 is configured. The data contained in each of related information databases 421 to 423 may be stored in table format rather than in database format. A single database may have a table area for storing data contained in related information database 421, a table area for storing data contained in related information database 422, and a table area for storing data contained in related information database 423.

[0056] As classification information, for example, the International Patent Classification (IPC) can be used. The IPC is an internationally standardized technology classification for classifying inventions for which patent applications have been filed. By using the IPC, technologies can be classified in detail by dividing them into hierarchical levels such as "section," "subsection," "class," "subclass," "main group," and "subgroup."

[0057] The IPC is originally an index used to classify patent documents. However, by using the IPC as an index to classify business case information and profile information, it is possible to classify business case information and profile information (especially skill information) using a uniform standard.

[0058] For example, suppose there is a business case such as "Research and development work for electric vehicle propulsion systems, electric vehicle propulsion devices, and electric vehicle battery charging means." Such a business case may be classified into "B06L (electric vehicle propulsion systems)," "B06L 50 / 00 (electric vehicle propulsion devices)," and "B06L 50 / 53 (electric vehicle battery charging means)."

[0059] As classification information, patent classifications such as F-terms, CPC, or FI may be used instead of IPC. Alternatively, the Japanese Decimal Classification (NDC) classification table for books and the International Decimal Classification may be used. Alternatively, if classification system 1 is used exclusively within a specific company, classification information used within the company may be used. Product classifications such as equipment may also be applied. In other words, classification information may be any of IPC, F-terms, CPC, FI, book classifications, and in-house classifications.

[0060] The microcomputer 410 accesses the RAG database 420 as needed, and provides information based on the RAG database 420 to the large-scale language model 430 in a dialogue between the user and the large-scale language model 430. As a result, the large-scale language model 430 responds to the user using accurate information based on the RAG database 420.

[0061] 1, an AI 450 is constructed by a microcomputer 410 and a storage 404. A user categorizes job case information and profile information through a chat dialogue with the AI ​​450, and searches for the job case information and profile information.

[0062] In particular, in the first embodiment, the task information and profile information are classified through a dialogue between the AI ​​450 and the user. A method for classifying the task information and profile information through a dialogue will be described below with reference to FIG. 4. Furthermore, a method for matching users with task information using the classification results will be described below.

[0063] 4 is a diagram for explaining the concepts of the classification method and matching method by classification system 1. Here, each method will be explained according to steps S100, S200, S300, and S400 shown in FIG.

[0064] (Step S100) Step S100 shows the flow of processing for constructing related information databases 421, 422 using business case information and classification information, and classifying business case information instructed by a user using the constructed related information databases 421, 422.

[0065] A user constructs the related information databases 421 and 422 in advance using a large amount of classification information and a large amount of business case information (see (i) in FIG. 4). The user who constructs the related information databases 421 and 422 is, for example, an administrator of the classification system 1. The administrator manages a large amount of business case information. The administrator constructs the related information database 421 using the business case information under his / her management.

[0066] A user who wishes to classify business case information interacts with the AI ​​450 using a chat screen provided by the user device 500. In particular, in the first embodiment, two types of interactions are assumed for classifying business case information. One is an interaction for summarizing the business case information (see (ii) in FIG. 4 ). The other is an interaction for determining classification information corresponding to the business case information based on the summarized business case information (see (iii) in FIG. 4 ).

[0067] The AI ​​450 summarizes the business case information specified by the user through dialogue with the user. When summarizing the business case information, the AI ​​450 refers to the related information database 421 as necessary. The AI ​​450 presents the summarized business case information to the user. The AI ​​450 re-summarizes the business case information in response to a user instruction. The AI ​​450 determines the summary content of the business case information in response to a user instruction.

[0068] Next, the AI ​​450 classifies the summarized business case information through interaction with the user. When classifying the business case information, the AI ​​450 references the related information database 422 as necessary. The AI ​​450 presents the user with classification information that is estimated to be appropriate, along with an overview of the technology encompassed by the classification information. The AI ​​450 re-estimates the classification information in response to a user instruction. The AI ​​450 determines the classification information corresponding to the summary of the business case information in response to a user instruction.

[0069] When the AI ​​450 determines appropriate classification information corresponding to the summarized business case information, it registers the business case information in association with the classification information in the business case database 123 (see (iv) in FIG. 4). As a result, the business case information is stored in the business case database 123 in a state where it is classified by the classification information.

[0070] (Step S200) Step S200 shows the flow of a process in which related information databases 422, 423 are constructed using profile information and classification information, and the constructed RAG database 420 is used to classify the profile information specified by the user. Step S200 is the same processing procedure as step S100, except that profile information is processed instead of business case information. Therefore, the same explanation as step S100 will not be repeated here. Note that the profile information includes the user's skill information. According to step S200, the profile information (especially skill information) is stored in the user database 122 in a state where it is classified by classification information.

[0071] (Step S300) Step S300 shows a process of searching for a job that matches the user's skills using classification information associated with the job information and classification information associated with the profile information (see (v) in FIG. 4). In step S300, so-called job matching is performed.

[0072] (Step 400) Step S400 shows the process of registering the results of business matching in the matching database 124 (see (vi) in FIG. 4). By step S400, business cases that are considered to be suitable for the user are registered in the matching database 124 for each user.

[0073] Note that uncategorized profile information may be stored in the user database 122. In this case, the AI ​​450 may call up the uncategorized profile information in response to a user specification, and carry out the two dialogues described above using the called-up profile information as the target. Alternatively, the AI ​​450 may accept profile information that is not stored in the user database 122, and carry out the two dialogues described above. The same applies to the case database 123. That is, uncategorized business case information may be stored in the case database 123. In this case, the AI ​​450 may call up business case information in response to a user specification, and carry out the two dialogues described above using the called-up business case information as the target. Alternatively, the AI ​​450 may accept business case information that is not stored in the case database 123, and carry out the two dialogues described above.

[0074] Here, an example has been described in which the related information databases 421 to 423 are constructed in the order of steps S100 and S200, but the related information databases 421 to 423 may be constructed in the order of steps S200 and S100.

[0075] Fig. 5 is a diagram showing an example of the company database 121. As shown in Fig. 5, the company database 121 registers company names, company addresses, etc. for each company ID.

[0076] FIG. 6 is a diagram showing an example of the user database 122. In the user database 122, user profile information is registered for each user ID. The profile information includes the ID of the company to which the user belongs, basic information, career information, and business ability information. The basic information includes the user's name, gender, age, and affiliation. The career information includes work history and educational background. The business ability information includes language ability, qualifications, skills, and the like. The server 400 registers the profile information in the user database 122 in association with classification information. The user database 122 is an example of a profile storage unit.

[0077] FIG. 7 is a diagram showing an example of the case database 123. In the case database 123, business case information is registered by case ID. The business case information includes the case title, case content, application requirements, recruitment start date, and recruitment end date. The application requirements include, for example, "Must Case," "Want Case," job rank, and age. The server 400 registers the business case information in the case database 123 in association with classification information. The case database 123 is an example of a business case storage unit.

[0078] FIG. 8 is a diagram showing an example of the matching database 124. Matching information is registered in the matching database 124 for each user ID. The matching information includes one or more case IDs. The case IDs included in the matching information indicate business cases that match the user. The server 400 uses the matching information to identify one or more business cases that match the user identified by the user ID.

[0079] 9 is a block diagram showing the configuration of the classification unit 800, AI construction unit 902, data import unit 903, acquisition unit 904, matching unit 905, and result display unit 906. The classification unit 800, AI construction unit 902, data import unit 903, acquisition unit 904, matching unit 905, and result display unit 906 are realized by the microcomputer 410 and data stored in the storage 404.

[0080] The classification unit 800 is an example of a classification unit that classifies the "business case information" acquired by the acquisition unit according to classification criteria. The classification unit 800 is also an example of a classification unit that classifies the "profile information" acquired by the acquisition unit according to classification criteria. The classification unit 800 includes a chat unit (summary) 803, a chat unit (classification) 804, and a classification storage unit 805.

[0081] Chat unit 803 includes chat unit 803A that processes business case information and chat unit 803B that processes profile information. Chat unit 804 includes chat unit 804A that processes business case information and chat unit 804B that processes profile information. Classification storage unit 805 includes classification storage unit 805A that processes business case information and classification storage unit 805B that processes profile information. The processing procedures of each unit will be described below with reference to Figures 10 to 17.

[0082] 10 is a flowchart showing the processing procedure of the acquisition unit 904. First, the acquisition unit 904 acquires business case information to be classified from a user and stores the acquired business case information in the case database 123 by user ID (step S1). At this stage, the business case information stored in the case database 123 is not associated with classification information.

[0083] Next, the acquisition unit 904 acquires profile information to be classified from the user and stores the acquired profile information by user ID in the user database 122 (step S2). At this stage, the profile information stored in the user database 122 is not associated with classification information.

[0084] The acquisition unit 904 may subject the business case information acquired from the user to classification processing by the classification unit 800 without storing it in the case database 123. Similarly, the acquisition unit 904 may subject the profile information acquired from the user to classification processing by the classification unit 800 without storing it in the user database 122. Step S1 is an example of an acquisition unit that acquires business case information. Step S2 is an example of an acquisition unit that acquires profile information. These acquisition units may be configured by the data import unit 903.

[0085] 11 is a flowchart showing the processing procedure of the data import unit 903. The data import unit 903 generates information sources required to construct the RAG database 420 according to the following procedure.

[0086] First, the data import unit 903 receives a large amount of business case information to be analyzed from a user such as an administrator (step S11). Next, the data import unit 903 saves the collected business case information in a specific area of ​​the storage 404 as an information source of the business case information (step S12).

[0087] Next, the data import unit 903 receives a large amount of profile information to be analyzed from a user such as an administrator (step S13). Next, the data import unit 903 saves the collected profile information in a specific area of ​​the storage 404 as an information source of the profile information (step S14).

[0088] Next, the data import unit 903 receives a large amount of classification information to be analyzed from a user such as an administrator (step S15). Next, the data import unit 903 saves the collected classification information in a specific area of ​​the storage 404 as an information source of the classification information (step S16).

[0089] 12 is a flowchart showing the processing steps of the AI ​​construction unit 902. First, the AI ​​construction unit 902 acquires an information source (business case information) from the storage 404 (step S21). Next, the AI ​​construction unit 902 uses the acquired information source (business case information) to construct an associated information database 421 related to the RAG (step S22). This generates the associated information database 421 in which data sources related to business case information are stored in a format conforming to the RAG.

[0090] Next, the AI ​​construction unit 902 acquires the information source (profile information) from the storage 404 (step S23). Next, the AI ​​construction unit 902 constructs the related information database 423 related to the RAG (step S24). This generates the related information database 423 in which the data source related to the profile information is stored in a format conforming to the RAG.

[0091] Next, the AI ​​construction unit 902 acquires the information source (classification information) from the storage 404 (step S25). Next, the AI ​​construction unit 902 constructs the related information database 422 related to the RAG (step S26). This generates the related information database 422 in which the data source related to the classification information is stored in a format conforming to the RAG.

[0092] 13 is a flowchart showing the processing steps of the chat unit (summary) 803. First, the chat unit 803 identifies the profile information or case information to be classified (step S31). Specifically, the chat unit 803 refers to the user database 122 to identify the profile information to be classified. Alternatively, the chat unit 803 refers to the case database 123 to identify the business case information to be classified.

[0093] The chat unit (summary) 803 may acquire profile information or business case information to be classified from the user through chat with the user.

[0094] Next, the chat unit 803 determines whether the classification target is profile information or business case information (step S32).

[0095] If the classification target is profile information, the chat unit 803 references the related information database 423 related to the profile information as necessary (step S33). Next, the chat unit 803 calls an AI algorithm (step S34). Next, the chat unit 803 summarizes the profile information through chat with the user while utilizing RAG (step S35). Next, the chat unit 803 passes the summary to the chat unit (classification) 804 (step S39).

[0096] Step S35 includes the steps of having the user confirm the summary of the profile information created by the chat unit 803, correcting the summary of the profile information in response to the user's instructions, and finalizing the summary of the profile information with the user's approval. That is, the chat unit 803 determines the content of the summary after interacting with the user regarding the summary of the profile information.

[0097] If the classification target is business case information, the chat unit 803 references the related information database 421 related to the business case information as necessary (step S36). Next, the chat unit 803 calls an AI algorithm (step S37). Next, the chat unit 803 summarizes the business case information through chat with the user while utilizing RAG (step S38). Next, the chat unit 803 passes the summary to the chat unit (classification) 804 (step S39).

[0098] Step S38 includes the steps of having the user confirm the summary of the business case information created by the chat unit 803, revising the summary of the business case information in response to the user's instructions, and finalizing the summary of the business case information with the user's approval. That is, the chat unit 803 determines the content of the summary after interacting with the user regarding the summary of the business case information. Note that the chat unit 803 may determine the content of the summary without obtaining explicit approval from the user.

[0099] 14 is a flowchart showing the processing procedure of the chat section (classification) 804. First, the chat section 804 determines whether the classification target is profile information or business case information (step S41).

[0100] If the classification target is profile information, the chat unit 804 refers to the related information database 422 related to the classification information as necessary (step S42). Next, the chat unit 804 calls an AI algorithm (step S43). Next, the chat unit 803 determines classification information corresponding to the profile information using a summary of the profile information while chatting with the user (step S44).

[0101] In this way, the chat unit (classification) 804 uses the large-scale language model 430 and the related information database 422 to determine category information corresponding to the profile information acquired by the acquisition unit. In particular, the chat unit (classification) 804 determines category information corresponding to the profile information based on a summary of the profile information. Next, the chat unit 804 displays the category information corresponding to the information to be classified (profile information) (step S48). More specifically, the chat unit 804 transmits data for displaying an image in which the profile information and the category information are associated to the user device 500. Based on the received data, the user device 500 displays the profile information on the display device in a state in which the profile information is associated with the category information. In this way, the chat unit 804 displays the profile information on the display device in a state in which the profile information is associated with the category information. Next, the chat unit 804 passes the determined content to the category storage unit 805 (step S49). The determined content includes the profile information and the category information corresponding to the profile information.

[0102] Step S44 includes a step of having the user confirm the classification information provisionally determined by the chat unit 804, a step of changing the classification information in response to an instruction from the user, and a step of finalizing the classification information with the user's approval. That is, the chat unit 804 determines the classification information corresponding to the profile information after having a conversation with the user regarding the classification information.

[0103] If the classification target is business case information, the chat unit 804 refers to the related information database 422 related to the classification information as necessary (step S45). Next, the chat unit 804 calls an AI algorithm (step S46). Next, the chat unit 803 determines classification information corresponding to the business case information using a summary of the business case information while chatting with the user (step S47).

