Information processing systems, information processing methods, and programs

Abstract

This invention provides an information processing system, method, and program for efficiently matching employers with job seekers. [Solution] In the information processing system, the recruiter agent construction unit of the server device constructs a recruiter task execution model for each recruiter using machine learning with recruiter learning data. The recruiter learning data includes information about recruiters registered in the recruiter database, information provided by recruiters, or publicly available information about recruiters on the network. The job seeker agent construction unit constructs a job seeker task execution model for each job seeker that is capable of executing tasks that handle job seeker information using machine learning with job seeker learning data. The job seeker learning data includes information about job seekers registered in the job seeker database, information provided by job seekers, or publicly available information about job seekers on the network. The task execution unit causes the recruiter task execution model or the job seeker task execution model to execute a task based on the task execution instruction.

Description

Technical Field

[0001] The present invention relates to an information processing system, an information processing method, and a program.

Background Art

[0002] Patent Document 1 discloses a technique for matching job seeker information with job requirements.

Prior Art Document

Patent Document

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] There is a need for a technique that can more efficiently match job providers and job seekers.

[0005] In view of the above circumstances, the present invention aims to provide an information processing system and the like that can efficiently match job providers and job seekers.

Means for Solving the Problems

[0006] According to one aspect of the present invention, an information processing system is provided, comprising at least one processor, the processor configured to perform the following steps by reading a program, wherein in the recruiter agent construction step, a recruiter task execution model capable of performing tasks that handle recruiter information is constructed for each recruiter by machine learning using recruiter learning data relating to recruiters, the recruiter learning data includes recruiter information registered in a recruiter database, information provided by recruiters, or publicly available information about recruiters on a network; in the job seeker agent construction step, a job seeker task execution model capable of performing tasks that handle job seeker information is constructed for each job seeker by machine learning using job seeker learning data relating to job seekers, the job seeker learning data includes job seeker information registered in a job seeker database, information provided by job seekers, or publicly available information about job seekers on a network; and in the task execution step, the information processing system is provided to cause the recruiter task execution model or the job seeker task execution model to perform a task based on an instruction to perform the task.

[0007] In this configuration, employers and job seekers can be efficiently matched through task execution models that are individually constructed for each party. [Brief explanation of the drawing]

[0008] [Figure 1] This is a diagram showing the configuration of Information Processing System 1. [Figure 2] This is a block diagram showing the hardware configuration of server device 10. [Figure 3] This block diagram shows the hardware configuration of the job seeker terminal 20 and the job applicant terminal 30. [Figure 4] This is a block diagram showing the functions realized by the server device 10 (control unit 11), the job seeker terminal 20 (control unit 21), and the job seeker terminal 30 (control unit 31). [Figure 5]This diagram illustrates the matching process between employers and job seekers using employer task execution models and job seeker task execution models. [Figure 6] This is an activity diagram showing an example of the flow of information processing (task execution processing) performed by Information Processing System 1. [Modes for carrying out the invention]

[0009] Embodiments of the present invention will be described below with reference to the drawings. The various features shown in the embodiments below can be combined with each other.

[0010] Incidentally, the program for implementing the software appearing in one embodiment may be provided as a non-transitory computer-readable medium, or it may be provided as a downloadable medium from an external server, or it may be provided so that the program is launched on an external computer and its functions are realized on a client terminal (so-called cloud computing).

[0011] Furthermore, in various information processing according to one embodiment, an input and an output corresponding to the input can be realized. Here, as long as an output is obtained as a result of the input, the form of the information referenced in such information processing (hereinafter referred to as "reference information") is not limited. The reference information may be, for example, rule-based information such as a database, a lookup table, or a predetermined function (including a decision formula such as a regression equation constructed by a statistical method), or a pre-trained model that has learned the correlation between input and output in advance, or a large-scale language model that can output a desired result by inputting a prompt.

[0012] Furthermore, in one embodiment, "part" may include, for example, hardware resources implemented by a circuit in a broad sense, and the information processing of software that can be specifically realized by these hardware resources. Also, in one embodiment, various types of information are handled, and this information can be represented, for example, by the physical values ​​of signal values ​​representing voltage and current, the high or low values ​​of signal values ​​as a set of binary bits composed of 0s or 1s, or by quantum superposition (so-called qubits), and communication and calculations can be performed on a circuit in a broad sense.

[0013] Furthermore, a circuit in a broad sense is a circuit realized by combining at least a suitable combination of circuits, circuits, processors, and memory. The processor may be a general-purpose processor or a dedicated circuit. In other words, it includes application-specific integrated circuits (ASICs), programmable logic devices (for example, simple programmable logic devices (SPLDs), complex programmable logic devices (CPLDs), and field programmable gate arrays (FPGAs)), etc.

[0014] 1. Hardware Configuration This section describes the hardware configuration.

[0015] <Information Processing System 1> Figure 1 is a configuration diagram representing information processing system 1. Information processing system 1 comprises a communication line 2, a server device 10, multiple employer terminals 20, and multiple job seeker terminals 30. The server device 10, employer terminals 20, and job seeker terminals 30 are configured to communicate with each other via the communication line 2. The connection between the server device 10, employer terminals 20, and job seeker terminals 30 may be wired or wireless.

[0016] The information processing system 1 constitutes at least a part of a job offer and job seeking system used by, for example, a plurality of job offerers (the first job offerer U1 and the second job offerer U2) and a plurality of job seekers (the first job seeker U3 and the second job seeker U4). The information processing system 1 mainly performs job seeker search by job offerers, job offer search by job seekers, mediation of communication between job offerers and job seekers, etc. In one embodiment, the information processing system 1 consists of one or more devices or components. Hereinafter, these components will be described.

[0017] <Server device 10> FIG. 2 is a block diagram showing the hardware configuration of the server device 10. As shown in FIG. 2, the server device 10 includes a control unit 11, a storage unit 12, a communication unit 13, and a communication bus 14. The control unit 11, the storage unit 12, and the communication unit 13 are electrically connected inside the server device 10 via the communication bus 14.

[0018] <Control unit 11> The control unit 11 performs processing and control of the overall operation related to the server device 10. The control unit 11 is, for example, a Central Processing Unit (CPU). The control unit 11 realizes various functions related to the server device 10 by reading a predetermined program stored in the storage unit 12. That is, the information processing by software stored in the storage unit 12 is specifically realized by the control unit 11, which is an example of hardware, and can be executed as each functional unit included in the control unit 11. These will be described in more detail in the next section. Note that the control unit 11 is not limited to being single, and the server device 10 may have a plurality of control units 11 for each function. Also, the server device 10 may be configured by a combination of these.

[0019] <Storage unit 12> The memory unit 12 stores various information defined as described above. This can be implemented, for example, as a storage device such as a Solid State Drive (SSD) that stores various programs and the like related to the server device 10 executed by the control unit 11, or as a memory such as a Random Access Memory (RAM) that stores temporarily necessary information (arguments, arrays, etc.) related to the operation of the program. The memory unit 12 stores various programs, variables, etc. related to the server device 10 executed by the control unit 11.

[0020] <Communication unit 13> Although wired communication means such as USB, IEEE1394, Thunderbolt (registered trademark), and wired LAN network communication are preferred for the communication unit 13, wireless LAN network communication, mobile communication such as LTE / 5G, and BLUETOOTH (registered trademark) communication may be included as needed. That is, it is more preferable to implement it as a collection of these multiple communication means. That is, the server device 10 may communicate various information from the outside via the communication unit 13 and the network.

[0021] The server device 10 may be in an on-premises form or in a cloud form. The server device 10 in the cloud form may provide the above-mentioned functions and processes, for example, in the form of Software as a Service (SaaS) or cloud computing.

[0022] <Job seeker terminal 20> FIG. 3 is a block diagram showing the hardware configuration of the job seeker terminal 20 and the job applicant terminal 30. The job seeker terminal 20 is an information processing terminal used by job seekers. Job seekers include organizations such as profit-making corporations (e.g., companies, etc.), non-profit corporations (e.g., cooperatives, foundations, etc.), public corporations (e.g., local governments, etc.) or their responsible persons. Also, job seekers include personnel intermediaries who mediate between job applicants and organizations as agents of the organization. Personnel intermediaries are also called headhunters, agents, etc.

[0023] As shown in Figure 3A, the job seeker terminal 20 comprises a control unit 21, a storage unit 22, a communication unit 23, an input unit 24, an output unit 25, and a communication bus 26. The control unit 21, storage unit 22, communication unit 23, input unit 24, and output unit 25 are electrically connected within the job seeker terminal 20 via the communication bus 26. The descriptions of the control unit 21, storage unit 22, and communication unit 23 are the same as the descriptions of each part in the server device 10 and are therefore omitted.

[0024] <Input section 24> The input unit 24 receives operation inputs made by the user. The operation inputs are transmitted as command signals to the control unit 21 via the communication bus 26. The control unit 21 can perform predetermined controls or calculations based on the transmitted command signals as needed. The input unit 24 may be included in the housing of the job seeker terminal 20 or it may be an external component. For example, the input unit 24 may be implemented as a touch panel integrated with the output unit 25. When the input unit 24 is implemented as a touch panel, the user can input tap operations, swipe operations, etc. to the input unit 24. Instead of a touch panel, the input unit 24 can be a switch button, mouse, trackpad, QWERTY keyboard, etc.

[0025] <Output section 25> The output unit 25 displays a graphical user interface (GUI) screen that can be operated by the user. The output unit 25 may be included in the housing of the job seeker terminal 20 or it may be an external device. Specifically, the output unit 25 can be implemented as a display device such as a CRT display, liquid crystal display, organic EL display, or plasma display. It is preferable that these display devices be used according to the type of job seeker terminal 20.

[0026] <Job seeker terminal 30> The job seeker terminal 30 is an information processing terminal used by job seekers. "Job seekers" include, for example, currently employed individuals (those seeking a career change), prospective graduates (job seekers), and students.

[0027] As shown in Figure 3B, the job seeker terminal 30 comprises a control unit 31, a storage unit 32, a communication unit 33, an input unit 34, an output unit 35, and a communication bus 36. The control unit 31, storage unit 32, communication unit 33, input unit 34, and output unit 35 are electrically connected within the job seeker terminal 30 via the communication bus 36. The descriptions of the control unit 31, storage unit 32, communication unit 33, input unit 34, and output unit 35 are the same as the descriptions of each part in the employer terminal 20 and are therefore omitted.

[0028] 2. Functional Configuration This section describes the functional configuration of this embodiment. Information processing by software stored in the memory unit 12 is specifically realized by the control unit 11, which is an example of hardware, and can be executed as each functional unit included in the control unit 11 (at least one processor provided by the information processing system 1).

[0029] Figure 4 is a block diagram showing the functions realized by the server device 10 (control unit 11), the job seeker terminal 20 (control unit 21), and the job seeker terminal 30 (control unit 31).

[0030] As shown in Figure 4A, the server device 10 (control unit 11) comprises a basic display control unit 111, a recruiter agent construction unit 112, a recruiter label creation unit 113, a job seeker agent construction unit 114, a job seeker label creation unit 115, a question registration unit 116, a platform agent construction unit 117, a task execution unit 118, and an artificial intelligence unit 120.

[0031] As shown in Figure 4B, the job seeker terminal 20 (control unit 21) includes a display unit 211 and an operation acquisition unit 212. As shown in Figure 4C, the job seeker terminal 30 (control unit 31) includes a display unit 311 and an operation acquisition unit 312.

[0032] <Basic display control unit 111> The basic display control unit 111 is configured to display various information on the employer terminal 20 or the job seeker terminal 30. For example, the basic display control unit 111 displays information about job seekers or employers registered in the database on the display unit 211 of the employer terminal 20 or the display unit 311 of the job seeker terminal 30, in response to requests from each user (employers U1, U2 or job seekers U3, U4).

[0033] <Recruiter Agent Development Department 112> The recruiter agent construction unit 112 is configured to construct a recruiter task execution model capable of performing tasks that handle recruiter information, for each recruiter registered in the recruiter database, using machine learning with recruiter learning data about recruiters. In this specification, "construction" of a model is not limited to creating a new model, but also includes updating an existing model.

[0034] The recruiter task execution model may be an output device that outputs the results of executing an input task, or it may be a so-called AI (artificial intelligence) agent (recruiter agent) that autonomously performs tasks such as data collection and conversation necessary for executing the tasks or goals input by the recruiter, using processing models built with natural language processing, machine learning, or general-purpose large-scale language models. The AI ​​agent may also be called an autonomous agent. Conversation with the recruiter task execution model takes place, for example, in the form of voice or text chat.

[0035] An "AI agent" is a model that, upon input of a goal (objective, purpose, etc.) such as "Teach me about XX" or a task such as "Produce the output of XX," breaks down the processes necessary to reach the goal or accomplish the task into subtasks, actions, etc., and performs necessary data collection and analysis, program generation and execution, etc. The AI ​​agent autonomously selects and executes tasks, actions, etc., and does not require user intervention (operation input). Furthermore, the AI ​​agent may autonomously plan and execute, and evaluate the execution results itself, thereby autonomously learning to aim for the achievement of the goal. For example, the AI ​​agent may autonomously update based on the execution results of subtasks (e.g., collected information, results of information analysis, etc.). The AI ​​agent may include a large-scale language model that includes a generative AI trained to perform these processes.

[0036] Specifically, an AI agent may perform a task in the following manner, for example. First, the AI ​​agent receives specific instructions from the user, including a goal or task. Next, the AI ​​agent divides the goal or task into multiple subtasks or actions necessary to achieve the goal or task, and generates subtasks or actions. After dividing the goal or task, the AI ​​agent collects information necessary to perform the multiple subtasks and actions (e.g., employer registration information, job seeker registration information, etc.). For example, the AI ​​agent collects information by accessing databases, the internet, web searches, API integration with external services, conversations with other AI agents, or referencing employer memory or job seeker memory as described later. After collecting the information, the AI ​​agent executes the multiple subtasks and actions based on the collected information (e.g., by generating and executing code), for example by using external tools, and outputs the task corresponding to the goal, or the execution result of the input task.

[0037] Furthermore, the AI ​​agent may verify or evaluate the results of a subtask or action itself, and if it determines that revision is necessary, it may modify or improve the subtask or action and re-execute it. For example, after generating code to perform a specific subtask, the AI ​​agent may verify the code itself and provide feedback to modify the code.

[0038] The tasks performed by the recruiter task execution model (hereinafter referred to as "recruiter tasks") may include, for example, answering questions about the recruiter (the source of the recruiter training data) to which the recruiter task execution model corresponds. In other words, the recruiter task execution model may be capable of answering questions about recruiters as part of the recruiter task. Furthermore, the recruiter task execution model may be capable of asking questions to job seekers or other AI agents (job seeker task execution model or platform task execution model) as part of the recruiter task or its subtasks. This allows the recruiter task execution model to conduct conversations or interviews with job seekers or other AI agents. As a result, for example, it becomes possible for the recruiter task execution model to conduct simulated interviews with highly popular and difficult-to-contact recruiters.

[0039] "Questions about the recruiter" include, for example, questions about the organization's performance, goals, structure, job descriptions, culture or atmosphere, job requirements, desired candidate profile, and recruitment process. In addition, questions that the recruiter task execution model itself asks to job seekers or other AI agents include questions to confirm the suitability (match) between the organization's goals, culture, job requirements, etc., and the person being asked (the job seeker).

[0040] The recruiter task may include output of information about recruiters specified by the task directive (for example, the task execution unit 118 described later). Alternatively, the recruiter task may also include information indicating a specific goal (objective, purpose, etc.) related to the recruiter, as specified by (for example, the task execution unit 118 described later). Such recruiter tasks may include instructions such as "present job seekers to recommend to recruiters" or information indicating a goal such as "I want to know which job seekers are recommended," and the information output by the recruiter task execution model may include, for example, information about job seekers recommended to recruiters.

[0041] "Job seekers recommended to employers" are, for example, job seekers who have a high degree of match with the employer or the job information the employer has, job seekers who are presumed to be likely to reply to a recruitment email sent by the employer, or job seekers who are presumed to be likely to be hired. The employer task execution model may determine which job seekers to recommend to employers by, for example, comparing the registration information of the job seeker to be judged, the job seeker memory described below, etc., with trained information about employers (for example, the employer memory described below, etc.) or information about job postings (for example, converting each piece of information into a vector and determining the similarity based on the cosine similarity between the vectors, etc.).

[0042] The recruiter training data used in the recruiter agent construction unit 112 includes information about recruiters registered in the recruiter database (registered information), information provided by recruiters (provided information), or publicly available information about recruiters on the network.

[0043] The "employer database" is a database used in the platform provided by Information Processing System 1 for matching employers with job seekers. In other words, the employer training data used in the employer agent construction unit 112 may include information on employers registered in the employer database on a single platform.

[0044] The recruiter registration information (hereinafter referred to as "recruiter registration information") includes, for example, information about the recruiter themselves (if the recruiter is an organization), the organization to which the recruiter belongs (if the recruiter is a representative of the organization), or the organization on which the recruiter acts as an agent (if the recruiter is a recruitment agency), such as information indicating the business content, industry, organizational size, performance, products or services handled, financial status, and management plan. In addition, the recruiter registration information may include information about the entire organization (e.g., the business content, industry, performance of the entire organization), or it may include information about a part of the organization (e.g., a business division, department to which the recruiter belongs) (e.g., the business content, industry, performance of a single business division).