[0104] In this way, the chat unit (classification) 804 uses the large-scale language model 430 and the related information database 422 to determine the classification information corresponding to the business case information acquired by the acquisition unit. In particular, the chat unit (classification) 804 determines the classification information corresponding to the business case information based on a summary of the business case information. Next, the chat unit 804 displays the classification information corresponding to the information to be categorized (business case information) (step S48). More specifically, the chat unit 804 transmits data for displaying an image in which the business case information and the classification information are associated to the user device 500. Based on the received data, the user device 500 displays the business case information associated with the classification information on the display device. In this way, the chat unit 804 displays the business case information associated with the classification information on the display device. Next, the chat unit 804 passes the determined content to the classification storage unit 805 (step S49). The determined content includes the business case information and the classification information corresponding to the business case information.

[0105] Step S47 includes a step of having the user confirm the classification information provisionally determined by the chat unit 804, a step of changing the classification information in response to a user instruction, and a step of finalizing the classification information with the user's approval. That is, the chat unit 804 determines the classification information corresponding to the profile information after having a dialogue with the user regarding the classification information. Note that the classification unit 800 may determine the classification information without having a dialogue with the user using the chat unit 804. For example, the classification unit 800 may collectively determine the classification information corresponding to each of a large number of pieces of profile information registered in advance, without having a dialogue with the user.

[0106] 15 is a flowchart showing the processing procedure of the category storage unit 805. First, the category storage unit 805 determines whether the storage target is profile information or business case information (step S51). If the storage target is profile information, the category storage unit 805 stores the profile information in the user database 122 in association with the category information (step S52). If the storage target is business case information, the category storage unit 805 stores the business case information in the case database 123 in association with the category information (step S53).

[0107] 16 is a flowchart showing the processing procedure of the matching unit 905. First, the matching unit 905 receives a matching instruction from a user (step S61). The user here is, for example, a user who is considering applying for a business case. The user instructs the matching unit 905 to perform matching using the user device 500.

[0108] Next, the matching unit 905 acquires profile information of the instructed user from the user database 122 (step S62). The profile information is associated with classification information. The matching unit 905 acquires business case information from the case database 123 (step S63). The business case information is associated with classification information.

[0109] Next, the matching unit 905 searches for business case information using the classification information associated with the profile information (step S64). More specifically, when the classification information associated with the profile information is classification information X, the matching unit 905 finds business case information associated with classification information X. The matching unit 905 estimates the found business case information as a matching result, and stores the matching result in the matching database 124 by user ID (step S65).

[0110] The matching unit 905 may execute step S63 before step S62. The matching unit 905 may accept a matching instruction on a web screen. The matching unit 905 may execute the process of deriving the matching result by batch processing.

[0111] 17 is a flowchart showing the processing procedure of the result display unit 906. First, the result display unit 906 receives an instruction from a user to display the matching results (step S71). Next, the result display unit 906 obtains the matching results corresponding to the user who received the instruction from the matching database 124 (step S72). The result display unit 906 uses the user ID to identify the matching results corresponding to the user who received the instruction.

[0112] Next, the result display unit 906 transmits the matching result to the user device 500 (step S73). Step S73 is an example of a transmission unit. The user device 500 displays the received matching result. As a result, the screen 550 shown in FIG. 1 is displayed on the user device 500. Note that the result display unit 906 may execute the processes of steps S72 and S73 even if it has not received a display instruction from the user (step S71).

[0113] [Modification 1] Next, Modification 1 will be described. Fig. 18 is a diagram for explaining the concepts of the classification method and matching method related to Modification 1. Modification 1 proposes a method of classifying advertisement information by classification information in addition to classifying job case information and profile information by classification information.

[0114] 18, in Modification 1, classification system 1 classifies profile information in step S200 and classifies advertising information in step S500. The procedure of step S200 has been described using FIG. 4, and therefore, the description thereof will not be repeated here. In Modification 1, classification system 1 also executes steps S100, S300, and S400 shown in FIG. 4, but the description thereof will not be repeated here.

[0115] In the first modification, advertisement database 125 is added to database 120 shown in FIG. 2, and related information database 424 is added to RAG database 420 shown in FIG.

[0116] (Step S500) Step S500 shows the flow of processing for constructing related information databases 422, 424 using advertisement information and classification information, and for classifying advertisement information designated by a user using the constructed related information databases 422, 424.

[0117] A user constructs related information databases 422, 424 in advance using a large amount of classification information and a large amount of advertising information (see (i) in FIG. 18). The user here is, for example, an administrator of the classification system 1. The advertising information includes, for example, various types of advertisements targeted at applicants. The advertising information may also be created targeting recruiters. Once the related information databases 422, 424 are constructed, preparations are made for classifying the advertising information.

[0118] A user who wishes to classify advertising information interacts with AI 450 using a chat screen provided by user device 500. In variant 1, two types of interactions are assumed for classifying advertising information. One is an interaction for summarizing advertising information (see (ii) in FIG. 18 ). The other is an interaction for determining classification information corresponding to advertising information based on the summarized advertising information (see (iii) in FIG. 18 ).

[0119] The AI ​​450 summarizes the advertising information specified by the user through dialogue with the user. When summarizing the advertising information, the AI ​​450 refers to the related information database 424 as necessary. The AI ​​450 presents the summarized advertising information to the user. The AI ​​450 re-summarizes the advertising information in response to a user instruction. The AI ​​450 determines the summary content of the advertising information in response to a user instruction.

[0120] Next, the AI ​​450 classifies the summarized advertising information through interaction with the user. When classifying the advertising information, the AI ​​450 refers to the related information database 422 as necessary. The AI ​​450 presents the user with classification information that is estimated to be appropriate, along with an overview of the technology encompassed by the classification information. The AI ​​450 re-estimates the classification information in response to a user instruction. The AI ​​450 determines classification information corresponding to the summary of the advertising information in response to a user instruction.

[0121] When AI 450 determines appropriate classification information corresponding to the summarized advertisement information, it registers the advertisement information in association with the classification information in advertisement database 125 (see (iv) in FIG. 18). As a result, advertisement information is stored in advertisement database 125 in a state where it is classified by the classification information. Advertisement database 125 is an example of an advertisement storage unit.

[0122] (Step S300A) Step S300A shows a process of searching for advertising information that matches the user's profile information using classification information associated with the advertising information and classification information associated with the profile information (see (v) in Figure 18).

[0123] (Step 400A) Step S400A shows a process of transmitting advertising information that matches the user's profile information to the user device 500 of the user (see (vi) in Figure 18). Step S400A causes an advertisement that matches the user's profile information to be displayed on the user device 500 of the user. Note that in step S400A, advertising information that matches the user's profile information may be registered in the matching database 124 for each user. Advertising information is an example of information to be classified. Similar to step S48 in Figure 14, the chat unit 804 may display the advertising information on the user device 500 in a state where it is associated with the classification information.

[0124] [Variation 2] Next, Variation 2 will be described. Fig. 19 is a diagram for explaining the concepts of the classification method and matching method related to Variation 2. Variation 2 proposes a method for classifying educational information by classification information in addition to classifying business case information and profile information by classification information. The educational information is, for example, educational videos for improving or broadening the user's skills.

[0125] As shown in Fig. 19, in Modification 2, classification system 1 classifies profile information in step S200 and classifies educational information in step S600. The procedure of step S200 has been described using Fig. 4, and therefore its description will not be repeated here. In Modification 2, classification system 1 also executes steps S100, S300, and S400 shown in Fig. 4, but the description thereof will not be repeated here.

[0126] In the second modification, an education database 126 is added to the database 120 shown in FIG. 2, and a related information database 425 is added to the RAG database 420 shown in FIG.

[0127] (Step S600) Step S600 shows the flow of processing for constructing the related information databases 422, 425 using the educational information and classification information, and classifying the educational information designated by the user using the constructed related information databases 422, 425.

[0128] The user constructs the related information databases 422, 425 in advance using a large amount of classification information and a large amount of educational information (see (i) in FIG. 19). The user here is, for example, an administrator of the classification system 1. The educational information includes, for example, various types of educational videos targeted at applicants. Once the related information databases 422, 425 are constructed, the educational information is ready to be classified.

[0129] A user who wishes to classify educational information interacts with the AI ​​450 using a chat screen provided by the user device 500. In the second modification, two types of interactions are assumed for classifying educational information. One is an interaction for summarizing the educational information (see (ii) in FIG. 19 ). The other is an interaction for determining classification information corresponding to the educational information based on the summarized educational information (see (iii) in FIG. 19 ).

[0130] The AI ​​450 summarizes the educational information specified by the user through dialogue with the user. When summarizing the educational information, the AI ​​450 refers to the related information database 425 as necessary. The AI ​​450 presents the summarized educational information to the user. The AI ​​450 re-summarizes the educational information in response to a user instruction. The AI ​​450 determines the summary content of the educational information in response to a user instruction.

[0131] Next, the AI ​​450 classifies the summarized educational information through dialogue with the user. When classifying the educational information, the AI ​​450 refers to the related information database 422 as necessary. The AI ​​450 presents the user with classification information that is estimated to be appropriate, along with an overview of the technology encompassed by the classification information. The AI ​​450 re-estimates the classification information in response to a user instruction. The AI ​​450 determines the classification information corresponding to the summary of the educational information in response to a user instruction.

[0132] When the AI ​​450 determines appropriate classification information corresponding to the summarized educational information, it registers the educational information in the educational database 126 in association with the classification information (see (iv) in FIG. 19). As a result, the educational information is stored in the educational database 126 in a state where it is classified by the classification information. The educational database 126 is an example of an educational storage unit.

[0133] (Step S300B) Step S300B shows a process of searching for educational information that matches the user's profile information using classification information associated with the educational information and classification information associated with the profile information (see (v) in Figure 19).

[0134] (Step 400B) Step S400B shows a process of transmitting educational information that matches the user's profile information to the user's user device 500 (see (vi) in Figure 19). Step S400B causes the educational information that matches the user's profile information to be displayed on the user's user device 500. Note that in step S400B, the educational information that matches the user's profile information may be registered in the matching database 124 for each user. Educational information is an example of information to be classified. Similar to step S48 in Figure 14, the chat unit 804 may display the educational information on the user device 500 in a state where it is associated with the classification information.

[0135] Here, an example is shown in which training information matching the user's profile information is provided to the user. However, to broaden the user's skill range, classification system 1 may search for training information stored in training database 126 and associated with classification information different from the classification information associated with the user's profile information. For example, if the user does not have the skills required for a business case in which the user is interested, classification system 1 may provide the user with training information for acquiring the skills required for the business case.

[0136] More specifically, when the classification information corresponding to the business case in which the user is interested differs from the classification information associated with the user's profile information, the classification system 1 may provide the user with an educational video classified by the classification information corresponding to the business case. Furthermore, the classification system 1 may perform the processing of the modification 2 in addition to the processing of the modification 1.

[0137] As described above, according to the first embodiment, it is possible to uniformly and appropriately classify business case information, profile information, advertising information, and educational information. Hereinafter, this information to be classified is referred to as "classification target information." Furthermore, according to the first embodiment, confidential information can also be classified by using RAG. Furthermore, by storing data related to the classification process in advance in the RAG database 420, the error rate of the large-scale language model 430 can be reduced. Note that, in this disclosure, "storage" includes the concept of continuously receiving and processing data. A transmitter that transmits data to a storage unit such as a memory may transmit all data to the storage unit in a lump. Alternatively, the transmitter may divide the data to be transmitted into packets or chunks and then transmit the divided data to the storage unit sequentially.

[0138] As described above, the classification system 1 determines the summary content of the information to be classified and the classification information corresponding to the information to be classified through a dialogue with the user. The dialogue between the classification system 1 and the user is realized by a chat function provided in the classification system 1. Such a chat function may not be necessary.

[0139] The classification system 1 may determine classification information corresponding to the information to be classified without going through a process of summarizing the information to be classified. When determining classification information corresponding to the information to be classified without going through a process of summarizing the information to be classified, the classification system 1 may determine the classification information through a dialogue with a user, or may determine the classification information without a dialogue with a user. Furthermore, in addition to RAG, Low-Rank Adaptation of Large Language Models (LoRA) or the like may be used.

[0140] 2, database 120 includes multiple databases. One or more of these multiple databases may be stored in a storage device separate from storage 404. Each of these multiple databases functions as a storage unit.

[0141] In the first embodiment, the classification unit 800 determines classification information corresponding to the information to be classified using the large-scale language model 430, related information databases 421, 423, 424, and 425 related to the information to be classified, and the related information database 422 in which data sources related to classification criteria are stored in a format conforming to RAG. However, the classification unit 800 may determine classification information corresponding to the information to be classified using the large-scale language model 430 and the related information database 422 without using the related information databases 421, 423, 424, and 425. In this case, the classification unit 800 may determine classification information corresponding to the information to be classified using, for example, the information to be classified obtained through a dialogue with the user, the large-scale language model 430, and the related information database 422. The information to be classified may include, for example, business case information, profile information, advertising information, and education information.

[0142] The user device 500 need not necessarily be equipped with all of the processor 501, memory 502, communication interface 503, and input / output interface 504, but may also be a thin client system using VDI (Virtual Desktop Infrastructure). A thin client system using VDI is a system in which a desktop environment on a server is transferred to a terminal in a remote location for use. The user device 500 and the server 400 do not necessarily have to be independent devices. When using such a thin client system, the functions of the user device 500 and the server 400 can be provided on the same aggregation server. Note that the classification system 1 may be configured without including the user device 500.

[0143] The server 400 is an example of a computing device. The computing device may be configured by a server (on-premise server, cloud server, etc.), a serverless system, etc. Here, an on-premise server is a server installed and managed in facilities managed within a company. A cloud server is a server (rented server) provided by another business via a network. A serverless system is a system in which computing and memory functions can be used only when necessary, without being aware of the existence of a server. Computing devices include servers and serverless systems. Servers include on-premise servers and cloud servers.

[0144] Second Embodiment Next, a second embodiment will be described with reference to FIGS. 20 to 50. In the second embodiment, the classification system 1000 will be described using, as an example, "work ability information" that is listed in a resume as a classification target. Here, the work ability information includes qualifications and skills acquired from past work experience. Career history, qualifications, and skills are examples of work ability information. The work ability information may be included in the profile information described as the first embodiment. The work ability information is also an example of "job-related information."

[0145] As will be described below as a "variant," the classification system 1000 has a function of classifying a user's work ability information using classification information such as IPC. More specifically, the classification system 1000 identifies (collects) classification information related to the user's work ability information and stores the identified classification information as part of the user's work history information. The work history information is expected to be used, for example, for work matching. Similar to the classification system 1, the classification system 1000 functions as a matching system that matches users with work cases.

[0146] [Overall Configuration] Fig. 20 is a block diagram showing an overview of a classification system 1000 according to the second embodiment. The classification system 1000 is used, for example, in crowdsourcing between companies. Crowdsourcing is generally a process of soliciting contributions from an unspecified number of people to obtain needed services, ideas, or content.

[0147] Many companies are encouraging employees to have side jobs in order to make effective use of their human resources. By using crowdsourcing between companies, the capabilities of employees can be utilized.

[0148] 20, the general configuration of the classification system 1000 will be described. The classification system 1000 includes a sharing server 100, recruiter devices 200A, 200B, 200C, . . ., applicant devices 300A, 300B, 300C, . . ., and a generating server 4000.

[0149] The sharing server 100 provides a matching service to many companies that matches business orders and receipts between companies. Figure 20 shows company A, company B, company C, etc. as examples of companies that use the matching service. Company A, company B, company C, etc. are registered as company members of the classification system 1000. Employees of company A, company B, company C, etc. who use the classification system 1000 are also individually registered as members of the classification system 1000.

[0150] The work arranged in classification system 1000 is, for example, temporary work that is expected to be completed within a predetermined period of time. Therefore, a person who accepts an order for work arranged in classification system 1000 will work in the specific department to which they belong within the company as their main job, and will engage in the work arranged in classification system 1000 as a side job, respectively. Note that in classification system 1000, for example, an applicant from company A can also accept an order for work from company A. Therefore, classification system 1000 also allows an applicant who belongs to a different department Y of company A to accept an order for work from department X of company A.

[0151] Hereinafter, a job for which contractors are being recruited in the classification system 1000 may be referred to as a "job being recruited" or a "job being recruited," a person who provides a job being recruited may be referred to as a "recruiter," and a person who applies to receive an order for a job being recruited may be referred to as an "applicant." Applying for a job being recruited may be referred to as an "application for the job being recruited" or an "application for the job being recruited."

[0152] An applicant who accepts a job offer is referred to as the "contractor," and a recruiter who places an order with a contractor is referred to as the "orderer." However, in the following, the term "contractor" may be used to refer to the "applicant," and the term "orderer" may be used to refer to the "contractor."

[0153] The sharing server 100 has a database 120 necessary for the matching service. The database 120 includes various databases in which information necessary for providing the matching service is registered. For example, the database 120 registers information about members and recruitment operations. The sharing server 100 is managed and operated by a company separate from the companies that use the matching service. Any of the companies that use the matching service may manage and operate the sharing server 100.

[0154] The recruiter device 200A is operated by an administrator of company A. The recruiter device 200B is operated by an administrator of company B. The recruiter device 200C is operated by an administrator of company C. Hereinafter, the recruiter devices 200A, 200B, 200C, etc. may be collectively referred to as "recruiter devices 200."

[0155] Applicant device 300A is operated by an applicant from company A. Applicant device 300B is operated by an applicant from company B. Applicant device 300C is operated by an applicant from company C. Hereinafter, applicant devices 300A, 300B, 300C, etc. may be collectively referred to as "applicant devices 300." Although FIG. 20 depicts two applicants for each company, the number of applicants is not limited to this. There may be more applicants for each company, or a company may have only one applicant. Sharing server 100 may also accept applicants who do not belong to a company, such as freelancers.

[0156] In the second embodiment, the managers of companies A, B, C, etc. will act as recruiters. Therefore, hereinafter, the managers of each company will be referred to as "recruiters." Recruiters can also act as applicants for jobs that other recruiters are recruiting for. In this case, the recruiter device 200 functions as the applicant device 300. In the second embodiment, when a company manager acts as a recruiter, the device that the manager uses to use the matching service will be referred to as the recruiter device 200.

[0157] Company A may have one or more administrators. When company A has administrators, each administrator may be given a recruiter device 200, or one recruiter device 200 may be shared by multiple administrators. The same applies to companies B, C, etc.

[0158] The sharing server 100 and the recruiter device 200 are configured to be able to communicate with each other via the Internet 50, which is an example of a communication network. The sharing server 100 and the applicant device 300 are configured to be able to communicate with each other via the Internet 50.

[0159] The sharing server 100 requires sign-in with input of a member ID and password when accepting access from the recruiter device 200. Similarly, the sharing server 100 requires sign-in with input of a member ID and password when accepting access from the applicant device 300. The sharing server 100 identifies each recruiter and applicant by the member ID notified at the time of sign-in.

[0160] The recruiter device 200 accepts various operations by the recruiter. For example, the recruiter device 200 accepts an operation to input a recruitment case (requested work), an operation to input an evaluation of a contractor who has completed the work, an operation to search for a member of the matching service, and the like.

[0161] The recruiter device 200 communicates with the sharing server 100 in response to each operation on the recruiter device 200. The sharing server 100 registers a recruitment request (requested work) in the database 120 in response to an operation to input the recruitment request, registers an evaluation of the target applicant (contractor) in the database 120 in response to an operation to input an evaluation, and provides member information to the recruiter device 200 in response to an operation to search for members.

[0162] The applicant device 300 accepts various operations by the applicant, such as an operation to search for a job offer, an operation to apply for a job offer, an operation to input work results, and an operation to input an evaluation of the recruiter (orderer).

[0163] The applicant device 300 communicates with the sharing server 100 in response to each operation on the applicant device 300. The sharing server 100 provides the applicant device 300 with appropriate job postings in response to an operation to search for job postings, issues a notice of acceptance or rejection to the applicant device 300 in response to an operation to apply for a job posting, registers the job postings in the database 120 in response to an operation to input job postings, and registers an evaluation of the target recruiter (orderer) in the database 120 in response to an operation to input an evaluation.

[0164] A recruiter belonging to a certain department of company A can use classification system 1000 to hire an applicant belonging to another department of company A as a contractor for a job. A recruiter belonging to company A can use classification system 1000 to hire an applicant belonging to company B as a contractor for a job.

[0165] Members who use the classification system 1000 access the sharing server 100 as recruiters or applicants. Hereinafter, members of the classification system 1000 may be referred to as "users." Also, below, the recruiter device 200 and applicant device 300 operated by members may be collectively referred to as "user device 500."

[0166] The sharing server 100 is communicably connected to the generation server 4000. The generation server 4000 provides the user with an interface to assist in creating a resume. The generation server 4000 has built therein a database 4420 necessary for providing such an interface to the user. The generation server 4000 may be managed by the company that manages the sharing server 100, or may be managed by a company different from the company that manages the sharing server 100. The sharing server 100 may include the functions of the generation server 4000.

[0167] The generation server 4000 is communicably connected to a user device 500 via the Internet 50. The user device 500 includes a recruiter device 200 and an applicant device 300. A user accesses the generation server 4000 using the user device 500.

[0168] Similar to the sharing server 100, the generation server 4000 requests sign-in involving input of a member ID and password when accepting access from the user device 500. The generation server 4000 identifies the user by the member ID notified at the time of sign-in.

[0169] The generation server 4000 provides the user with an interface for assisting in the creation of a career history. As a result, a career history creation tool is displayed on the screen 1550 of the user device 500. The generation server 4000 is an example of a career history information collection device.

[0170] The generation server 4000 has the function of collecting all the work history information necessary for creating a resume from the user while interacting with the user by displaying a question box 1551 and an answer box 1552 on the screen 1550. The generation server 4000 creates a resume in a standardized format based on the collected work history information. The generation server 4000 displays the created resume on the screen 1550, giving the user an opportunity to check the resume. The generation server 4000 stores the resume after the user has checked it in the generation server 4000. When a user applies for a job posting, the generation server 4000 cooperates with the sharing server 100 and transmits the resume to the recruiter device 200.

[0171] The background to the introduction of the generation server 4000 into the classification system 1000 is the "complexity of creating a curriculum vitae" and the "difficulty of standardizing curriculum vitae." Each of these will be described in detail below.

[0172] [The complexity of creating a resume] In order to efficiently utilize in-house human resources, it is effective to clarify each individual's skills and work experience. To clarify each individual's skills and work experience, it is necessary to collect as much personal information as possible without leaving anything out and compile it into a standard document format.

[0173] However, recalling skills and work experience and then compiling them into a standardized document format can be a very cumbersome process for job seekers.

[0174] Of course, those who wish to officially change jobs will be able to maintain a certain level of motivation for the tedious process described above, as they will have to prepare for rigorous document screening and interviews. On the other hand, employees who apply for internal projects or wish to take on side jobs outside the company will find it difficult to maintain motivation for the tedious process described above. This is because, in addition to not involving a formal job change, there is no requirement for rigorous document screening and multifaceted interviews.

[0175] In this way, in so-called "closed circle" cases, applicants have little motivation to go through the troublesome process described above. In such cases, there is a problem in that job seekers cannot collect all the individual skills and work experience and compile the collected information into a standard document format.

[0176] To create a high-quality resume, more information is needed regarding an individual's skills and experience. However, in private cases, even if a lot of information is available, many job seekers are not motivated to write a resume and find it troublesome (cumbersome) to write about their skills and experience. In such cases, the amount of information included in the resume tends to be small. Therefore, even if a job seeker has a lot of information available, they are unable to organize that information into a standardized document format, which presents a challenge.

[0177] [Difficulty in Standardizing Resumes] Consistent accuracy of resumes makes it easier for evaluators to evaluate each employee's work history. Furthermore, consistent accuracy of resumes allows companies to group employees' work histories for effective talent management. However, because resumes are typically written in free format, resume styles vary depending on the creator. Even if creators are required to write in a standard format, it is difficult to standardize the accuracy of resumes due to differences in individual document writing ability. There is also a risk that duplicate entries may be mixed into a resume due to the creator's carelessness. As a result, evaluators spend a lot of time evaluating each employee's work history based on their resumes. Furthermore, it is difficult for companies to group employees' work histories based on their resumes.

[0178] Based on the above-described background, the generation server 4000 is introduced into the classification system 1000. The generation server 4000 collects all work history information from the user in an interactive manner and creates a resume in a standardized format. The created resume contains all work history information related to required information without omission or duplication.

[0179] According to the second embodiment, it is possible to support a user in creating a resume that includes all the necessary information and does not contain any duplicate information. In addition, according to the second embodiment, it is possible to support a user in creating a resume that properly reflects the user's work history, regardless of the user's document creation ability. According to the second embodiment, it is possible to collect work history information that includes all the necessary information and does not contain any duplicate information.

[0180] FIG. 21 is a block diagram showing the configurations of the sharing server 100, the recruiter device 200, and the applicant device 300.

[0181] [Configuration of Sharing Server 100 ] The sharing server 100 includes a processor 101 , a memory 102 , a storage 103 , and a communication interface 104 .

[0182] The memory 102 may include a random access memory (RAM), a read only memory (ROM), a flash memory, or any other suitable memory system. The memory 102 stores programs required for the arithmetic processing of the processor 101, temporary data calculated in the arithmetic processing, and the like.

[0183] The storage 103 is configured with a hard disk drive, a solid state drive, etc. A database 120 is stored in the storage 103. The database 120 includes multiple types of databases. The multiple types of databases include a company database (company DB) 121, a user database (user DB) 122, a community database (community DB) 127, and a project database (project DB) 123.

[0184] Some of these multiple types of databases may be stored in storage provided separately from the sharing server 100. For example, the sharing server 100 may be connected to a cloud service separate from the sharing server 100, and some of the multiple types of databases shown in FIG. 21 may be stored on that cloud. In this case, the sharing server 100 can access the necessary databases by communicating with that cloud via the Internet 50.

[0185] Processor 101 connects to Internet 50 via communication interface 104 in accordance with a program stored in memory 102. Processor 101 connects to Internet 50 and communicates with recruiter device 200 and applicant device 300. Processor 101 accesses database 120 and executes processes such as extracting necessary data, registering new data in database 120, and updating data registered in database 120.

[0186] [Configuration of Recruiter Device 200] The recruiter device 200 includes a processor 201, a memory 202, a communication interface 203, an input / output interface 204, a display 205, and an operation unit 206. The operation unit 206 includes a mouse, a keyboard, and the like.

[0187] The memory 202 may include a random access memory (RAM), a read only memory (ROM), a flash memory, or any other suitable memory system. The memory 202 stores programs required for the arithmetic processing of the processor 201, temporary data calculated in the arithmetic processing, and the like.

[0188] The processor 201 connects to the Internet 50 via the communication interface 203 in accordance with a program stored in the memory 202. The processor 201 connects to the Internet 50 and communicates with the sharing server 100. The processor 201 communicates with the sharing server 100 and executes processes such as sending a job offer, displaying information about applicant members on the display 205, placing an order for work with a contractor selected from the applicants, and sending the contents of an evaluation of the contractor entered by the recruiter to the sharing server 100.

[0189] Information input by operating the operation unit 206 is notified to the processor 201 via the input / output interface 204 .

[0190] [Configuration of Applicant Device 300] Applicant device 300 comprises a processor 301, memory 302, a communication interface 303, an input / output interface 304, a display 305, and an operation unit 306. Operation unit 306 is made up of a mouse, a keyboard, and the like.

[0191] The memory 302 may include a random access memory (RAM), a read only memory (ROM), a flash memory, or any other suitable memory system. The memory 302 stores programs required for the arithmetic processing of the processor 301, temporary data calculated in the arithmetic processing, and the like.

[0192] The processor 301 connects to the Internet 50 via the communication interface 303 in accordance with a program stored in the memory 302. The processor 301 connects to the Internet 50 and communicates with the sharing server 100. The processor 301 communicates with the sharing server 100 and executes processes such as applying for a job offer, displaying on the display 305 a notice of acceptance or rejection of the job offer, transmitting the performance of the work order to the sharing server 100, and transmitting the contents of the evaluation of the recruiter entered by the applicant to the sharing server 100.

[0193] Information input by operating the operation unit 306 is notified to the processor 301 via the input / output interface 304 .

[0194] [Database 120] The following describes database 120. Company database 121 stores information about companies that are members of classification system 1000. User database 122 stores information about members who use classification system 1000. Many members are employees of companies that are members of classification system 1000.

[0195] Members registered in the user database 122 can act as recruiters (orderers) or applicants (recipients) by using the classification system 1000. Members may include employees of companies registered in the company database 121 as well as individuals (freelancers) who are not affiliated with any company.

[0196] The community database 127 stores information for identifying companies that belong to a community. A community is formed by agreement between companies. Therefore, multiple communities can be formed depending on how the companies agree. The number of companies that belong to one community can also be set in various ways. Companies that have a community relationship form a relationship of trust within the scope determined by how they agree to form the community. The community database 127 registers information for each community that identifies companies that belong to the community.

[0197] The job database 123 registers jobs (jobs for which contractors are being recruited). Employees of each company can work in their own departments while also accepting orders for jobs from other departments of their own company or jobs from other companies as members of the classification system 1000. In this case, members accept orders for jobs from other departments of their own company or jobs from other companies as a side job.

[0198] 22 is a block diagram showing the configuration of the generation server 4000. The generation server 4000 includes a processor 4001, a memory 4002, a storage 4003, and a communication interface 4004.