[0045] Furthermore, the recruiter registration information may include data on job postings created by the organization, data on job seekers approached by the recruiter (e.g., sending recruitment emails, conducting job interviews, and successful job placements), data from the applicant tracking system (ATS) (e.g., attributes or characteristics of personnel hired by the organization), and personnel evaluation data within the organization (e.g., attributes or characteristics of highly-rated personnel within the organization).

[0046] Thus, the recruiter learning data may include at least one of the following as information about recruiters registered in the recruiter database: data on recruiters' talent acquisition used in the recruitment management system, and data on recruiters' evaluations of personnel within the organization used in the human resources system. This makes it possible to construct a recruiter task execution model that can perform tasks based on the attributes or characteristics of personnel working in the organization.

[0047] Information provided by recruiters includes, for example, information created or acquired by recruiters during recruitment activities (e.g., organizational explanatory videos, scripts for information sessions, audio recordings of recruiters during information sessions, interviews, or meetings, items to be emphasized in interviews or meetings, questions asked during interviews, requirements verbally explained by recruiters, and verbal responses from job seekers). The recruiter agent building unit 112 acquires the provided information by, for example, accepting input, uploads, and specification of storage locations on the network from the recruiter terminal 20. The recruiter agent building unit 112 may also provide recruiters with incentives for inputting information, such as granting them benefits that can be used on the platform.

[0048] "Public information" refers to information that is publicly available on websites accessible via the internet. The types of websites from which a job seeker's public information can be obtained are not limited to those that contain at least some information about the job seeker (organization), and include official websites, unofficial websites, and other websites.

[0049] An official website is a website established by the recruiting organization itself, and examples include corporate websites, investor relations (IR) websites, product or service websites, brand websites, promotional websites, owned media websites, campaign websites, e-commerce websites, member websites, recruitment websites, and landing pages. An unofficial website is a website established by an organization other than the recruiting organization (an external organization), and examples include industry information websites, news websites, economic information websites, and commercial database websites. Unofficial websites include both free and paid access pages. Publicly available information from recruiters may include, for example, interviews with the organization's employees.

[0050] The recruiter agent building unit 112 may automatically collect publicly available information by keyword searching using organization names, crawling, or web scraping on the internet. Alternatively, the recruiter agent building unit 112 may accept access information (e.g., URL, IP address, etc.) indicating the network location of the publicly available information from the recruiter terminal 20 and obtain the publicly available information based on the entered access information.

[0051] The recruiter task execution model is a learning model included in the artificial intelligence unit 120 that has been machine-trained to take a recruiter task as input and output the result of executing the recruiter task. The recruiter task execution model is a learning model that has been machine-trained using recruiter training data and data necessary for executing the recruiter task (for example, a combination of input and output examples for the recruiter task) as training data.

[0052] The training data for the job seeker task execution model may further include statistical information on job seekers on the platform provided by Information Processing System 1 (e.g., the number of applications for each job by industry, occupation, etc.).

[0053] The recruiter task execution model may include a large-scale language model, including a generative AI. If the recruiter task execution model is a large-scale language model, the recruiter agent construction unit 112 may create at least one recruiter memory for each recruiter from the recruiter training data, which the recruiter task execution model will refer to when generating output. This allows the recruiter task execution model to be constructed using an existing large-scale language model. It also facilitates updating the recruiter task execution model.

[0054] Even when the recruiter task execution model is an AI agent, the recruiter agent construction unit 112 may create at least one recruiter memory for each recruiter from the recruiter training data, which the recruiter task execution model will refer to when generating output. This allows for the construction of a recruiter task execution model tailored to the characteristics of each individual recruiter, using a pre-built AI agent.

[0055] The recruiter memory is information representing the characteristics or attributes of a recruiter, and it is a parameter that customizes (determines the characteristics of) the recruiter task execution model. The recruiter task execution model executes the recruiter task as a recruiter agent that has learned the characteristics represented in the recruiter memory. The recruiter memory is registered, for example, in the recruiter database, linked to the recruiter registration information.

[0056] The recruiter memory may include at least some information about the culture of the organization making the job. This allows for the construction of a recruiter task execution model that can respond while considering the match between the recruiter and the organizational culture.

[0057] The recruiter memory includes information such as the organization's growth rate, growing areas, organizational policies, business advantages, and the type of personnel the organization seeks (e.g., personality and characteristics). Specific examples of recruiter memory content include "a rapidly growing HR company," "a venture company," and "we want to hire honest and diligent people." The recruiter memory can be information written in natural language, or code information such as computer languages, programming languages, or machine code, as long as it can be referenced by the recruiter task execution model. The number of recruiter memories prepared for one recruiter task execution model (one recruiter) may be on the order of 1000, for example.

[0058] The recruiter agent building unit 112, for example, inputs recruiter training data into a memory creation model and causes the memory creation model to output recruiter memory. The memory creation model is a learning model included in the artificial intelligence unit 120 that has been machine-trained to take recruiter training data as input and output recruiter memory. The memory creation model is a learning model that has been machine-trained using recruiter training data and corresponding recruiter memory as training data.

[0059] The memory creation model may be a large-scale language model that includes a generative AI. In this case, the recruiter agent construction unit 112 takes recruiter training data as input, inputs a prompt to the memory creation model that includes an instruction to output a recruiter memory corresponding to the recruiter training data, and causes the memory creation model to output the recruiter memory. The recruiter agent construction unit 112 may also generate a prompt that gives the memory creation model an instruction to output a recruiter memory corresponding to the recruiter training data, and input this prompt to the memory creation model. In addition to the instructions for generating and outputting recruiter memories and the recruiter training data, the recruiter agent construction unit 112 may also input a prompt to the memory creation model that includes, for example, one or more samples of recruiter training data and one or more samples of corresponding recruiter memories as examples, samples, or training data of input and output pairs.

[0060] The recruiter agent construction unit 112 may cause the memory generation model, which is a large-scale language model, to output recruiter memory by inputting input information including input to the recruiter task execution model, output from the recruiter task execution model for said input, or a combination thereof, and an instruction to create recruiter memory representing the characteristics or attributes of recruiters based on said input information. This allows for the addition or updating of recruiter memory based on the execution results of tasks in the recruiter task execution model, thereby improving the execution accuracy of tasks in the recruiter task execution model.

[0061] The recruiter agent construction unit 112 may present the recruiter with questions to create a recruiter memory and create the recruiter memory based on the answers the recruiter provides to those questions. For example, the recruiter agent construction unit 112 may input the recruiter's answers into a memory generation model and have the memory generation model output the recruiter memory. Alternatively, the recruiter agent construction unit 112 may use the results of creating the recruiter memory, which is a subtask executed by the recruiter task execution model, an AI agent, when performing a recruiter task, such as extracting job seekers to recommend to the recruiter.

[0062] The recruiter agent construction unit 112 may create a recruiter memory representing the characteristics or attributes of each recruiter from the recruiter training data, and may also construct a recruiter task execution model using machine learning based on the recruiter memory. This allows for the construction of a recruiter task execution model specific to each individual recruiter based on the recruiter memory. The recruiter task execution model constructed using the recruiter memory may be a dedicated model that has been machine-trained to take a task as input and output the execution result of the task, or it may be a large-scale language model (specifically, for example, a general-purpose large-scale language model that has undergone reinforcement learning or fine-tuning using the recruiter training data and the recruiter memory).

[0063] The recruiter agent building unit 112 may build a recruiter task execution model using machine learning with recruiter labels created by the recruiter label creation unit 113 (described later) and edited by the recruiter. This allows recruiters to customize the recruiter task execution model and modify its characteristics, thereby improving the accuracy of matching between recruiters and job seekers using the recruiter task execution model.

[0064] The recruiter agent building unit 112 may build a recruiter task execution model by machine learning using further recorded data of interviews between recruitment agencies (headhunters) and job seekers or employers. This allows the recruiter task execution model to learn the know-how of recruitment agencies regarding interviews, matching, etc., thereby improving the accuracy of matching between job seekers and employers using the recruiter task execution model.

[0065] Interviews between recruitment agencies and employers may include, for example, an exchange of information about the desired candidate profile. Interviews between recruitment agencies and job seekers may include consultations not only about recruitment but also about the job seeker's career, life, and comparisons with other job seekers (their relative position). "Interview record data" may include, for example, video, audio, and text data.