[0199] The memory 4002 may include a random access memory (RAM), a read only memory (ROM), a flash memory, or any other suitable memory system. The memory 4002 stores programs necessary for the arithmetic processing of the processor 4001, temporary data calculated in the arithmetic processing, and the like.

[0200] The storage 4003 is configured with a hard disk drive, a solid state drive, etc. The storage 4003 stores a database 4420 and a large-scale language model 430. The large-scale language model 430 is an example of a natural language processing algorithm. The storage 4003 is an example of a storage unit that stores a natural language processing algorithm.

[0201] The large-scale language model 430 is a language model (trained language model) that has been trained in advance by machine learning. A huge amount of text data is used to train the large-scale language model 430. The large-scale language model 430 is formed as an autoregressive model that uses a transformer such as a GPT (Generative Pre-trained Transformer). The large-scale language model 430 according to the second embodiment may include Bard and the like in addition to GPT (GPT-2, GPT-3, GPT-4).

[0202] The database 4420 includes a plurality of types of databases, including a work history database (work history DB) 4421, a dialogue information database (dialogue information DB) 4422, and a dictionary database (dictionary DB) 4423.

[0203] The memory 4002 stores a program (algorithm) 4050. The processor 4001 executes the program (algorithm) 4050 to utilize the large-scale language model 430 and create a resume.

[0204] The program 4050 includes an interaction program 4051, a confirmation program 4052, and a creation program 4053. By executing the interaction program 4051, the processor 4001 utilizes the large-scale language model 430 to interact with the user. As a result, the processor 4001 acquires a large amount of work history information from the user. By executing the confirmation program 4052, the processor 4001 utilizes the large-scale language model 430 to confirm that all work history information necessary for creating a career history has been acquired. By executing the creation program 4053, the processor 4001 utilizes the large-scale language model 430 to create a career history in accordance with a predetermined format.

[0205] [Database 4420] The database 4420 will be described below. In the work history database 4421, "work history information" and "work history" are registered for each user's membership ID. In this disclosure, "work history information" refers to information used to create a "work history." For example, the generation server 4000 can receive work history information in file format from the user via the user device 500 before interacting with the user. When the generation server 4000 receives work history information from the user, it registers the work history information in the work history database 4421 for each user's membership ID.

[0206] If the work history information registered in the work history database 4421 includes all the information necessary to create a work history, the generation server 4000 creates the work history without interacting with the user.

[0207] The generation server 4000 interacts with the user to acquire work history information from the user. The "work history information" acquired through the interaction is registered in the interaction information database 4422 for each member ID. The generation server 4000 creates a "work history" using the "work history information" registered in the interaction information database 4422. If "work history information" is registered in the work history database 4421, the generation server 4000 creates a "work history" using the "work history information" registered in the work history database 4421 and the "work history information" registered in the interaction information database 4422. The generation server 4000 registers the created "work history" in the work history database 4421. The configuration of the work history database 4421 will be described in detail later using FIG. 27.

[0208] The dictionary database 4423 stores terms and their meanings extracted during the dialogue between the generation server 4000 and the user. The generation server 4000 associates terms registered in the dictionary from the resume with the dictionary database 4423. The generation server 4000 (or the sharing server 100) displays the resume on the user device 500. When the user clicks with the mouse on a term registered in the dictionary from the resume, the generation server 4000 (or the sharing server 100) displays the meaning of the term on the user device 500.

[0209] The large-scale language model 430 or part of the multiple types of databases may be stored in storage provided separately from the generation server 4000. For example, the generation server 4000 may be connected to a cloud service separate from the generation server 4000, and part of the multiple types of databases shown in FIG. 22 or the large-scale language model 430 may be stored on that cloud. In this case, the generation server 4000 can access the necessary database or large-scale language model 430 by communicating with the cloud via the Internet 50.

[0210] The processor 4001 connects to the Internet 50 via the communication interface 4004 in accordance with a program stored in the memory 4002. The processor 4001 connects to the Internet 50 and communicates with the user device 500 (see FIG. 20 ). The processor 4001 accesses the database 4420 and executes processes such as extracting necessary data, registering new data in the database 4420, and updating data registered in the database 4420.

[0211] The processor 4001 communicates with the sharing server 100. For example, in response to a request from the sharing server 100, the processor 4001 transmits the career history of the member (user) registered in the career history database 4421 to the sharing server 100.

[0212] [Company Database 121] Fig. 23 is a diagram showing an example of the company database 121. In the company database 121, a company ID for identifying the company, a company name, a company address, etc. are registered for each company. In the second embodiment, members are permitted to apply for jobs recruited by various companies and departments and to receive orders for those jobs.

[0213] 24 is a diagram showing an example of the user database 122. Various information about members is registered in the user database 122. The various information about members includes a member ID for identifying the member, the ID of the company to which the member belongs, the member name, the member's authority, and the department to which the member belongs.

[0214] The types of member authority include administrator and applicant. A member with administrator authority is given the authority to use the classification system 1000 as a recruiter and an applicant. A member with applicant authority is given the authority to use the classification system 1000 as an applicant, but not as a recruiter. A department head within a company is given administrator authority to manage the side jobs of his or her subordinates within the department. A manager with administrator authority is given the authority to approve applications from his or her subordinate applicants. Therefore, the manager functions as an approver.

[0215] [Community Database 127] Fig. 25 is a diagram showing an example of the community database 127. Information on communities formed between companies is registered in the community database 127. The community information includes a community ID for identifying the community, a community name, and a list of IDs of companies belonging to the community. Each company can form various communities by agreeing with other companies. A company belonging to a community can change the companies that belong to the community by agreeing with the other companies.

[0216] 26 is a diagram showing an example of the project database 123. Information on open projects is registered in the project database 123. The information on open projects includes a project ID for identifying the open project, the ID of the company to which the recruiter who registered the open project belongs, a list of non-disclosure company IDs, a disclosure level, a project title, an estimated number of man-hours, an estimated period, and project details.

[0217] The non-disclosure company ID list registers the IDs of companies that prohibit the disclosure of job postings. The disclosure level can be set to one of three levels: "Company," "Within the community," or "All." If the disclosure level is set to "All," applicants outside the community will also be disclosed.

[0218] The IDs of companies that can view the recruitment project are shown on the right side of the project database 123 in Fig. 26. For example, the disclosure level for the recruitment project corresponding to project ID = 001 is set to "our company." In this case, only members belonging to the company that registered the recruitment project (company ID = 00A) can view the recruitment project corresponding to project ID = 001.

[0219] Hereinafter, using the case ID, the recruiting cases corresponding to each case ID may be referred to as case 001, case 002, case 003, etc. Similarly, using the community ID, the communities corresponding to each community ID may be referred to as community 01, community 02, community 03, etc., and using the member ID, the members corresponding to each member ID may be referred to as member P1, member P2, member P3, etc. Also, hereinafter, using part of the company ID, the companies corresponding to each company ID may be referred to as company A, company B, company C, etc.

[0220] The disclosure level for case 002 is set to "within the community." According to the community database 127 shown in Fig. 25, the companies that have a community relationship with company A, which registered case 002, are company B and company C. Therefore, as shown in Fig. 26, only members belonging to company A, company B, or company C can view case 002.

[0221] Case 003 has the same registered companies and disclosure level as case 002. However, case 003 has "00B" registered in the non-disclosure company ID list. Therefore, as shown in Figure 26, only members belonging to either company A or company C can view case 003, and members belonging to company B are not authorized to view case 003.

[0222] If the disclosure level of a job posting is set to "all," all members can view the job posting. This is the case for job posting 005 shown in Figure 26. If one or more company IDs are registered in the non-disclosure company ID list for job posting 005, members belonging to the companies with those company IDs will not be authorized to view job posting 005.

[0223] The estimated effort and duration are used by the applicant and classification system 1000 to estimate the time to process a job posting.

[0224] Each of the company database 121, the user database 122, and the community database 127 may be shared between the sharing server 100 and the generating server 4000. The job database 123 may register the member ID of a recruiter corresponding to a job.

[0225] [Work History Database] FIG. 27 is a diagram showing an example of the work history database 4421. As shown in FIG. 27, the work history database 4421 registers the "work history information" and "work history" of each member (user) for each member ID. As described above, the "work history information" refers to information used to create a "work history." The "work history" is created based on the "work history information." Note that in the second embodiment, the "work history information" may be acquired in advance before the user interacts with the generation server 4000, or may be acquired through the user interacting with the generation server 4000.

[0226] The format of the resume is predetermined by the designer, and includes the types of information that must be included and the order in which the information must be included.

[0227] The required information to be included in a resume includes "formal items" and "substantive items." For example, the formal items include "name," "age," "gender," "work history summary," "work history (duration and content)," "qualifications and skills," and "self-promotion." As shown in FIG. 27 , these formal items are arranged in the resume in the following order: "name," "age," "gender," "work history summary," "work history (duration and content)," "qualifications and skills," and "self-promotion." In the resume created by the generation server 4000, the "formal items" shown in FIG. 27 are listed in the order shown in FIG. 27 . Note that the types and order of the "formal items" shown in FIG. 27 are merely examples.

[0228] A resume includes content corresponding to each of the "formal items." The content to be included in a resume includes predetermined "substantial items." In the second embodiment, an item called "STAR" is introduced as an example of a "substantial item."

[0229] "STAR" is generally known as one of the methods used by interviewers to effectively interview job seekers. "STAR" is a coined word created by combining the initials of "S (Situation)," "T (Task)," "A (Action)," and "R (Result)." The generation server 4000 acquires work history information from the user so that the user's work history items related to each of the four items intended by "STAR" are included, and creates a work history using the acquired work history information.

[0230] The work history database 4421 stores resume format information in addition to the resumes (including content) created for each user. This format information includes the necessary information (formal and substantive information) to be included in a resume. The generation server 4000 executes a process related to supporting the creation of a resume while referring to the formats registered in the work history database 4421.

[0231] The work history information registered in the work history database 4421 is information that serves as the basis for creating a work history, similar to the work history information obtained through dialogue between the generation server 4000 and the user. If work history information is not registered in the work history database 4421, the generation server 4000 creates a work history based on the "work history information" obtained through dialogue with the user. Therefore, in the present disclosure, it is not essential that work history information be registered in the work history database 4421. The work history database 4421 is an example of a work history database in which work history information collected before dialogue by the collection unit is registered.

[0232] The "work history information" is, for example, information about the name, department, job description, work experience, and employment period of the company where the user has previously worked, and includes information about the above-mentioned "STAR." Alternatively, the "work history information" is information about the qualifications held by the user.

[0233] "Work history information" includes fragmentary information about the user's work history. Fragmentary information is, for example, information about one of the multiple required items included in a work history. One of the multiple required items is, for example, information about a qualification held by the user. Alternatively, fragmentary information is information that constitutes part of the required items. For example, if a user holds three qualifications, fragmentary information is information about one of the three qualifications.

[0234] The work history database 4421, together with the dialogue information database 4422 and the dictionary database 4423, constitutes the database 4420. Here, the work history database 4421 out of the databases 4420 has been described in detail using the drawings. The dialogue information database 4422 and the dictionary database 4423 have already been described in a manner that allows the understanding of their respective configurations, and therefore the description of the dialogue information database 4422 and the dictionary database 4423 will not be repeated here.

[0235] 28 and 29 are diagrams showing an example of a dialogue between the generation server 4000 and a user. The user accesses the generation server 4000 using the user device 500. The generation server 4000 uses the large-scale language model 430 to dialogue with the user in the following procedure. In particular, the generation server 4000 diverges the dialogue with the user so that all of the user's work history is acquired without omission from the perspective of STAR (Situation, Task, Action, Result) related to the required information. In this way, the generation server 4000 acquires all of the user's work history information necessary for creating a resume.

[0236] Here, an example will be described in which the generation server 4000 interacts with the user using the screen of the user device 500. That is, the generation server 4000 inquires of the user about their work history by displaying a request or question on the screen of the user device 500. The user responds to the inquiry by inputting text using a keyboard or the like of the user device 500. The generation server 4000 may also interact with the user by voice.

[0237] First, the generation server 4000 displays a request message on the screen saying, "Please enter your work history" (step SA1). The user responds to the request in step SA1 by saying, "I was a lead designer at XX company" (step SA2).

[0238] Based on the response in step SA2, the generation server 4000 displays the query "What was the work environment like?" on the screen (step SA3). The user responds to the query in step SA3 by saying, "The team was short-staffed and had a large number of backlogged cases. The account manager set an unreasonable deadline, which caused stress for the team and led to low morale" (step SA4).

[0239] Based on the response in step SA4, the generation server 4000 displays on the screen the query, "Please explain the responsibilities and roles you played in the situation and task at that time" (step SA5). In response to the query in step SA5, the user responds, "My role as a team leader was not only to ensure that the team worked within the deadline, but also to communicate our processing capabilities to other departments and maintain the motivation of the team" (step SA6).

[0240] In the same manner, the generation server 4000 communicates with the user in a chat format to acquire all of the work history information necessary to create a resume from the user. The generation server 4000 then creates a resume based on the work history information. The generation server 4000 displays a request message on the screen stating, "I have created your work history. Please check it." (Step SA7). The generation server 4000 then presents the created resume to the user (Step SA8). The user checks the resume on the screen of the user device 500.

[0241] In particular, the generation server 4000 diverges the interaction between the generation server 4000 and the user so that the entire work history of the user is acquired without omission from the perspective of STAR (Situation, Task, Action, Result).

[0242] Step SA3 is an example of an inquiry for obtaining work history information corresponding to a "Situation" from a user. Step SA5 is an example of an inquiry for obtaining work history information corresponding to a "Task" from a user. Although not illustrated in FIGS. 28 and 29, an "Action" relates to the method by which the user overcame the "Situation" or "Task." A "Result" relates to the results obtained by the user's "Action."

[0243] The dialogue program 4051 includes parameters for allowing the dialogue to diverge in terms of STAR (Situation, Task, Action, Result). Such parameters are designed, for example, by the system administrator of the generation server 4000.

[0244] In addition, in addition to, or instead of, chat-style dialogue, the generation server 4000 may allow the user to send a CSV (Comma Separated Value) file or the like related to work history information from the user device 500 to the generation server 4000.

[0245] Some examples of "venting" are as follows. For example, when information acquired from a user is expanded into a sentence, a situation may arise in which the number of characters in the sentence does not meet the number of characters specified in the required information. More specifically, when asking the question "What is the work environment like?", the answer obtained may not reach the specified number of characters (e.g., 100 characters or more). In such a case, the generation server 4000 asks the user an additional question such as "How was your relationship with your direct supervisor?" or "How were your relationships with your colleagues?"

[0246] Alternatively, if the answer obtained from the user is not specific enough, the generation server 4000 asks the user an additional question. For example, if the reason is not clear from the user's answer, the generation server 4000 asks the user an additional question. For example, if the user answers "The team is short-staffed..." but the user's answer does not contain a "keyword" or the like indicating the reason for the shortfall, the generation server 4000 asks the user an additional question such as "Why were there a shortfall?"

[0247] If the answer obtained from the user does not contain enough information about the "reason," the generation server 4000 may present the user with options for the reason after organizing the information using the large-scale language model 430. For example, suppose the answer obtained from the user is, "The team was short-staffed and had a large number of backlogged cases. The account manager set an unreasonable deadline, which caused stress for the team and lowered morale."

[0248] In this case, the generation server 4000 determines, for example, that (a) the reason for the shortage of personnel and (b) the reason why the user has a large number of unprocessed cases are insufficient, and presents the user with options to obtain each reason.

[0249] To obtain "(a) Reason for personnel shortage," the generation server 4000 may, for example, ask the user, "I understand that there was a personnel shortage. Could the reason be one of the following? (Multiple choices allowed)," and may also present the user with the following options: "A: Even after hiring, we were unable to attract new employees," "B: There was a sudden increase in orders," "C: Many people are quitting," and "D: Other (free description)."

[0250] To obtain "(b) the reason why the user ended up with a large number of unprocessed cases," the generation server 4000 may, for example, ask, "I understand that you had a large number of unprocessed cases. Could one of the following be the reasons? (Multiple choices allowed)," and may also present the user with the following options: "A: There was a lack of information sharing between sales and development, and too many cases were accepted compared to the number of developers," "B: The account manager lacked the management skills," "C: The development staff lacked the skills," and "D: Other (free description)."

[0251] The user may select three options, A, B, and C, as answers to (a) above, and two options, A and B, as answers to (b) above.

[0252] As a result, the generation server 4000 can extract more detailed information from the user about the reasons for (a) and (b) above. For example, before asking the user additional questions, the information obtained from the user might be, "The team was short-staffed and had a large backlog of cases. The account manager had set an unreasonable deadline, which caused stress for the team and low morale."

[0253] In response to this, by asking additional questions regarding (a) and (b) above, the generation server 4000 can extract information from the user such as, for example, "There was a lack of information coordination between sales and development, the number of projects was excessive and too rapid compared to the number of developers, and the account managers lacked the management skills to handle all the projects. Because there were so many projects and working hours tended to be excessive, many people quit, which led to an even greater shortage of personnel. Furthermore, the company was looking to hire mid-career employees who could contribute immediately, but was unable to find anyone."

[0254] 30 and 31 are diagrams showing an example of a dialogue (with dictionary registration) between the generation server 4000 and a user. Here, the operation of the generation server 4000 when the generation server 4000 detects a term to be registered in the dictionary database 4423 in the dialogue shown in FIGS. 28 and 29 will be described.

[0255] 30, the generation server 4000 detects the term "lead designer" from the input of step SA2 as a term that is likely to be misunderstood. In this case, the generation server 4000 displays the query "Does lead designer mean XXXXXX?" on the screen (step SA2a).

[0256] In response to the inquiry in step SA2a, the user replies, "No, I mean XXXXXX." (step SA2b). Based on the reply in step SA2b, the generation server 4000 displays the sentence, "I understand." on the screen (step SA2c). Furthermore, in step SA2c, the generation server 4000 registers the term "lead designer" and its meaning as a set in the dictionary database 4423.

[0257] The generation server 4000 then continues to interact with the user and acquires all of the work history information necessary to create the resume from the user. The generation server 4000 then creates the resume based on the work history information. The generation server 4000 displays a request message on the screen stating, "Your work history has been created. Please confirm it." (Step SA7). The generation server 4000 then presents the created resume to the user (Step SA8).

[0258] As shown in FIG. 31 , the term "lead designer" is underlined in the resume. The user understands that the underlined term is linked to the dictionary. When the user clicks on the term displayed on the screen with a mouse or the like, the meaning of the term registered in the dictionary database 4423 is displayed on the screen. The user checks the resume, which includes a link to the dictionary, on the screen of the user device 500. The generation server 4000 links the resume approved by the user to the user's membership ID and registers it in the resume database 4421.

[0259] The resume registered in the work history database 4421 is transmitted to the recruiter device 200 in response to a user's request. The recruiter device 200 displays the resume including a link to the dictionary on the recruiter device 200. When the recruiter clicks a term displayed on the screen with a mouse or the like, the meaning of the term registered in the dictionary database 4423 is displayed on the screen. This allows the recruiter to accurately understand the terms written in the resume.

[0260] [Functions of the Generation Server] FIG. 32 is a diagram for explaining the content of the processing executed by the generation server 4000 from the viewpoint of the divergence phase and the convergence phase.

[0261] The generation server 4000 has a large-scale language model 430 and a program 4050 (a dialogue program 4051, a confirmation program 4052, and a creation program) that executes processing to support the creation of a curriculum vitae using the large-scale language model 430. The processing to support the creation of a curriculum vitae is divided into a divergence phase and a convergence phase, as shown in FIG.

[0262] The dialogue program 4051 and the large-scale language model 430 are combined to form a "dialogue AI (Artificial Intelligence) 4110." The dialogue AI 4110 acquires from the user a large amount of work history information necessary for creating a resume. Furthermore, the dialogue AI 4110 inquires of the user whether the created resume meets the user's intentions.

[0263] The "confirmation AI 4120" is configured by combining the confirmation program 4052 and the large-scale language model 430. The confirmation AI 4120 confirms that all work history information corresponding to the necessary items required to create a resume has been obtained without omission.

[0264] The "creation (summarization) AI 4130" is configured by combining the creation program 4053 and the large-scale language model 430. The creation (summarization) AI 4130 creates a resume in accordance with a predefined resume format.

[0265] In the divergence phase, the dialogue AI 4110 chats with the user and acquires from the user work history information related to the necessary items to be included in the resume. At this time, the dialogue AI 4110 controls the dialogue so that the dialogue diverges from the perspective of "STAR (Situation, Task, Action, Result)" according to parameters designed by the system administrator.

[0266] In the divergence phase, the confirmation AI 4120 reads the resume format. The resume format contains the necessary information required to create a resume. The confirmation AI 4120 verifies that all work history information corresponding to the necessary information required to create a resume has been obtained without omission. If work history information is missing, the confirmation AI 4120 instructs the dialogue AI 4110 to continue the dialogue regarding the missing work history information.

[0267] The dialogue AI 4110 stores the dialogue history. Therefore, when the dialogue AI 4110 is instructed by the confirmation AI 4120 to continue the dialogue, it can continue the dialogue from where it left off without having to restart the dialogue that has already been completed.

[0268] In the divergence phase, divergent dialogue by the dialogue AI 4110 and confirmation by the confirmation AI 4120 are repeated until all the work history information corresponding to the necessary items to be written in creating a resume is obtained without omission.

[0269] After completing the divergence phase, the generation server 4000 moves to the convergence phase. In the convergence phase, the creation AI (summarization AI) 4130 has the function of arranging the acquired information into a predetermined format. More specifically, the creation AI 4130 creates a resume that conforms to a pre-defined resume format. The dialogue AI 4110 presents the created resume to the user. The user checks the resume. If the user determines that the resume does not meet the user's intentions, the dialogue AI 4110 instructs the dialogue AI 4110 to make corrections. In this case, the creation AI 4130 corrects the resume based on the user's correction instructions. The dialogue AI 4110 presents the corrected resume to the user. The user checks the resume.

[0270] In the convergence phase, the creation (modification) of the resume by the creation AI 4130 and the dialogue between the dialogue AI 4110 and the user are repeated until the resume intended by the user is created.

[0271] Fig. 33 is a diagram showing the functional configuration of the generation server 4000. As shown in Fig. 33, the generation server 4000 includes a control unit 4010. The control unit 4010 is realized by the processor 4001, memory 4002, and communication interface 4004 shown in Fig. 22.

[0272] The control unit 4010 can function as the dialogue AI 4110, the confirmation AI 4120, and the creation AI 4130 shown in Fig. 32 by utilizing the large-scale language model 430. The control unit 4010 accesses the database 4420 to execute processing related to resume creation support.

[0273] The control unit 4010 includes a data import unit 4011 , an information acquisition unit 4012 , a dictionary information registration unit 4013 , a dictionary information display unit 4014 , an information organization unit 4015 , and a resume registration unit 4016 .

[0274] The data import unit 4011 imports work history information already owned by the user into the dialogue AI 4110. For example, the data import unit 4011 can receive work history information in a predetermined file format from the user via the user device 500. The control unit 4010 uses the imported work history information to create a resume. In this way, by utilizing the work history information already owned by the user, the dialogue cost of the dialogue AI 4110 can be reduced.

[0275] The information acquisition unit 4012 acquires work history information from the user using the dialogue AI 4110. The information acquisition unit 4012 uses the confirmation AI 4120 to confirm that all work history information corresponding to the necessary items to be written in creating a resume has been acquired without omission.

[0276] The dictionary information registration unit 4013 registers unique terms obtained during the dialogue between the user and the dialogue AI 4110 in the dictionary database 4423. Here, the unique terms may be, for example, information unique to an organization such as a company.

[0277] If a term registered in the dictionary database 4423 is included in the resume, the dictionary information display unit 4014 displays the term in the resume in the form of a link. The information organizing unit 4015 uses the creation AI 4130 to create the resume in a predetermined format.

[0278] The curriculum vitae registration unit 4016 stores the created curriculum vitae in a curriculum vitae database 4421. In particular, when a term registered in the dictionary database 4423 is written in the curriculum vitae, the curriculum vitae registration unit 4016 generates a link that associates the term written in the curriculum vitae with the meaning of the term registered in the dictionary database 4423, and then stores the curriculum vitae information in the curriculum vitae database 4421.

[0279] 34 to 39, an example of the processing procedure of the control unit 4010 included in the generation server 4000 will be described. Here, as the processing procedure of the control unit 4010, the processing procedures of the data import unit 4011, the information acquisition unit 4012, the dictionary information registration unit 4013, the dictionary information display unit 4014, the information organization unit 4015, and the resume registration unit 4016 will be described.

[0280] 34 is a flowchart showing the processing steps of the data import unit 4011. First, the data import unit 4011 acquires, from the work history database 4421, format information for the resume and work history information registered before the dialogue between the dialogue AI 4110 and the user (step SA11). Step SA11 is an example of a collection unit that collects work history information related to the user's work history. Furthermore, step SA11 is an example of an acquisition unit that acquires format information including the necessary information to be written in the resume.

[0281] The "career history information registered before the dialogue between the dialogue AI 4110 and the user" refers to the "career history information" shown in Figure 27. In step SA11, the data import unit 4011 may read the user's career history information from a specified local file. The resume format information includes the necessary information to be included in the resume and the order in which the necessary information should be arranged.

[0282] Next, the data import unit 4011 inputs the work history information acquired in step SA11 into the dialogue AI 4110 as prior information (step SA12). Next, the data import unit 4011 determines whether the work history information acquired in step SA11 includes all the information necessary to create a resume (step SA13). The data import unit 4011 calls the confirmation AI 4120 and makes the determination in step SA13 by referring to the format information of the resume.

[0283] If the work history information acquired in step SA11 contains all the information necessary to create a curriculum vitae, no dialogue between the dialogue AI 4110 and the user is required. Therefore, if the work history information acquired in step SA11 contains all the information necessary to create a curriculum vitae, the data import unit 4011 passes the process to the information organizing unit 4015. The information organizing unit 4015 creates the curriculum vitae.

[0284] If the work history information acquired in step SA11 does not include all the information necessary to create a work history, the data import unit 4011 passes the process to the information acquisition unit 4012.

[0285] 35 is a flowchart showing the processing procedure of the information acquisition unit 4012. First, the information acquisition unit 4012 calls the dialogue AI 4110 and the confirmation AI 4120 using the large-scale language model 430 (step SA21).

[0286] Next, the information acquisition unit 4012 uses the dialogue AI 4110 to engage in a dialogue with the user (step SA22). As a result, the information acquisition unit 4012 acquires dialogue information (including work history information) from the user. Step SA22 is an example of a collection unit that collects work history information related to the user's work history. Furthermore, step SA22 is an example of a dialogue unit that collects work history information by dialogue with the user. Next, the information acquisition unit 4012 stores the dialogue information in the dialogue information database 4422 (step SA23). The dialogue information is information indicating the history of dialogue with the user.

[0287] Next, the information acquisition unit 4012 determines whether all of the work history information corresponding to the required items necessary for creating a curriculum vitae has been acquired without omission, based on the dialogue information (step SA24). At this time, the information acquisition unit 4012 makes the determination of step SA24 by referring to the format information acquired in step SA11. Note that the information acquisition unit 4012 may acquire format information from the work history database 4421 in a step separate from step SA11. If all of the work history information corresponding to the required items has not been acquired without omission, the information acquisition unit 4012 returns the process to step SA22.

[0288] When all work history information corresponding to the required fields has been acquired without omission, the information acquisition unit 4012 confirms with the user whether there is other work history information (step SA25). If there is other work history information, the information acquisition unit 4012 returns the process to step SA22. If there is no other work history information, the information acquisition unit 4012 passes the process to the information organization unit 4015. The information organization unit 4015 creates a work history.

[0289] 36 is a flowchart showing the processing procedure of the dictionary information registration unit 4013. First, the dictionary information registration unit 4013 detects words to be registered in the dictionary during a dialogue between the dialogue AI 4110 and the user (step SA31). Words to be registered in the dictionary include proper nouns and polysemous word information. Words to be registered in the dictionary may also be information specific to an organization such as a company. Next, the dictionary information registration unit 4013 confirms the meaning of the word with the user (step SA32).

[0290] Next, the dictionary information registration unit 4013 acquires the user's answer (step SA33). Next, the dictionary information registration unit 4013 registers the word detected in step SA31 and its meaning based on the user's answer as dictionary information in the dictionary database 4423 (step SA34), and ends the processing based on this flowchart. Step SA34 is an example of a registration unit that registers the meaning of the term included in the work history information in the dictionary database.

[0291] 37 is a flowchart showing the processing steps of the dictionary information display unit 4014. First, the dictionary information display unit 4014 acquires the user's career history from the career history database 4421 (step SA41). Next, the dictionary information display unit 4014 acquires terms present in the career history from the dictionary database 4423 (step SA42). Next, the dictionary information display unit 4014 displays the career history with a link on the screen of the user device 500 (step SA43).

[0292] Next, the dictionary information display unit 4014 displays the dictionary information corresponding to the link on the screen of the user device 500 in response to the user's operation (step SA44), and ends the processing based on this flowchart. The dictionary information display unit 4014 is an example of a display unit that displays the resume on a display device. The user device 500 is an example of a display device. Note that the dictionary information display unit 4014 may display the resume on a display device provided in the generation server 4000 instead of or in addition to the user device 500.