[0066] The recruiter agent building unit 112 may prepare training data for each recruiter's department, job type, or job that the recruiter handles, and build a recruiter task execution model for each department, job type, or job. This allows recruiter task execution models to be built at the department, job type, or job level in large organizations, enabling the recruiter task execution models to perform more detailed recruiter tasks.

[0067] <Job Posting Label Creation Section 113> The recruiter label creation unit 113 is configured to output at least one recruiter label by inputting a recruiter memory and an instruction to generate a recruiter label that represents the contents of the recruiter memory in natural language to the memory generation model.

[0068] A job seeker label represents the content of an individual job seeker memory using natural language (words or sentences) that job seekers can understand. Note that multiple job seeker labels may be generated from a single job seeker memory. Job seeker labels are registered, for example, in the job seeker database, linked to the job seeker registration information.

[0069] The recruiter label creation unit 113, for example, takes a recruiter memory as input and inputs a prompt to the memory creation model that includes an instruction to output a recruiter label corresponding to the said recruiter memory, causing the memory creation model to output the recruiter label. The recruiter label creation unit 113 may also generate a prompt that gives the memory creation model an instruction to output a recruiter label corresponding to the recruiter memory, and input this prompt to the memory creation model. In addition to the instructions for generating and outputting recruiter labels and the recruiter memory, the recruiter label creation unit 113 may also input a prompt to the memory creation model that includes, for example, one or more samples of recruiter memories and one or more samples of corresponding recruiter labels as examples, samples, or training data of input and output pairs.

[0070] The recruiter label creation unit 113 may acquire recruiter labels created by the recruiter task execution model, which is an AI agent. The AI ​​agent may perform the generation of recruiter labels as a subtask when executing a recruiter task, such as extracting job seekers to recommend to recruiters. The AI ​​agent may also perform the generation of recruiter memory and recruiter labels as subtasks. Furthermore, the recruiter label creation unit 113 may cause the recruiter task execution model to create a prompt for the memory creation model to create recruiter labels, and input this prompt to the memory creation model.

[0071] The recruiter label creation unit 113 presents all or part of the multiple recruiter labels generated for each recruiter to the corresponding recruiter (recruiter terminal 20). This allows the recruiter to confirm the components (features or attributes) of their recruiter task execution model (AI agent).

[0072] For example, the recruiter label creation unit 113 may determine whether multiple recruiter labels contain positive or negative content, and present only the recruiter labels determined to contain positive content to the recruiter terminal 20. Alternatively, the recruiter label creation unit 113 may have a large-scale language model, including a generation AI, determine whether a recruiter label contains positive or negative content. Furthermore, the recruiter label creation unit 113 may utilize the results of a subtask performed by the recruiter task execution model, which is an AI agent, when performing recruiter tasks such as extracting job seekers to recommend to recruiters, specifically the determination of whether a recruiter label contains positive or negative content.

[0073] Furthermore, the employer label creation unit 113 may present to the employer terminal 20 employer labels that have a high degree of relevance to items that are important in matching with job seekers, and to items that employers want to appeal to job seekers (for example, employer labels ranked from highest to lowest relevance). These items may be extracted from employer learning data (employer registration information), for example, or entered by the employer. In addition, the employer label creation unit 113 may present to the employer terminal 20 an employer label summary, which summarizes combinations of several employer labels.

[0074] The recruiter label creation unit 113 may accept edits of recruiter labels from the corresponding recruiter (recruiter terminal 20). This allows for customization of the recruiter task execution model based on edits made by the recruiter to the recruiter label. Editing also includes deleting recruiter labels.

[0075] <Job Seeker Agent Development Department 114> The job seeker agent building unit 114 is configured to build a job seeker task execution model capable of performing tasks that handle job seeker information, for each job seeker registered in the job seeker database, using machine learning with job seeker learning data about the job seeker.

[0076] The job seeker task execution model may be an output device that outputs the results of executing an input task, or it may be a so-called AI agent (job seeker agent) that autonomously performs subtasks such as data collection and conversation necessary for executing the task or goal entered by the job seeker, using processing models built with natural language processing, machine learning, or general-purpose large-scale language models. Conversation with the job seeker task execution model may take place, for example, in the form of voice or text chat.

[0077] The tasks performed by the job seeker task execution model (hereinafter referred to as "job seeker tasks") may include, for example, answering questions about the job seeker (the source of the job seeker training data) to which the job seeker task execution model corresponds. In other words, the job seeker task execution model may be capable of answering questions about the job seeker as a job seeker task. Furthermore, the job seeker task execution model may be capable of asking questions to employers or other AI agents (employer task execution model or platform task execution model) as a job seeker task or its subtask. This allows the job seeker task execution model to engage in conversations or interviews with employers or other AI agents. As a result, for example, it becomes possible for the job seeker task execution model to conduct simulated interviews with popular job seekers who are difficult to contact.

[0078] "Questions about the job seeker" include, for example, questions about desired conditions (industry, job type, duties, annual salary, working conditions, etc.), past experience (industry, job type, duties, position, skills, qualifications, etc.), selling points, career plan, personality, values, etc. In addition, questions that the job seeker task execution model itself asks to employers or other AI agents include, for example, questions to confirm the suitability (degree of match) between the job seeker's desired conditions, personality, values, etc. and the person being asked (employer).

[0079] The job seeker task may include, for example, the revision or suggestion of revisions to the work history of a job seeker, which is the source of the job seeker training data, and which is the job seeker task execution model that handles. In other words, the job seeker task execution model may be capable of performing the revision or suggestion of revisions to the work history of a job seeker as a job seeker task.

[0080] A job seeker task may include output of information about the job seeker, as specified by the task instructioner (for example, the task execution unit 118 described later). Alternatively, a job seeker task may also include information indicating a specific goal (objective, purpose, etc.) related to the job seeker, as specified by the task execution unit 118 described later. Such a job seeker task may include instructions such as "present recommended job openings to the job seeker" or information indicating an objective such as "I want to know what recommended job openings are," and the information output by the job seeker task execution model may include, for example, information on recommended job openings to the job seeker.

[0081] "Recommended job postings for job seekers" are, for example, job postings from employers with a high degree of match with the job seeker, job postings that are presumed to have a high probability of passing the screening process, or job postings that are presumed to have a high probability of resulting in a successful placement. The job seeker task execution model may determine which job postings to recommend to the job seeker by comparing information about the job posting to be evaluated, the employer's memory for that job posting, etc., with information about the job seeker that has been trained (for example, job seeker memory, etc.) (for example, by converting each piece of information into a vector and determining the similarity based on the cosine similarity between the vectors, etc.).

[0082] The job seeker learning data used in the job seeker agent construction unit 114 includes information about job seekers registered in the job seeker database (registered information), information provided by job seekers (provided information), or publicly available information about job seekers on the network.

[0083] The "job seeker database" is a database used in the platform provided by Information Processing System 1, along with the employer database. In other words, the job seeker learning data used in the job seeker agent construction unit 114 may include information on job seekers registered in the job seeker database on a single platform.

[0084] Job seeker registration information (hereinafter referred to as "job seeker registration information") includes, for example, the job seeker's resume, work history, and other profile information. A "resume" is a document that mainly describes the job seeker's profile, current situation, educational background, work history, and desired working conditions. A "work history," also called a resume, is a document in which the job seeker communicates their past work experience, skills, qualifications, etc., to the employer. Job seeker registration information may also include conditions such as the industry and job type that the job seeker desires.

[0085] Furthermore, the job seeker registration information may include a history of the job seeker's actions (for example, viewing job postings, applying for jobs, opening recruitment emails, replying to recruitment emails, conducting interviews, and securing employment). In addition, the job seeker registration information may include records of the content of interviews or meetings with employers.

[0086] Thus, the job seeker learning data may include at least one of the following as information about job seekers registered in the job seeker database: job postings viewed or applied for by the job seeker, and recruitment emails opened or replied to. This makes it possible to construct a job seeker task execution model that can perform tasks based on the job seeker's job-seeking behavioral history.

[0087] Information provided by job seekers may include, for example, the results of aptitude tests, personality tests, and written descriptions of the job seeker's values. The job seeker agent building unit 114 acquires the information by, for example, receiving input, uploading, and specifying a storage location on the network from the job seeker terminal 30. The job seeker agent building unit 114 may also provide incentives to job seekers for providing information, such as granting them benefits that can be used on the platform.

[0088] The types of websites from which job seekers' publicly available information can be obtained are not limited to websites that contain at least some information about the job seeker, and include the job seeker's own blog or social networking site, or third-party websites that publish information about the job seeker (e.g., interview content).

[0089] The job seeker agent building unit 114 may automatically collect publicly available information by keyword searching, crawling, or web scraping on the internet using job seeker names, etc. Alternatively, the job seeker agent building unit 114 may receive access information from the job seeker terminal 30 indicating the network location of the publicly available information, and obtain the publicly available information based on the entered access information.