[0293] 38 is a flowchart showing the processing procedure of the information organizing unit 4015. First, the information organizing unit 4015 calls the creation AI 4130 using the large-scale language model 430 (step SA51).

[0294] Next, the information organizing unit 4015 reads the dialogue information from the dialogue information database 4422 (step SA52). If pre-registered work history information exists in the work history database 4421, the information organizing unit 4015 reads the work history information from the work history database 4421 in step SA52.

[0295] Next, information organizing unit 4015 creates a curriculum vitae from the dialogue information (step SA53). If information organizing unit 4015 has read work history information from work history database 4421, it creates a curriculum vitae from the work history information and dialogue information in step SA53. Step SA53 is an example of a creation unit that creates a curriculum vitae.

[0296] The information organizing unit 4015 references the format information acquired in step SA11 and creates a resume in a format that conforms to the format information. Note that the information acquiring unit 4012 may acquire the format information from the work history database 4421 in a step separate from step SA11. Next, the information organizing unit 4015 displays the resume on the screen of the user device 500 (step SA54).

[0297] Next, the information organizing unit 4015 determines whether or not an approval operation from the user has been received (step SA55). If an approval operation from the user has been received, the information organizing unit 4015 ends the processing based on this flowchart.

[0298] If the information organizing unit 4015 has not received an approval operation from the user, it determines whether or not a non-approval operation from the user has been received (step SA56). If the information organizing unit 4015 has not received a non-approval operation from the user, it returns the process to step SA55. If the information organizing unit 4015 has received a non-approval operation from the user, it passes the process to the information acquisition unit 4012. The information acquisition unit 4012 again calls the dialogue AI 4110 and acquires new dialogue information from the user by interacting with the user.

[0299] The information organizing unit 4015 may not only acquire new dialogue information from the user, but also re-engage in a dialogue with the user without changing the conditions before the dialogue, because a large-scale language model may generally generate multiple answers (sentences) with different nuances for the same question and conditions.

[0300] 39 is a flowchart showing the processing steps of the resume registration unit 4016. First, the resume registration unit 4016 acquires the user's resume from the work history database 4421 (step SA61). Next, the resume registration unit 4016 references the dictionary database 4423 to generate a link between the resume and dictionary information (step SA62). Next, the resume registration unit 4016 stores the linked resume in the work history database 4421 (step SA63).

[0301] Next, the resume registration unit 4016 receives a resume output request from the user (step SA64). Next, the resume registration unit 4016 outputs the resume to the user device 500 (step SA65), and ends the processing based on this flowchart. Note that the user device 500 may display the resume on the screen of the user device 500. Alternatively, the user device 500 may store the resume data in memory until instructed by the user.

[0302] [Processing sequence executed by classification system 1000] Fig. 40 is a timing chart showing the processing procedures related to resumes of the sharing server 100, the recruiter device 200, the applicant device 300, and the generation server 4000. The processing flow of the classification system 1000 related to resumes will be explained using the timing chart shown in Fig. 40.

[0303] First, the applicant accesses the generation server using applicant device 300 and creates a curriculum vitae (step SA101). Generation server 4000 registers the created curriculum vitae in curriculum vitae database 4421 (step SA102). The detailed processing procedures of steps SA101 and SA102 have already been described using Figures 34 to 39, and therefore will not be repeated here.

[0304] Next, the applicant accesses the sharing server 100 using the applicant device 300, searches for the job opening, and then decides where to apply (step SA103). Next, the applicant operates the applicant device 300 to send a command requesting the applicant's resume from the applicant device 300 to the sharing server 100 (step SA104). Upon receiving the command, the sharing server 100 identifies the applicant (step SA105). More specifically, the sharing server 100 identifies the applicant's member ID.

[0305] Next, the sharing server 100 transmits a command requesting transmission of the resume to the generation server 4000 (step SA106). This command includes the applicant's member ID. The generation server 4000 searches the applicant's resume from the work history database 4421 based on the member ID included in the command (step SA107).

[0306] Next, the generation server 4000 transmits the resume found as a result of the search to the sharing server 100 (step SA108). As a result, the sharing server 100 acquires the applicant's resume (step SA109). Next, the sharing server 100 transmits the acquired resume to the applicant device 300 (step SA110).

[0307] Applicant device 300 displays the received resume on the screen of applicant device 300 (step SA111). If the resume contains a term registered in dictionary database 4423, the resume including a link to the dictionary is displayed on the screen.

[0308] The applicant confirms that there are no problems with the resume displayed on the screen. For example, the applicant confirms that there is no information that needs to be updated in the resume. If there are no problems with the resume, the applicant performs an approval operation using a keyboard, etc. If there are problems with the resume, the applicant performs a disapproval operation using a keyboard, etc.

[0309] The applicant device 300 determines whether an approval operation has been detected (step SA112). If the applicant device 300 detects a non-approval operation rather than an approval operation, it returns to step SA101. In this case, the process of creating a resume is executed again. However, it is preferable that the generation server 4000 not execute the process of creating a resume from the beginning, but rather interact with the applicant regarding items that need to be corrected and correct the contents of the resume.

[0310] When the applicant device 300 detects the approval operation, it transmits a command to the sharing server 100 to instruct the sending of the resume (step SA113). The sharing server 100 identifies the recruiter to whom the resume should be sent (step SA114). For example, the sharing server 100 identifies the recruiter to whom the resume should be sent based on the application destination determined in step SA103.

[0311] Next, the sharing server 100 transmits the resume to the recruiter device 200 (step SA115). The recruiter device 200 displays the received resume on the screen of the recruiter device 200 (step SA116). If the resume contains a term registered in the dictionary database 4423, the resume including a link to the dictionary is displayed on the screen. The recruiter uses the resume to understand the applicant's work history. When the applicant clicks with the mouse on a term registered in the dictionary in the resume, the meaning of the term is displayed on the recruiter device 200. Therefore, if the resume contains a term that the applicant does not understand, such as internal jargon of the company where the applicant works, the recruiter can understand the meaning of the term.

[0312] 40, the sharing server 100 and the generation server 4000 may be configured as a single server. That is, the sharing server 100 may be provided with the functions of the generation server 4000, or the generation server 4000 may be provided with the functions of the sharing server 100. Steps SA110 and SA115 are an example of an output unit configured to output the resume created by the creation unit to a user device.

[0313] As described above, according to the second embodiment, it is possible to assist a user in creating a resume that includes all the necessary information and does not contain any duplicate information. In addition, according to the second embodiment, it is possible to assist a user in creating a resume that appropriately reflects the user's work history, regardless of the user's document creation ability. Furthermore, according to the second embodiment, the following effects are achieved.

[0314] a. It is possible to provide a user with a resume that adequately expresses the user's skills and experience, regardless of the user's document creation ability.

[0315] b) A wide variety of work history information of the user can be collected without omission. c) By providing the user with an interactive interface, the burden on the user to appropriately extract work history information can be reduced.

[0316] d) It reduces the burden on users when creating a resume. e) Because resumes are created in a standardized format, it is possible to improve the efficiency of managing and searching a large number of resumes.

[0317] f. It is possible to provide an interface for creating a high-quality resume to users who find it difficult to stay motivated to complete the cumbersome task of creating a resume, such as employees applying for an internal project or employees looking to take on a side job outside the company.

[0318] g. If the resume contains a term registered in the dictionary database 4423, the resume will be displayed on the screen with a link to the dictionary, so that if the resume contains a term that the recruiter does not understand, they can understand the meaning of the term.

[0319] Note that part of the configuration of the generation server 4000 may be configured by a device separate from the generation server 4000. For example, at least one of the multiple databases included in the database 4420 and the large-scale language model 430 may be placed on a cloud separate from the generation server 4000. In this case, it is sufficient that the generation server 4000 and the cloud are connected to each other via a network so as to be able to communicate with each other.

[0320] [Modification] In the second embodiment, an example has been shown in which the generation server 4000 is arranged in the classification system 1000. However, a generation server that creates a curriculum vitae may be constructed independently of the classification system 1000. An example of a generation server that creates a curriculum vitae independently of the classification system 1000 will be described.

[0321] 41 is a diagram showing the functional configuration of a generation server 4000A according to a modified example. Similar to the generation server 4000, the generation server 4000A has components such as a processor, a memory, a communication interface, and storage, and includes a control unit 4010A realized by these components.

[0322] The generation server 4000A is communicably connected to a user device 500A used by a user via the Internet or the like.

[0323] The generation server 4000A includes a satisfying requirement database (satisfying requirement DB) 426, a raw information storage database (raw information storage DB) 427, a generative format database (generative format DB) 428, and large-scale language models 430A and 430B. These databases and large-scale language models are configured in the storage of the generation server 4000A.

[0324] Like the large-scale language model 430, the large-scale language models 430A and 430B are natural language processing models trained using large amounts of text data. In the present disclosure, BERT (registered trademark) (LaMBDA) by Google, GPT-4 by OpenAI, or the like may be adopted as the large-scale language model. In the present disclosure, the large-scale language model is an example of a natural language processing algorithm. A language model other than the large-scale language model may be adopted as the natural language processing algorithm. A model generated by machine learning such as pattern matching may be adopted as the natural language processing algorithm. Furthermore, instead of using two models, the large-scale language models 430A and 430B, a single large-scale language model may be used as the large-scale model. For example, the large-scale language model 430 may be used.

[0325] As shown in FIG. 41, the control unit 4010A includes an input unit 4021, a content sufficiency determination unit 4022, an input request unit 4023, a raw information storage unit 4024, a generation unit 4025, a confirmation request unit 4026, and an output unit 4027.

[0326] The input unit 4021 accepts information input by the user using the user device 500. The information input by the user includes work history information. The input unit 4021 may accept work history information in chat format, for example. The input unit 4021 may also accept work history information in audio file format and other file formats. When using GPT-3 or the like as a large-scale language model and a chat-style UI such as ChatGPT, a file may be attached to the chat. In this case, the generation server 4000 may be provided with a function for converting the file into a chat-style document.

[0327] The output unit 4027 outputs various information to the user device 500. The output unit 4027 may output not only chat-style information but also other files such as audio to the user device 500. When the control unit 4010A interacts with the user in chat style, the input of information to the input unit 4021 and the output of information from the output unit 4027 are repeatedly executed.

[0328] The requirements fulfillment database 426 stores information about the requirements (requirements fulfillment) that the large-scale language model 430A uses to determine whether all of the work history information corresponding to the necessary entries required for creating a resume has been acquired without omission.

[0329] Information regarding multiple types of fulfillment requirements may be registered in the fulfillment requirement database 426. In this case, the control unit 4010A may transmit multiple types of fulfillment requirements to the user device 500A and allow the user to select the fulfillment requirement of their choice. Alternatively, the control unit 4010A may accept fulfillment requirements set by the user via the user device 500A.

[0330] The control unit 4010A may also generate a trained model that functions as an AI using the contents of a large number of fulfillment requirements as training data. In this case, the generated trained model may be used instead of the large-scale language model. The fulfillment requirements may also be used in the prompts of an existing large-scale language model 430A. The fulfillment requirements may include specific information, such as the number of companies the user wants to include in their resume.

[0331] The content sufficiency determination unit 4022 determines whether the information obtained by the input unit 4021 satisfies the sufficiency requirements using the large-scale language model 430. If the content sufficiency determination unit 4022 determines that the information obtained by the input unit 4021 does not satisfy the sufficiency requirements, it instructs the input request unit 4023 to acquire additional information. If the content sufficiency determination unit 4022 determines that the information obtained by the input unit 4021 satisfies the sufficiency requirements, it outputs the information obtained by the input unit 4021 to the raw information storage unit 4024.

[0332] Based on the instruction of the content sufficiency determination unit 4022, the input request unit 4023 instructs the large-scale language model 430A to continue the dialogue with the user so that additional information (missing information) can be obtained from the user. When multiple pieces of additional information are required, the large-scale language model 430A may dialogue with the user so that the multiple pieces of additional information are acquired all at once, or may dialogue with the user so that each piece of additional information is acquired sequentially. For example, when information A, B, and C are missing, the large-scale language model 430A may first inquire about information A from the user.

[0333] The large-scale language model 430A may automatically generate a dialogue rule for prompting the user to input additional information, or may select a dialogue rule from preset options.

[0334] The raw information storage unit 4024 temporarily stores work history information that is the basis for generating a career history. The raw information storage unit 4024 outputs the temporarily stored work history information to the raw information storage database 427. The raw information storage database 427 registers the work history information. The generation unit 4025 creates a career history using the work history information registered in the raw information storage database 427. Note that the generation unit 4025 may have a function of creating a career history without waiting for a determination by the content sufficiency determination unit 4022.

[0335] For example, if the user has a draft of a resume, the user may only want to have the resume finalized or formatted. In this case, the input unit 4021 may accept data of the draft of the resume from the user via the user device 500. Furthermore, the input unit 4021 may register the accepted data in the raw information storage database 427. The generation unit 4025 may create the resume based on the draft data registered in the raw information storage database 427.

[0336] The generation format database 428 stores the formats required for the large-scale language model 430B to generate work histories.

[0337] Note that multiple types of formats may be registered in the generation format database 428. In this case, the control unit 4010A may transmit multiple types of formats to the user device 500A and allow the user to select a format of their choice. Alternatively, the control unit 4010A may accept a format set by the user via the user device 500A.

[0338] The control unit 4010A may also generate a trained model that functions as an AI using multiple formats as training data. In this case, the generated trained model may be used instead of the large-scale language model. Alternatively, a format may be used for the prompt of an existing large-scale language model 430A.

[0339] The generation unit 4025 generates a resume using the large-scale language model 430B. The format registered in the generation format database 428 is input to the large-scale language model 430B. The generation unit 4025 references the raw information storage database 427 and creates a resume that reflects the career history information in a predefined format. Note that the large-scale language model 430B may create the resume instead of the generation unit 4025. The resume generated by the generation unit 4025 is output to the confirmation request unit 4026.

[0340] The confirmation request unit 4026 requests the user to confirm the resume. The confirmation request unit 4026 transmits the resume to the user device 500 via the output unit 4027. The user confirms the content of the resume displayed on the user device 500. The confirmation request unit 4026 may divide the content of the resume into multiple items and have the user confirm the content of the resume for each item.

[0341] For example, if the resume contains entries for Company A and Company B, the confirmation request unit 4026 may have the user confirm the entry for Company A, and then have the user confirm the entry for Company B. Of course, the confirmation request unit 4026 may have the user confirm all of the contents of the resume at once. The large-scale language model 430B may automatically generate a dialogue rule for having the user confirm the resume, or may select a dialogue rule from preset options.

[0342] Note that part of the configuration of the generation server 4000A may be configured separately from the generation server 4000A. For example, at least one of the satisfying requirement database 426, the raw information storage database 427, the generation format database 428, the large-scale language model 430A, and the large-scale language model 430B may be configured separately from the generation server 4000A. In this case, it is sufficient that the device configured separately from the generation server 4000A and the generation server 4000A are communicably connected via a network.