[0090] The job seeker task execution model is a learning model included in the artificial intelligence unit 120 that has been machine-trained to take a job seeker task as input and output the result of executing the job seeker task. The job seeker task execution model is a learning model that has been machine-trained using job seeker learning data and data necessary for executing the job seeker task (for example, a combination of input and output examples for the job seeker task) as training data.

[0091] The job seeker task execution model may include a large-scale language model, including a generative AI. If the job seeker task execution model is a large-scale language model, the job seeker agent construction unit 114 may create at least one job seeker memory for each job seeker from the job seeker training data, which the job seeker task execution model will refer to when generating output. This allows the job seeker task execution model to be constructed using an existing large-scale language model (for example, a large-scale language model common to the employer task execution model). It also facilitates updating the job seeker task execution model.

[0092] Even when the job seeker task execution model is an AI agent, the job seeker agent construction unit 114 may create at least one job seeker memory for each job seeker from the job seeker training data, which the job seeker task execution model will refer to when generating output. This allows for the construction of a job seeker task execution model tailored to the individual characteristics of each job seeker, using a pre-built AI agent.

[0093] Job seeker memory is information representing the characteristics or attributes of a job seeker, and it is a parameter that customizes (determines the characteristics of) the job seeker task execution model. The job seeker task execution model executes the job seeker task as a job seeker agent that has learned the characteristics represented in the job seeker memory. Job seeker memory is registered, for example, in the job seeker database, linked to the job seeker registration information.

[0094] The job seeker memory may contain at least some information about the job seeker's personality. This makes it possible to construct a job seeker task execution model that can respond while considering the matching of the job seeker's personality. In particular, if the employer memory contains information about the organization's culture, as mentioned above, it becomes possible to match the culture of the job seeker's organization with the job seeker's personality through a combination (dialogue) of the employer task execution model and the job seeker task execution model.

[0095] The job seeker memory includes information such as the job seeker's values, values ​​they dislike, their life goals, essential working conditions, and preferred working conditions. Specific examples of the contents of the job seeker memory include "value personal charm," "value work-life balance," and "want to make an impact on society." The job seeker memory may be information written in natural language, or it may be code information such as computer language, programming language, or machine code, as long as it can be referenced by the job seeker task execution model. The number of job seeker memories prepared for one job seeker task execution model (one job seeker) may be on the order of 1000, for example.

[0096] The job seeker agent construction unit 114, for example, inputs job seeker learning data into a memory creation model and causes the memory creation model to output job seeker memory. The memory creation model is a learning model similar to that used in the employer agent construction unit 112. If the memory creation model is a large-scale language model including generative AI, the job seeker agent construction unit 114 inputs a prompt into the memory creation model that includes an instruction to output job seeker memory corresponding to the job seeker learning data, and causes the memory creation model to output job seeker memory. The specific procedure for outputting job seeker memory (example of prompt) is the same as the procedure for outputting employer memory.

[0097] The job seeker agent construction unit 114 may cause the memory generation model, which is a large-scale language model, to output job seeker memory by inputting input information including input to the job seeker task execution model, output from the job seeker task execution model for said input, or a combination thereof, and an instruction to create a pre-job seeker memory representing the characteristics or attributes of the job seeker based on said input information. This allows the job seeker memory to be added or updated based on the execution results of the tasks of the job seeker task execution model, thereby improving the execution accuracy of the tasks of the job seeker task execution model.

[0098] The job seeker agent construction unit 114 may present the job seeker with questions to create a job seeker memory and create the job seeker memory based on the answers the job seeker provides to those questions. For example, the job seeker agent construction unit 114 may input the job seeker's answers into a memory generation model and have the memory generation model output the job seeker memory. Alternatively, the job seeker agent construction unit 114 may use the results of creating the job seeker memory, which is a subtask executed by the job seeker task execution model, an AI agent, when performing job seeker tasks such as extracting recommended job postings for the job seeker.

[0099] The job seeker agent construction unit 114 may create a job seeker memory for each job seeker representing the characteristics or attributes of the job seeker from the job seeker learning data, and may also construct a job seeker task execution model using machine learning based on the job seeker memory. This makes it possible to construct a job seeker task execution model specific to each job seeker based on the job seeker memory. The job seeker task execution model constructed using the job seeker memory may be a dedicated model that has been machine-trained to take a task as input and output the execution result of the task, or it may be a large-scale language model (specifically, for example, a general-purpose large-scale language model that has undergone reinforcement learning or fine-tuning using the job seeker learning data and the job seeker memory).

[0100] The job seeker agent construction unit 114 may construct a job seeker task execution model by machine learning using job seeker labels created by the job seeker label creation unit 115 (described later) and edited by the job seeker. This allows job seekers to customize the job seeker task execution model and modify its characteristics, thereby improving the accuracy of matching employers and job seekers using the job seeker task execution model.

[0101] The job seeker agent building unit 114 may build a job seeker task execution model by machine learning using further recorded data of interviews between recruitment agencies and employers or job seekers. This allows the job seeker task execution model to learn from questions from recruitment agencies and the answers to those questions, thereby improving the accuracy of matching employers and job seekers using the job seeker task execution model.

[0102] <Job seeker label creation section 115> The job seeker label creation unit 115 is configured to output at least one job seeker label by inputting a job seeker memory and an instruction to generate a job seeker label that represents the contents of the job seeker memory in natural language to the memory generation model.

[0103] A job seeker label represents the content of each job seeker memory in natural language (words or sentences) that the job seeker can understand. Note that multiple job seeker labels may be generated from a single job seeker memory. Job seeker labels are registered, for example, in the job seeker database, linked to the job seeker registration information.

[0104] The job seeker label creation unit 115, for example, takes a job seeker memory as input and inputs a prompt to the memory creation model that includes an instruction to output a job seeker label corresponding to the job seeker memory, causing the memory creation model to output the job seeker label. The specific procedure for outputting the job seeker label (example of prompt) is the same as the procedure for outputting the employer label.

[0105] The job seeker label creation unit 115 may acquire job seeker labels created by the job seeker task execution model, which is an AI agent. The AI ​​agent may perform the generation of job seeker labels as a subtask when executing a job seeker task, such as extracting job postings to recommend to job seekers. The AI ​​agent may also perform the generation of job seeker memory and job seeker labels as a subtask, for example. Furthermore, the job seeker label creation unit 115 may cause the job seeker task execution model to create a prompt for the memory creation model to create job seeker labels, and input the prompt to the memory creation model.

[0106] The job seeker label creation unit 115 presents all or part of the multiple job seeker labels generated for each job seeker to the corresponding job seeker (job seeker terminal 30). This allows the job seeker to confirm the components (features or attributes) of their own job seeker task execution model (AI agent).

[0107] For example, the job seeker label creation unit 115 may determine whether multiple job seeker labels contain positive or negative content, and present only the job seeker labels determined to contain positive content to the job seeker terminal 30. Alternatively, the job seeker label creation unit 115 may have a large-scale language model, including a generation AI, determine whether a job seeker label contains positive or negative content. Furthermore, the job seeker label creation unit 115 may utilize the results of a subtask performed by a job seeker task execution model, which is an AI agent, when performing job seeker tasks such as extracting recommended job postings for job seekers, specifically the determination of whether a job seeker label contains positive or negative content.

[0108] Furthermore, the job seeker label creation unit 115 may present to the job seeker terminal 30 job seeker labels that have a high degree of relevance to items that are important in matching with employers, and to items that job seekers want to highlight to employers (for example, job seeker labels ranked from highest to lowest relevance). These items may be extracted from, for example, job seeker learning data (job seeker registration information), or they may be entered by the job seeker. In addition, the job seeker label creation unit 115 may present to the job seeker terminal 30 a job seeker label summary, which summarizes combinations of several job seeker labels.

[0109] The job seeker label creation unit 115 may accept edits of job seeker labels from the corresponding job seeker (job seeker terminal 30). This allows for customization of the job seeker task execution model based on edits made by the job seeker. Editing also includes deleting job seeker labels.

[0110] <Question Registration Section 116> The question registration unit 116 is configured to register pre-employment questions received from job seekers for employers. Specifically, the question registration unit 116 receives the selection of an employer and the input of the pre-employment questions for that employer from the job seeker terminal 30. The question registration unit 116 links the job seeker who created the pre-employment questions, the selected employer (the target of the pre-employment questions), and the pre-employment questions, and registers them, for example, in the job seeker database. Note that one pre-employment question (a common question for multiple employers) may be registered for multiple employers.

[0111] <Platform Agent Construction Department 117> The platform agent construction unit 117 is configured to construct a platform task execution model capable of performing tasks that handle information of any employer or any job seeker by using machine learning with employer training data and job seeker training data.

[0112] The platform task execution model is constructed on a platform-by-platform basis provided by the information processing system 1. The platform task execution model may be an output device that outputs the execution results of an input task, or it may be a so-called AI agent (platform agent) that autonomously performs tasks such as data collection and conversation.

[0113] The tasks executed by the platform task execution model (hereinafter referred to as "platform tasks") include employer tasks executed by the employer task execution model and job seeker tasks executed by the job seeker task execution model. In addition, platform tasks may include tasks that do not target individual employers or job seekers, but rather target multiple employers or job seekers registered on the platform (for example, tasks related to the aggregation of statistical data of employers or job seekers).

[0114] The platform task execution model may be capable of answering questions about employers or job seekers as platform tasks. In other words, the platform task execution model may be capable of conversing or interviewing employers, job seekers, or other AI agents.

[0115] If the platform task execution model is an AI agent, it may generate subtasks based on the goals or tasks entered by the employer or job seeker, and for each subtask, select the optimal model (e.g., a job seeker task execution model or an employer task execution model) to perform the processing. The platform task execution model may also generate prompts necessary for the execution of the subtasks and input the generated prompts into the selected model, thereby causing that model to perform the processing.

[0116] The employer training data used in the platform agent construction unit 117 may include information on multiple employers registered in the employer database used in the employer agent construction unit 112. Similarly, the job seeker training data used in the platform agent construction unit 117 may include information on multiple job seekers registered in the job seeker database used in the job seeker agent construction unit 114. As a result, the employer task execution model and the job seeker task execution model, and the platform task execution model are constructed based on common employer training data and job seeker training data. Therefore, by coordinating the employer task execution model or the job seeker task execution model with the platform task execution model, it becomes possible to efficiently match employers and job seekers based on information from the entire platform.

[0117] The employer training data used in the platform agent construction unit 117 may include information on employers for whom an employer task execution model has not been built. Similarly, the job seeker training data used in the platform agent construction unit 117 may include information on job seekers for whom a job seeker task execution model has not been built. This makes it possible to provide users (employers or job seekers) with platform task execution results based on information on employers for whom an employer task execution model has not been built, or on job seekers for whom a job seeker task execution model has not been built.

[0118] The platform task execution model is a learning model included in the artificial intelligence unit 120 that has been trained to take a platform task as input and output the execution result of the platform task. The platform task execution model is a learning model that has been trained using, for example, employer training data and job seeker training data and data necessary for executing the platform task (for example, combinations of input examples and output examples of the platform task) as training data.

[0119] The training data for the platform task execution model includes employer training data containing information on multiple employers and job seeker training data containing information on multiple job seekers, and therefore includes employer statistics and job seeker statistics on the platform provided by Information Processing System 1.

[0120] Furthermore, the training data for the platform task execution model may include external information publicly available outside the platform regarding workers' employment. Examples of such external information include labor statistics and policies of government and other public institutions. This enables the platform task execution model to produce outputs (responses) that take into account the correction of matching biases and responses to policies (e.g., increasing the number of female executives).

[0121] The platform task execution model may be a generative AI, or a large-scale language model. If the platform task execution model is a large-scale language model, the platform agent construction unit 117 may create at least one platform memory from the employer training data and the job seeker training data that the platform task execution model will refer to when generating output.

[0122] Platform memory is information that represents the overall statistical characteristics of the platform and is a parameter that customizes (determines the characteristics of) the platform task execution model. Platform memory includes, for example, statistics on employers such as the number of job postings, the number of applications, and the number of successful placements (success rate) for each job attribute (industry, job type, duties, requirements, etc.), and statistics on job seekers such as the number of recruitment emails received, the number of recruitment email replies (reply rate), the number of job applications, and the number of successful job placements (job placement success rate) for each job seeker attribute (background, experience, skills, age, etc.).

[0123] The platform agent construction unit 117 inputs, for example, employer training data and job seeker training data into the memory creation model and causes the memory creation model to output platform memory. The memory creation model is a training model similar to the one used in the employer agent construction unit 112. If the memory creation model is a large-scale language model, the platform agent construction unit 117 inputs prompts into the memory creation model that include instructions to input employer training data and job seeker training data and output platform memory corresponding to said employer training data and said job seeker training data, causing the memory creation model to output platform memory. The specific procedure for outputting platform memory (example of prompt) is the same as the procedure for outputting employer memory.

[0124] The platform agent construction unit 117 may create a platform memory representing the characteristics of the entire platform from the employer training data and the job seeker training data, and may also construct a platform task execution model using machine learning with the platform memory. The platform task execution model constructed using the platform memory may be a dedicated model that has been machine-trained to take a task as input and output the execution result of the task, or it may be a large-scale language model (specifically, for example, a general-purpose large-scale language model that has been fine-tuned using the platform memory, etc.).

[0125] <Task Execution Unit 118> The task execution unit 118 is configured to cause the recruiter task execution model or the job seeker task execution model to execute a task (recruiter task or job seeker task) based on an execution instruction for the job seeker task execution model or the job seeker task execution model.

[0126] The instructions for executing a recruiter task or a job seeker task may be information indicating the task (e.g., questions to the AI ​​agent) directly entered from the recruiter terminal 20 or the job seeker terminal 30, or they may be generated by the task execution unit 118 based on the content of the task entered from the recruiter terminal 20 or the job seeker terminal 30 (e.g., a prompt). Alternatively, the instructions for executing a recruiter task or a job seeker task may be automatically generated by the task execution unit 118 without input (instructions) from the recruiter terminal 20 or the job seeker terminal 30 (e.g., periodic task execution instructions).

[0127] Furthermore, the recruiter task or job seeker task may be autonomously generated by the recruiter task execution model or job seeker task execution model (AI agent) in response to the objectives (purpose, goal, etc.) input by the user. In this case, the prompts, which are instructions to the recruiter task execution model or job seeker task execution model, are also generated by the recruiter task execution model or job seeker task execution model itself.

[0128] For example, if the job seeker task execution model is an AI agent, the task execution unit 118 inputs the task or goal included in the execution instructions input by the job seeker to the AI ​​agent, which has been built by learning from the job seeker learning data. Based on the input, the AI ​​agent generates the necessary subtasks. The AI ​​agent may generate prompts for executing the subtasks and input these prompts to an external large-scale language model to obtain the execution results of the subtasks from the large-scale language model, or it may execute the generated prompts itself. Alternatively, the AI ​​agent may create input information to cause an external learning model that performs a specific process to perform a specific process, and input this input information to the learning model to obtain the execution results of the subtasks from the learning model. The AI ​​agent presents the job seeker with information corresponding to the input task or goal.

[0129] If the employer task execution model and the job seeker task execution model are large-scale language models, the task execution unit 118 may input the employer memory or job seeker memory as reference information along with the task (employer task or job seeker task) to the employer task execution model or job seeker task execution model that is the target of the task execution instruction, and cause the employer task execution model or job seeker task execution model to execute the task.

[0130] Specifically, the task execution unit 118 takes a task and the recruiter memory or job seeker memory as input, inputs a prompt to the recruiter task execution model or job seeker task execution model that includes an instruction to output the execution result of the task (e.g., an answer to a question), and causes the recruiter task execution model or job seeker task execution model to output the execution result. The task execution unit 118 may also generate a prompt that gives the recruiter task execution model or job seeker task execution model an instruction to output the execution result of the task, and input the prompt to the recruiter task execution model or job seeker task execution model. In addition to the task execution instruction and the task and recruiter memory or job seeker memory, the task execution unit 118 may also input a prompt to the recruiter task execution model or job seeker task execution model that includes, for example, one or more sample combinations of tasks and recruiter memory or job seeker memory, and one or more corresponding sample execution results, as examples, samples, or training data of input and output pairs.

[0131] If the employer task or job seeker task is an answer to a question about the employer or job seeker, the task execution unit 118 may receive the question for the employer task execution model or job seeker task execution model as a task execution instruction, and cause the employer task execution model or job seeker task execution model to output an answer to the question.

[0132] The task execution unit 118 may request the employer corresponding to the employer task execution model or the job seeker corresponding to the job seeker task execution model to converse with the user, depending on the content of the conversation between the user (employer or job seeker) and the employer task execution model or the job seeker task execution model. This allows the conversation partner for the user to be switched from the employer task execution model or the job seeker task execution model (AI agent) to the user themselves, in cases where the employer task execution model or the job seeker task execution model cannot answer a question, or in situations where a direct conversation with the person in question is necessary.