[0343] [Variations Related to the Use of IPC Classification] Generally, an engineer's resume contains a detailed description of the skills the engineer possesses. However, with the rapid development of IT-related technology, previously little-known new technologies are emerging one after another, and technological diversification is progressing rapidly. On the other hand, the granularity of the skills possessed by users varies from field to field. As a result, consistency in the descriptions of skills in resumes is disappearing. This is hindering recruiters when analyzing applicants' resumes.

[0344] Similar problems exist within companies as well. It is important for those in a position to evaluate employees to understand and systematically organize the skills possessed by employees. This is thought to enable the company to make the most of its technical resources. However, due to the circumstances described above, even if each employee is required to submit a resume, it is difficult to uniformly manage each employee's skills based on a set standard.

[0345] Therefore, an index that can classify various skills using uniform criteria is needed. One possible index to achieve this is the International Patent Classification (IPC). The IPC is an internationally unified technology classification for classifying inventions that have been applied for patents. By using the IPC, technologies can be classified in detail by dividing them into hierarchical levels such as "section," "subsection," "class," "subclass," "main group," and "subgroup."

[0346] IPC is originally an index used to classify patent documents. However, by using IPC as an index to classify engineer skills, it is thought that it would be possible to classify a wide variety of engineer skills, including the latest technologies, using uniform standards. However, because IPC is a highly specialized classification index used in the patent industry, it is difficult for users who are unfamiliar with IPC to accurately select an IPC that corresponds to their skills from the many IPC options.

[0347] Therefore, we propose a system that allows a user to select an appropriate IPC that corresponds to their skills through a dialogue between the generation server 4000 and the user. Here, the IPC is assumed as an example of the required information or format information for a resume, and the user's career information required to obtain an IPC that corresponds to the user's skills is assumed as an example of work history information. The generation server 4000 collects the user's work history information required to obtain the IPC by dialogue with the user using a natural language processing algorithm. Note that skills are an example of work ability information.

[0348] 42 and 43 are diagrams showing another example of a dialogue between the generation server 4000 and a user. In Fig. 42 and Fig. 43, an example is shown in which the generation server 4000 acquires an appropriate IPC corresponding to the user's skills through a dialogue with the user.

[0349] First, generation server 4000 displays a request message on the screen saying, "Please enter your work history." (Step SA201). The user responds to the request in Step SA1 with their own work history (Step SA202). In Step SA202, for example, the user responds with work history related to the development of lithium-ion secondary batteries.

[0350] Based on the answer in step SA2, the generation server 4000 organizes the information using the large-scale language model 430 and presents to the user IPCs that may correspond to the user's skills (step SA203). In step SA203, for example, one or more "IPCs" corresponding to each of the information obtained through the dialogue, such as "research and development of lithium-ion secondary batteries," "development of new positive electrode materials," "development of lithium cobalt-nickel-manganese oxide," "development of silicon negative electrodes," "development of electrolytes that suppress degradation," and "development and practical application of lithium-ion secondary batteries with lithium titanate negative electrodes," are presented to the user.

[0351] Next, the generation server 4000 presents several IPCs from the IPCs presented in step SA203 to the user as selection candidates (step SA204). At this time, the generation server 4000 may present the selection candidates along with check boxes, as shown in FIG. 43 . The user considers whether there is an IPC that seems appropriate among the options. If there is an IPC that seems appropriate among the options, the user responds by checking the check box corresponding to that IPC. In this case, a button 601 as shown in FIG. 43 may be displayed on the screen of the user device 500. After checking the check box, the user clicks the button 601. When the generation server 4000 detects the click of the button 601, it determines the IPC to be included in the user's resume based on the user's response.

[0352] In this way, the generation server 4000 suggests to the user IPCs that are considered relevant from among a large number of IPCs based on the work history acquired through correspondence. Therefore, even if the user does not have specialized knowledge of IPCs, they can select IPCs that are relevant to their own skills. While the example shown here uses "subgroups," higher-level symbols such as "subclasses" or "groups" may be used depending on the desired granularity of classification. Furthermore, descriptions such as "subclasses," "groups," or "main groups" may be added to the description of "subgroups," or buttons such as question marks may be added to allow the user to check the contents of higher levels as needed.

[0353] Although an example has been given here in which the user is prompted to select an IPC related to his or her own technical field, the IPC presented to the user may be any of "section," "subsection," "class," "subclass," "main group," and "subgroup." For example, a subclass may be presented to the user, such as "H01: Electrical element." Furthermore, together with the classification symbol, the name of the technical field classified by that classification symbol may also be presented to the user.

[0354] If there is no suitable IPC among the options proposed in step SA205, the user can request generation server 4000 to present other candidates or to start over with entering work history. In this case, buttons 602 and 603 as shown in FIG. 43 may be displayed on the screen of user device 500.

[0355] When the generation server 4000 detects a click operation on the button 602, it selects an IPC different from the options presented in step SA204 from the IPCs presented in step SA203 and presents the selected IPC to the user. When the generation server 4000 detects a click operation on the button 603, it returns the process to step SA204 and prompts the user to input their work history.

[0356] Furthermore, IPC candidate generation may be performed automatically multiple times. The generation server 4000 may generate different answers to the same sentence (or the same question). For this reason, before proposing IPC candidates in step SA204, the IPC generation from step SA202 may be performed automatically multiple times to widen the range of IPC candidates. This increases the likelihood that the user will be able to select an IPC that is suitable for them.

[0357] As described above, the generation server 4000 supports the user's actions to identify IPCs corresponding to the user's skills through a dialogue between the generation server 4000 and the user. Note that the processing of step SA203 is not essential in embodiment 2. That is, the generation server 4000 may detect, through a dialogue, multiple IPCs that are thought to correspond to the user's skills, and then present some of the detected multiple IPCs to the user in step SA204 without executing the processing of step SA203.

[0358] The generation server 4000 may present other candidates to the user autonomously, not only when receiving a request from the user to present other candidates again. For example, the generation server 4000 may execute the process of generating option candidates multiple times and present multiple candidates to the user. Generally, answers obtained from a large-scale language model in response to a query may differ even if the content of the first query and the content of the second query are the same. Therefore, by having the generation server 4000 execute the process of generating option candidates multiple times, the generation server 4000 may derive multiple candidates from different perspectives. If the generation server 4000 can obtain a user's selection for each of these multiple candidates, the generation server 4000 can obtain a wider range of answers regarding IPC from the user.

[0359] FIG. 44 is a diagram showing the functional configuration of the generation server 4000, including an IPC classification unit 4017. In this modification, the IPC classification unit 4017 is added to the control unit 4010 of the generation server 4000 shown in FIG. 33. The IPC classification unit 4017 determines the user's IPC to be included in the resume. The IPC classification unit 4017 can present the user with IPCs that are considered to correspond to the user's skills using the user's work history information obtained through dialogue. The IPC classification unit 4017 can also present the user with IPCs that are considered to correspond to the user's skills using the work history information imported by the data import unit 4011. The IPC classification unit 4017 presents the user with multiple IPCs and then determines, through dialogue, an IPC that is appropriate for the user's skills. The determined IPC is imported into the information organization unit 4015. The information organization unit 4015 uses the creation AI 4130 to create a resume including the IPC in a predetermined format. The curriculum vitae registration unit 4016 stores the created curriculum vitae in a curriculum vitae database 4421 .

[0360] 45 is a flowchart showing the processing procedure of the IPC classification unit 4017. First, the IPC classification unit 4017 calls the dialogue AI 4110 using the large-scale language model 430 (step SA71).

[0361] Next, the IPC classification unit 4017 interacts with the user using the interaction AI 4110 (step SA72). As a result, the IPC classification unit 4017 acquires interaction information (including work history information) from the user. Step SA72 is an example of a collection unit that collects work history information related to the user's work history. Furthermore, step SA72 is an example of a dialogue unit that collects work history information by interacting with the user. Information related to IPCs may be stored in the dictionary database 4423. In this case, the IPC classification unit 4017 may refer to the dictionary database 4423 to determine the IPC to be presented to the user.

[0362] Next, the IPC classification unit 4017 presents to the user several IPCs that are considered to correspond to the user's skills (step SA73). Next, the IPC classification unit 4017 determines whether the user has performed an operation to select an IPC from the presented IPCs (step SA74).

[0363] If the user has not performed an operation to select an IPC, the IPC classification unit 4017 determines the user's request (step SA76). More specifically, if the user has requested the re-creation of an IPC, the IPC classification unit 4017 returns the process to step SA73, changes the IPC to be presented, and presents several IPCs to the user again. This process is executed, for example, when a click on button 602 shown in FIG. 43 is detected. If the user has requested a redo of the dialogue, the IPC classification unit 4017 returns the process to step SA72 and engages in a dialogue with the user again. This process is executed, for example, when a click on button 603 shown in FIG. 43 is detected.

[0364] For example, if the IPC classification unit 4017 detects a click on button 601 shown in FIG. 43, it determines that the user is selecting an IPC. In this case, the IPC classification unit 4017 confirms the IPC to be included in the resume based on the user's operation (step SA75) and terminates the processing based on this flowchart. The generation server 4000 registers the IPC confirmed in step SA75 in the work history database 4421 as part of the work history information. Therefore, the work history information includes the IPC. In this flowchart, steps SA72 to SA75 are an example of a collection unit that collects patent classification information related to the user's work based on information acquired from the user through dialogue. As in steps SA72 to SA75, the collection unit presents the user with multiple IPCs related to the user's work based on information acquired from the user through dialogue, and then confirms the IPC selected by the user as the user's work history information. This allows the user's qualifications, skills, etc. to be classified by IPC.

[0365] Next, several examples of resumes will be described with reference to Figures 46 to 49. Figures 46 to 49 are diagrams showing examples of resumes including IPC classifications. The resumes shown in Figures 46 to 49 include basic information, job content, and IPCs related to the job content. In particular, Figure 46 shows a resume format in which IPC items are listed separately from job content items. In contrast, Figure 47 shows a resume format in which the corresponding IPCs are incorporated into the description of job content.

[0366] A format may be adopted in which IPC items are listed separately from job content items, and the corresponding IPC is incorporated into the job content description. Figure 48 shows an example of a resume created using such a format.

[0367] If the user has patent applications for which the user is named as an inventor, the resume may include a patent section, as shown in Figure 49. The patent section may include the patent information along with the corresponding IPC. By listing the user's patent information along with the IPC in the resume, the user can more strongly demonstrate their expertise.

[0368] In this way, IPCs corresponding to the user's skills can be entered in the resume in various formats. The generation server 4000 stores information regarding job content, work experience, and employment period as work history information in the work history database 4421, linking the information with the IPC. When the generation server 4000 acquires information regarding the user's patents, as shown in FIG. 49, the generation server 4000 stores the patent information as work history information in the work history database 4421, linking the IPC with the patent information. Furthermore, the generation server 4000 stores the resumes illustrated in FIGS. 46 to 49 in the work history database 4421. The generation server 4000 may allow the user to select the style of the resume through a dialogue. Note that the dialogue is a concept that includes expressions of intent (such as questions and answers) exchanged between the generation server 4000 and the user.

[0369] Here, patent classification has been described as an example of format information containing the necessary information for a resume. Patent classification is not limited to IPC, and F-terms, FI-terms, and CPC (Cooperative Patent Classification) may also be used. Patent classification is an example of classification information for classifying a user's business capability information. As in the first embodiment, classification information used within a company may be used as classification information, or product classifications such as equipment may be applied. In other words, classification information may be any of IPC, F-terms, book classifications, and in-house classifications.

[0370] [Other Modifications] Fig. 50 is a diagram showing an example in which RAG (Retrieval-Augmented Generation) technology is utilized in the functional configuration of the generation server 4000. As already described, RAG is a technology in which a large-scale language model is configured to access a knowledge source containing the latest accurate information and the large-scale language model generates an answer based on the knowledge source. As is well known, large-scale language models may occasionally output inaccurate or misleading information, which is called hallucination. RAG can compensate for such imperfect operation of the large-scale language model and improve the quality of the answers generated by the large-scale language model.

[0371] To improve the accuracy of IPC selection, an IPC database 429 storing IPCs, patent information, etc. may be provided in the generation server 4000, as shown in Fig. 50. The large-scale language model 430 acquires IPCs from the IPC database 429, and identifies IPC candidates that are thought to correspond to the user's skills based on the acquired IPCs.

[0372] More specifically, the IPC database 429 is configured with the related information database 422 shown in Fig. 3. The IPC classification unit 4017 uses the large-scale language model 430 and a data source related to classification criteria stored in the IPC database 429, using the method already described with reference to Fig. 3 and other figures, to determine classification information corresponding to the user's skills.

[0373] This prevents the large-scale language model 430 from outputting an answer that corresponds to hallucination. As a result, the accuracy and reliability of the large-scale language model 430 can be improved. Furthermore, the system administrator can easily access the information source from which the large-scale language model 430 derived an answer. This allows the system administrator to easily determine the accuracy of the answer obtained from the large-scale language model 430.

[0374] In general, by adopting RAG technology, it is possible to create an interactive AI that can specialize in the desired function and obtain more precise answers. For example, by including internal company confidential information, it is possible to create an interactive AI that is intended for use exclusively within the company. It is also possible to input large amounts of information according to the purpose into the large-scale language model 430. Another advantage is that the type of large-scale language model 430 can be easily switched.

[0375] In general, in order to obtain appropriate answers from generative AI, it is considered important to improve the quality of prompts input by a user when interacting with the generative AI. Prompt engineering is known as a technique for improving the quality of such prompts. Prompt engineering is a technique for developing and optimizing prompts to be provided to a large-scale language model in order to efficiently use the large-scale language model. Prompt engineering is a method for optimizing output for a specific task, assuming the use of an existing model.

[0376] In contrast to this, there is the concept of fine-tuning. Fine-tuning is a technique for improving the performance of an existing model for a specific task by additional training. In fine-tuning, at least a portion of a trained model generated based on one dataset is additionally trained based on another dataset. This fine-tunes the parameters of the machine learning model for a specific task. In a broad sense, fine-tuning can also be interpreted as a type of transfer learning. However, the two differ in that fine-tuning is a technique for fine-tuning the weights of all layers of a trained model, while transfer learning is a technique for fixing the weights of a trained model and training using only the added layers.

[0377] Fine-tuning has the problem of requiring enormous computational resources because it requires additional training of a large-scale language model with many parameters. Prompt tuning solves this problem by taking a different approach from fine-tuning. In prompt tuning, the prompt itself is the learning target. In other words, in prompt tuning, the parameters corresponding to the prompt are the targets of optimization.

[0378] RLHF (Reinforcement Learning from Human Feedback) is a model learning method that combines "supervised learning," "reinforcement learning," and "inverse reinforcement learning." RLHF allows AI to learn complex tasks such as natural language processing while minimizing the need for human involvement, such as in supervised learning. Therefore, the large-scale language model 430 may be trained using such RLHF.