[0133] If, for example, the recruiter task execution model or the job seeker task execution model is unable to generate a response (i.e., it provides a response indicating that it cannot generate a response), the task execution unit 118 outputs a notification to the recruiter terminal 20 of the recruiter corresponding to the recruiter task execution model or the job seeker terminal 30 of the job seeker corresponding to the job seeker task execution model, requesting them to participate in a conversation with the user. If the recruiter or job seeker accepts this request, communication is initiated between the recruiter terminal 20 or the job seeker terminal 30 and the user's terminal to conduct a conversation (e.g., video conference, voice call, etc.).

[0134] Furthermore, the task execution unit 118 may request the employer or job seeker to participate in the conversation if either the user's question or the response from the employer task execution model or the job seeker task execution model contains a pre-set specific important keyword (including similar keywords). Important keywords include, for example, keywords related to job applications, hiring procedures, etc.

[0135] Furthermore, if the employer or job seeker does not accept the request to converse with the user, or if there is no response from the employer or job seeker, the task execution unit 118 may present the user's questions to the employer or job seeker and subsequently accept the user's answers.

[0136] The task execution unit 118 may perform a job seeker presentation process by having the recruiter task execution model execute a task to extract job seekers recommended to the recruiter corresponding to the recruiter task execution model, or a job presentation process by having the job seeker task execution model execute a task to extract job postings recommended to the job seeker corresponding to the job seeker task execution model. This makes it possible to present recommended job seekers based on recruiter information (recruiter learning data) or recommended job postings based on job seeker information (job seeker learning data), thereby increasing the efficiency of matching between recruiters and job seekers. The recommended job postings may be job postings offered by recruiters recommended to job seekers. In other words, the task execution unit 118 may use the job seeker task execution model to extract recruiters recommended to job seekers.

[0137] The task execution unit 118 displays at least one recommended job seeker extracted by the job seeker presentation process on the employer terminal 20. The task execution unit 118 also displays at least one recommended job opening extracted by the job opening presentation process on the job seeker terminal 30.

[0138] If the employer task execution model is capable of answering questions about employers, and the job seeker task execution model is capable of answering questions about job seekers, the task execution unit 118 may have the employer task execution model and the job seeker task execution model converse, and based on the content of the conversation, determine which job seekers to recommend to employers, or which job openings to recommend to job seekers. This makes it possible to present recommended job seekers or recommended job openings through a simulated interview between employers and job seekers (an interview between AI agents).

[0139] The conversation between the employer task execution model and the job seeker task execution model is initiated, for example, by the task execution unit 118 causing one model to ask questions to the other. For example, when determining a recommended job seeker, the task execution unit 118 causes the employer task execution model to ask questions to the job seeker task execution model. Similarly, when determining a recommended job opening, the task execution unit 118 causes the job seeker task execution model to ask questions to the employer task execution model.

[0140] If the employer task execution model and the job seeker task execution model are AI agents, the task execution unit 118 may, for example, receive a task execution instruction from the employer, such as "Find (list) matching job seekers," and input this task execution instruction as a target to the employer task execution model corresponding to that employer. The employer task execution model, as a subtask for the input target, sequentially asks questions to multiple job seeker task execution models (the other AI agents), extracts necessary job seeker information, vectorizes it, calculates similarity, etc., to determine job seekers with a high degree of match, and outputs or visualizes the results. Here, the employer task execution model and the job seeker task execution model each autonomously generate the necessary subtasks (for example, questions to determine compatibility), and further generate and execute the code necessary to execute the subtasks. Subtasks include, for example, generating and presenting questions to the other AI agents necessary for matching, and analyzing the answers from the AI ​​agents to the questions. The task execution unit 118 presents the recommended job seekers determined by the employer task execution model to the user.

[0141] Furthermore, if the employer task execution model and the job seeker task execution model are AI agents, the task execution unit 118 may, for example, receive a task execution instruction from a job seeker such as "Find (list) matching employers or job postings," and input this task execution instruction as a target to the job seeker task execution model corresponding to that job seeker. The job seeker task execution model, as a subtask for the input target, sequentially asks questions to multiple employer task execution models (the other AI agents), extracts necessary employer information, vectorizes it, calculates similarity, etc., to determine employers or job postings with a high degree of match, and outputs or visualizes the results. The task execution unit 118 then presents the recommended employers or job postings determined by the employer task execution models to the user.

[0142] Furthermore, the task execution unit 118 may, in response to the task execution instruction "Find a matching job seeker," input the task "Calculate the degree of match with each of the multiple job seeker task execution models" to the employer task execution model, causing the employer task execution model to calculate the degree of match with each job seeker. In addition, the task execution unit 118 may present job seekers with a degree of match of a predetermined level or higher to the user as recommended job seekers. The same applies when presenting recommended employers or recommended job postings.

[0143] The task execution unit 118, for example, records conversation logs in the employer task execution model and the job seeker task execution model, and analyzes the logs to determine whether the job seeker corresponding to the job seeker task execution model is a recommended job seeker for the employer corresponding to the employer task execution model, or whether the job posting from the employer corresponding to the employer task execution model is a recommended job posting for the job seeker corresponding to the job seeker task execution model.

[0144] The task execution unit 118 may determine a recommended job seeker or recommended job based on the number of agreements in the conversation log. Here, "agreement" may be interpreted as an affirmative or positive response to a question. For example, if the number of agreements exceeds a predetermined number, it is determined to be a recommended job seeker or recommended job. Alternatively, the task execution unit 118 may have the employer task execution model or job seeker task execution model ask specific questions and determine a recommended job seeker or recommended job based on the answers to those questions. Specific questions may, for example, be questions to confirm the employer's willingness to hire (likelihood of hiring) or the job seeker's willingness to apply (likelihood of applying).

[0145] The task execution unit 118 may determine which job seekers to recommend to employers, or which job openings to recommend to job seekers, based on the similarity calculated from the features of the employer training data and the features of the job seeker training data. This makes it possible to extract recommended job seekers or recommended job openings based on quantitative judgments using information from both job seekers and employers.

[0146] The features used for the training data are either individual pieces of information for each category included in the training data (for example, industry, job title, duties, skills, qualifications, etc., included in job postings or job seeker registration information) or vector data obtained by vectorizing the entire training data. Vectorization is performed by quantification using known methods such as natural language processing using morphological analysis or encoding. The similarity between the employer training data and the job seeker training data is represented by the difference in features of the same category or the total features (for example, the vector distance). The task execution unit 118 determines, for example, that job postings or job seekers of employers whose similarity (for example, cosine similarity) between the employer training data and the job seeker training data is above a threshold are recommended job postings or recommended job seekers.

[0147] The task execution unit 118 may determine which job seekers to recommend to employers, or which job openings to recommend to job seekers, based on the similarity calculated from the features of the employer memory as features of the employer training data and the features of the job seeker memory as features of the job seeker training data. This allows for the determination of recommended job seekers or recommended job openings based on the characteristics or attributes of employers and job seekers, thereby improving the accuracy of the determination.

[0148] The feature quantities for employer memory and job seeker memory may be vectorized representations of individual employer memories and individual job seeker memories, respectively, or they may be vectorized representations of all employer memories for one employer and all job seeker memories for one job seeker.

[0149] The task execution unit 118 may use the number of combinations of employer memory and job seeker memory with small differences in individual features as the similarity score, or it may use the small difference between the total features of the employer memory and the total features of the job seeker memory as the similarity score. The task execution unit 118 determines that employers whose similarity score is equal to or greater than a predetermined value are recommended employers or recommended job seekers.

[0150] Figure 5 shows an image of matching employers and job seekers using an employer task execution model and a job seeker task execution model. The task execution unit 118 estimates the degree of matching between employers and job seekers by the conversation between the constructed employer task execution model and the job seeker task execution model, or by comparing registered employer labels (employer memory) and job seeker labels (job seeker memory).

[0151] The task execution unit 118 may present the results of the conversation between the employer task execution model and the job seeker task execution model to at least one of the employers corresponding to the employer task execution model and the job seekers corresponding to the job seeker task execution model. This allows the employer or job seeker to confirm the responses of their agent. Furthermore, depending on the confirmation results, the employer or job seeker can customize the employer task execution model or the job seeker task execution model (AI agent) by editing employer labels, job seeker labels, etc.

[0152] For example, the task execution unit 118 displays the recruiter task execution model, the job seeker task execution model, and the conversation log on the recruiter terminal 20 or the job seeker terminal 30 in response to a request from the recruiter or job seeker. The task execution unit 118 may, if it receives a refusal from the job seeker (preferably along with the reason for refusal), refrain from displaying the conversation results (log) to the recruiter.