[0379] In the second embodiment, a large-scale language model is given as an example of a natural language processing algorithm. However, algorithms that can be adopted as natural language processing algorithms are not limited to large-scale language models. For example, instead of large-scale language models, algorithms generated by rule-based techniques such as pattern matching may be adopted.

[0380] The user devices 500 (recruiter device 200 and applicant device 300) may not only be equipped with all of the processor, memory, communication interface, and input / output interface shown in FIG. 21 , but may also be thin client systems using VDI (Virtual Desktop Infrastructure). A thin client system using VDI is a system in which a desktop environment on a server is transferred to a terminal in a remote location for use. The user devices 500 (recruiter device 200, applicant device 300), sharing server 100, and generation server 4000 do not necessarily have to be independent devices. When using a thin client system such as the one described above, the functions of the user device 500, sharing server 100, and generation server 4000 can be provided on the same aggregation server.

[0381] The databases 120 and 4420 are not limited to relational databases, and object-type or NoSQL-type databases may also be used.

[0382] Each of the sharing server 100 and the generation server 4000 is an example of a computing device. A computing device may be configured by a server (on-premise server, cloud server, etc.), a serverless system, etc. Here, an on-premise server is a server installed and managed in facilities managed within a company. A cloud server is a server (rented server) provided by another business operator via a network. A serverless system is a system in which computing and memory functions can be used only when necessary, without being aware of the existence of a server. Computing devices include servers and serverless systems. Servers include on-premise servers and cloud servers.

[0383] Examples of components included in the present disclosure are listed below: (a) The work history information collection device (generation server 4000, 4000A) includes an acquisition unit (step SA11) that acquires format information including required information for a resume, a collection unit (steps SA11 and SA22) that collects work history information related to the user's work history, and a memory unit (storage 4003) that stores a natural language processing algorithm, and the collection unit collects work history information corresponding to the required information by interacting with the user using the natural language processing algorithm stored in the memory unit (step SA22).

[0384] (b) In addition to the work history information collection device related to (a) above, the work history information collection device further includes a creation unit (step SA53) that creates a work history using the work history information collected by the collection unit, and when the collection unit has completed collecting the work history information related to the necessary information, the creation unit creates the work history in accordance with the format information using a natural language processing algorithm stored in the memory unit (step SA53).

[0385] (c) In addition to the work history information collection device related to (b) above, the work history information collection device further includes a work history database (work history database 4421) in which work history information collected before the dialogue by the collection unit is registered, and if the work history information registered in the work history database contains information corresponding to the required items, the creation unit creates a work history based on the work history information registered in the work history database (step SA13).

[0386] (d) The work history information collection device, in addition to the work history information collection device related to (b) or (c) above, further includes an output unit configured to output the work history created by the creation unit to a user device (step SA110, step SA115: the generation server 4000 may have the functions of the sharing server 100), and the user device is a recruiter device (200) operated by a recruiter or an applicant device (300) operated by an applicant in a matching system that matches recruiters who are recruiting contractors for work with applicants.

[0387] (e) In addition to the work history information collection device relating to any of (a) to (d) above, the work history information collection device further includes a dictionary database (dictionary database 4423), a registration unit (step SA34) that registers the meanings of terms included in the work history information in the dictionary database, and a display unit (step SA43) that displays the work history on a display device (user device 500), and if a term registered in the dictionary database is included in the work history, the display unit displays the meaning of the term on the display device (step SA44).

[0388] (f) In addition to the functions of the work history information collection device related to (e) above, the collection unit queries the user about the meaning of terms while interacting with the user (step SA32).

[0389] (g) The work history information collection device includes a work history information collection device related to any one of (a) to (f) above, and the natural language processing algorithm includes a large-scale language model (large-scale language models 430, 430A, 430B).

[0390] (h) In addition to being a work history information collection device relating to any of (a) to (g) above, the collection unit collects patent classification information related to the user's work based on information obtained from the user through dialogue (step SA75), and the work history information includes the patent classification information.

[0391] (i) In addition to the work history information collection device related to (h) above, the collection unit of the work history information collection device presents the user with multiple patent classification information related to the user's work based on information obtained from the user through dialogue, and then determines the patent classification information selected by the user as work history information (steps SA72 to SA75).

[0392] (j) A method for creating a work history information collection includes: having a computer perform the steps of obtaining format information containing required information for a resume; and collecting work history information related to a user's work history, wherein the collecting step includes collecting work history information corresponding to the required information by interacting with the user using a natural language processing algorithm.

[0393] [Aspects] Aspects of the present disclosure are listed below.

[0394] (Clause 1) The classification system described in clause 1 is a classification system that classifies information to be classified that is used by a user, and includes an acquisition unit that acquires the information to be classified, a classification unit that classifies the information to be classified in accordance with classification criteria, and a display device, wherein the classification unit includes a language processing model and a RAG data storage unit that stores a data source for causing the language processing model to perform language processing based on RAG (Retrieval-Augmented Generation), the RAG data storage unit includes a classification criteria data storage unit that stores a data source related to the classification criteria in a format that conforms to RAG, and the data source related to the classification criteria includes multiple pieces of classification information defined by the classification criteria, and the classification unit uses the language processing model and the data source related to the classification criteria to determine classification information corresponding to the information to be classified, and displays the information to be classified on the display device in a state where it is associated with the classification information.

[0395] (Section 2) In addition to the classification system described in Section 1, the classification system described in Section 2 has the following features: the RAG data storage unit includes a classification target data storage unit in which a data source regarding the information to be classified is stored in a format conforming to RAG; and the classification unit determines classification information corresponding to the information to be classified using a language processing model, a data source regarding classification criteria, and a data source regarding the information to be classified.

[0396] (Clause 3) In addition to the classification system described in clause 3, the classification system functions as a matching system that matches users with business cases, and the information to be classified includes one of business case information, user profile information, educational information for educating users, and advertising information provided to users.

[0397] (4) In addition to the classification system described in paragraph 3, the classification system described in paragraph 4 includes the profile information that includes the user's job ability information.

[0398] (Section 5) In addition to the classification system described in Section 3 or 4, the classification system described in Section 5 further includes a business case storage unit that stores business case information, a profile storage unit that stores profile information, a user device operated by a user, a search unit that searches for business case information stored in the business case storage unit using classification information associated with the profile information, and a transmission unit that transmits the business case information found by the search unit to the user device.

[0399] (Clause 6) In addition to the classification system described in Clause 3 or 4, the classification system described in Clause 6 further includes an advertisement storage unit that stores the advertisement information, a profile storage unit that stores the profile information, a user device operated by a user, a search unit that searches for advertisement information stored in the advertisement storage unit using classification information associated with the profile information, and a transmission unit that transmits advertisement information found by the search unit to the user device.

[0400] (Clause 7) In addition to the classification system described in claim 3 or claim 4, the classification system described in clause 7 further includes an education storage unit that stores the education information, a profile storage unit that stores the profile information, a user device operated by a user, a search unit that searches for the education information stored in the education storage unit using classification information associated with the profile information, and a transmission unit that transmits the education information found by the search unit to the user device.

[0401] (Section 8) The classification system described in Section 8 is, in addition to the classification system described in Section 7, characterized in that the search unit searches for educational information stored in the education storage unit in association with classification information different from the classification information associated with the user's profile information.

[0402] (Clause 9) The classification system described in clause 9 is, in addition to the classification system described in any one of clauses 1 to 8, characterized in that the classification unit generates summary information of the information to be classified, and determines classification information corresponding to the information to be classified based on the generated summary information, and the classification unit generates the summary information while interacting with the user.

[0403] (Clause 10) The classification system described in clause 10 is the same as the classification system described in any one of clauses 1 to 9, in that the classification unit determines classification information corresponding to the information to be classified while interacting with the user.

[0404] (Clause 11) The classification system described in clause 11 is a classification system described in any one of clauses 1 to 10, and in addition, the classification information is any of IPC, F-term, CPC, FI, book classification, and in-house classification.

[0405] (Clause 12) The method described in clause 12 is a method for classifying information to be classified that is used by a user, the method causing a computer to execute steps of acquiring the information to be classified and classifying the information to be classified in accordance with a classification criterion, the computer being configured to access a Retrieval-Augmented Generation (RAG) data storage unit, the RAG data storage unit including a classification criterion data storage unit in which a data source related to the classification criterion is stored in a format conforming to RAG, the data source related to the classification criterion including a plurality of pieces of classification information defined by the classification criterion, the classifying step including a step of determining classification information corresponding to the information to be classified using a language processing model and the data source related to the classification criterion, and a step of displaying the information to be classified on a display device in association with the classification information.

[0406] The embodiments disclosed herein should be considered to be illustrative in all respects and not restrictive. The scope of the present invention is defined by the claims, not by the description of the above embodiments, and is intended to include all modifications within the meaning and scope of the claims.

[0407] 1,10000 Classification system, 50 Internet, 100 Sharing server, 101 Processor, 102 Memory, 103 Storage, 104 Communication interface, 120 Database, 121 Company database, 122 User database, 123 Job database, 124 Matching database, 125 Advertisement database, 126 Education database, 127 Community database, 200 Recruiter device, 200A to 200C Recruiter device, 201 Processor, 202 Memory, 203 Communication interface, 204 Input / output interface, 205 Display, 206 Operation unit, 300 Applicant device, 300A to 300C Applicant device, 301 Processor, 302 Memory, 303 Communication interface, 304 Input / output interface, 305 Display, 306 Operation unit, 400 Server, 401 Processor, 402 Memory, 403 Communication interface, 404 Storage, 410 Microcomputer, 420 RAG database, 421 Related information database, 422 Related information database, 423 Related information database, 424 Related information database, 425 Related information database, 426 Satisfying requirement database, 427 Original information storage database, 428 Generative format database, 429 IPC database, 430, 430A, 430B Large scale language model (LLM), 450 AI, 500, 500A User device, 501 Processor, 502 Memory, 503 Communication interface, 504 Input / output interface, 505 Display, 506 Operation unit, 550 Screen, 601 to 603 Buttons, 800A Profile classification unit, 800B Case classification unit, 803A, 803B Chat unit (summary), 804A, 804B Chat unit (classification), 805A, 805B classification storage unit, 902 AI construction unit, 903 data import unit, 904 acquisition unit, 905 matching unit, 906 result display unit, 1550 screen, 1551 question frame, 1552 answer frame, 4000, 4000A generation server, 4001 processor, 4002 memory, 4003 storage, 4004 communication interface, 4010,4010A control unit, 4011 data import unit, 4012 information acquisition unit, 4013 dictionary information registration unit, 4014 dictionary information display unit, 4015 information organization unit, 4016 resume registration unit, 4017 IPC classification unit, 4021 input unit, 4022 content sufficiency determination unit, 4023 input request unit, 4024 original information storage unit, 4025 generation unit, 4026 confirmation request unit, 4027 output unit, 4050 program, 4051 dialogue program, 4052 confirmation program, 4053 creation program, 4110 dialogue AI, 4120 confirmation AI, 4130 creation AI, 4420 database, 4421 work history database, 4422 dialogue information database, 4423 dictionary database.

Claims

1. A classification system that classifies information to be classified that is used by a user, comprising: an acquisition unit that acquires the information to be classified; a classification unit that classifies the information to be classified in accordance with a classification criterion; and a display device, wherein the classification unit includes a language processing model; and a RAG data storage unit in which a data source for causing the language processing model to perform language processing based on RAG (Retrieval-Augmented Generation), the RAG data storage unit includes a classification criterion data storage unit in which a data source related to the classification criterion is stored in a format that conforms to the RAG, and the data source related to the classification criterion includes a plurality of pieces of classification information defined by the classification criterion, and the classification unit uses the language processing model and the data source related to the classification criterion to determine classification information corresponding to the information to be classified, and displays the information to be classified on the display device in a state in which it is associated with the classification information.

2. The classification system of claim 1, wherein the RAG data storage unit includes a classification target data storage unit in which a data source relating to the classification target information is stored in a format conforming to the RAG, and the classification unit determines classification information corresponding to the classification target information using the language processing model, a data source relating to the classification criteria, and a data source relating to the classification target information.

3. The classification system of claim 1 or claim 2, wherein the classification system functions as a matching system that matches the user with business cases, and the classification target information includes one of business case information, profile information of the user, educational information for educating the user, and advertising information provided to the user.

4. The classification system of claim 3, wherein the profile information includes job performance information of the user.

5. The classification system according to claim 3 or claim 4, wherein the information to be classified includes the business case information and the profile information, and the classification system further comprises a business case storage unit that stores the business case information, a profile storage unit that stores the profile information, a user device operated by the user, a search unit that searches for business case information stored in the business case storage unit using classification information associated with the profile information, and a transmission unit that transmits the business case information found by the search unit to the user device.

6. The classification system of claim 3 or claim 4, wherein the information to be classified includes the advertising information and the profile information, and the classification system further comprises an advertising storage unit that stores the advertising information, a profile storage unit that stores the profile information, a user device operated by the user, a search unit that searches for advertising information stored in the advertising storage unit using classification information corresponding to the profile information, and a transmission unit that transmits the advertising information found by the search unit to the user device.

7. The classification system of claim 3 or claim 4, wherein the information to be classified includes the educational information and the profile information, and the classification system further comprises an education storage unit that stores the educational information, a profile storage unit that stores the profile information, a user device operated by the user, a search unit that searches for educational information stored in the education storage unit using classification information corresponding to the profile information, and a transmission unit that transmits the educational information found by the search unit to the user device.

8. The classification system according to claim 7, wherein the search unit searches for education information stored in the education storage unit in association with classification information different from classification information associated with the user's profile information.

9. A classification system as described in any one of claims 1 to 8, wherein the classification unit generates summary information of the information to be classified, and determines classification information corresponding to the information to be classified based on the generated summary information, and the classification unit generates the summary information while interacting with the user.

10. A classification system according to any one of claims 1 to 9, wherein the classification unit determines classification information corresponding to the classification target information while interacting with the user.

11. A classification system according to any one of claims 1 to 10, wherein the classification information is any one of IPC, F-term, CPC, FI, book classification, and in-house classification.

12. A method for classifying information to be classified used by a user, the method comprising: causing a computer to execute a step of acquiring the information to be classified; and a step of classifying the information to be classified in accordance with a classification criterion; the computer is configured to access a Retrieval-Augmented Generation (RAG) data storage unit; the RAG data storage unit includes a classification criterion data storage unit in which a data source related to the classification criterion is stored in a format conforming to the RAG; the data source related to the classification criterion includes a plurality of classification information defined by the classification criterion; and the classifying step includes a step of determining classification information corresponding to the information to be classified using a language processing model and the data source related to the classification criterion; and a step of displaying the information to be classified on a display device in association with the classification information.

Images