[0153] The task execution unit 118 may present the log of the conversation between the user (job seeker) and the recruiter task execution model to the recruiter corresponding to the recruiter task execution model. Alternatively, the task execution unit 118 may present the log of the conversation between the user (job seeker) and the job seeker task execution model to the job seeker corresponding to the job seeker task execution model.

[0154] The task execution unit 118 may cause the recruiter task execution model, which corresponds to the recruiter specified by the job seeker, to execute the answers to the pre-registered questions registered by the question registration unit 116 as a recruiter task. This allows the job seeker to obtain answers from the recruiter task execution model at any time of their choosing.

[0155] The task execution unit 118 may, for example, cause the employer task execution model to answer the pre-questions when the job seeker views information about the employer that the job seeker has designated as the target of the pre-questions (e.g., organizational information, job posting, etc.). The task execution unit 118 may also cause the employer task execution model to answer the pre-questions when the job seeker takes an action other than viewing information about the employer (e.g., logging into a platform service, applying for a job, etc.).

[0156] The task execution unit 118 may, for example, input a goal such as "I want to revise my resume" entered by the job seeker into the job seeker task execution model, which is an AI agent. The job seeker task execution model generates and executes subtasks such as acquiring documents related to work experience, such as resumes, that the job seeker has registered, analyzing the documents related to work experience, identifying areas that need revision, generating revision proposals, identifying missing information, generating questions to obtain missing information, presenting questions to the job seeker, receiving answers, and generating revision proposals based on the answers. As a result, the job seeker task execution model can revise the documents related to the job seeker's work experience while conversing or interviewing the job seeker, thereby improving the quality of the job seeker's work experience.

[0157] The task execution unit 118 causes the platform task execution model to execute the platform task based on the execution instruction for the platform task to the platform task execution model.

[0158] The platform task execution instruction may be information indicating a task, directly input from the recruiter terminal 20 or the job seeker terminal 30, or it may be something generated by the task execution unit 118 based on the content of the task input from the recruiter terminal 20 or the job seeker terminal 30 (for example, a prompt). Alternatively, the platform task execution instruction may be something automatically generated by the task execution unit 118 without input (instruction) from the recruiter terminal 20 or the job seeker terminal 30 (for example, a periodic task execution instruction).

[0159] When the task execution unit 118 has the recruiter task execution model or the job seeker task execution model execute a task (recruiter task or job seeker task), it may also have the platform task execution model execute the same task or a task related to that task. This allows the user to be presented with the execution results of the task by the platform task execution model in addition to the execution results of the task by the individual recruiter task execution model or job seeker task execution model. As a result, the user can search for job seekers or job postings using execution results that depend on a specific recruiter or job seeker (e.g., answers to questions) and execution results that take into account information from the entire platform (e.g., statistical insights).

[0160] "Tasks related to employer tasks or job seeker tasks" include, for example, if the employer task or job seeker task involves answering questions about employers or job seekers, the presentation of statistical information related to keywords included in those questions. Furthermore, "Tasks related to employer tasks or job seeker tasks" also include, for example, if the employer task or job seeker task involves extracting recommended job seekers or recommended job postings, the presentation of recommended job seekers or recommended job postings based on platform-wide statistical information (for example, to mitigate hiring bias, achieve policies, etc.).

[0161] If each task execution model is an AI agent, the task execution unit 118 may cause multiple AI agents to mutually input task execution instructions from one AI agent to another. For example, an AI agent may execute prompts generated by other AI agents for subtask execution, and at the same time input prompts it has generated to other AI agents for them to execute. In this way, the task execution unit 118 may cause multiple task execution models to mutually input prompts and exchange their respective outputs.

[0162] <Artificial Intelligence Department 120> The artificial intelligence unit 120 is configured to receive input from each functional unit and return the instructed output. The artificial intelligence used by each functional unit of the server device 10 may be common to all units, or it may be prepared individually for each functional unit.

[0163] The artificial intelligence unit 120 is an AI (Artificial Intelligence) equipped with learning models such as transformers including GPT (Generative Pretrained Transformer, including GPT-1, GPT-2, GPT-3, and GPT-4), BERT (Bidirectional Encoder Representations from Transformers), BART (Bidirectional and Auto-regressive Transformer), and language models such as recurrent neural networks (RNNs), and may include generative AI including large-scale language models. Large-scale language models are a type of generative AI and include models provided by services such as OpenAI's GPT, Google's Gemini, and Microsoft's Azure AI Studio. In addition, the artificial intelligence unit 120 can include any machine learning model, deep learning model, artificial intelligence model, etc.

[0164] The language model is an example of a learning model using a machine learning algorithm. Specific machine learning algorithms include nearest neighbors, naive Bayes, decision trees, support vector machines, and deep learning using neural networks. The artificial intelligence unit 120 can apply the above algorithms as appropriate.

[0165] The artificial intelligence unit 120 may have a trained model constructed by a learning method such as supervised learning, unsupervised learning, or self-supervised learning. In supervised learning, machine learning is performed using training data. Training data consists of pairs of input data and output data (correct answer data) for training. Furthermore, the language model may not only be one trained for a specific task, but also a general-purpose model that can be used universally for a wide range of tasks.

[0166] The artificial intelligence unit 120 may be a general-purpose natural language processing learning model, such as a Large Language Model (LLM), which has learned from a vast amount of data. An LLM is a learning model that has been pre-trained on a large amount of data consisting of text data, etc. (for example, (i) web content on the internet, or (ii) data stored in a predetermined database), and can perform various language processing tasks by being given a task. It can perform a wide range of natural language processing tasks, such as understanding sentence patterns and context, responding to questions, and generating sentences, according to the given prompts. Such a general-purpose learning model includes language models that can handle various tasks without fine-tuning using One-shot Learning or Few-shot Learning. Furthermore, the general-purpose learning model may also be configured to handle various tasks using Zero-shot Learning. The artificial intelligence used in each functional unit of the control unit 11 may be a separate learning model, or it may be a common general-purpose learning model.

[0167] The learning models included in the artificial intelligence unit 120 (such as the job seeker task execution model and the job applicant task execution model used in each functional unit) can undergo additional learning through transfer learning or fine-tuning. For example, the artificial intelligence unit 120 may perform additional learning and fine-tuning each time new data is registered, using this data as new training data. This improves the accuracy of the information output from the learning model.

[0168] The learning model included in the artificial intelligence unit 120 may be a learning model (distilled model) obtained by knowledge distillation using the original learning model. In knowledge distillation, a pre-trained model, such as a large-scale language model, is used as the teacher model, and the parameters of the student model are adjusted so that the output loss of the student model (distilled model) relative to the output (Soft Target Loss) of the teacher model is small. The student model is then trained, and this student model becomes the distilled model. Alternatively, the student model may be trained so that the output loss of the student model relative to the correct labels (Hard Target) of the teacher data (combinations of input and output data of the learning model) is small. Compared to the original learning model (teacher model), the distilled model has similar performance to the original learning model, but with fewer parameters and a lower processing load. Therefore, using a distilled model can reduce the cost of the information processing system 1.

[0169] For example, the learning model used in each functional unit may be a distilled model that has been trained using combinations of input and output data from a large-scale language model as training data. Alternatively, when the information processing system 1 is introduced, a large-scale language model may be used as the learning model in each functional unit, and once training data from the large-scale language model has been accumulated, the distilled model obtained by knowledge distillation using that training data may be used as the learning model in each functional unit.

[0170] <Display section> The display unit 211 of the job seeker terminal 20 and the display unit 311 of the job seeker terminal 30 each display the screen indicated by the screen data transmitted from the server device 10.

[0171] <Operation acquisition section> The operation acquisition unit 212 of the employer terminal 20 receives operations from the employer using the employer terminal 20. The operation acquisition unit 312 of the job seeker terminal 30 receives operations from the job seeker using the job seeker terminal 30.

[0172] 3. Information Processing Methods This section describes the information processing method of the server device 10. In this information processing method, each part of the server device 10 is executed by a computer as a step.

[0173] This information processing comprises a recruiter agent construction step, a job seeker agent construction step, and a task execution step. In the recruiter agent construction step, a recruiter task execution model capable of performing tasks that handle recruiter information is constructed for each recruiter using machine learning with recruiter training data about the recruiter. In the job seeker agent construction step, a job seeker task execution model capable of performing tasks that handle job seeker information is constructed for each job seeker using machine learning with job seeker training data about the job seeker. In the task execution step, the recruiter task execution model or the job seeker task execution model is instructed to perform the task based on the task execution instructions given to the recruiter task execution model or the job seeker task execution model.

[0174] Figure 6 is an activity diagram showing an example of the flow of information processing (task execution processing) performed by the information processing system 1. The information processing will be explained below in accordance with each activity in this activity diagram.

[0175] Task execution processing begins with the construction of the employer task execution model and the job seeker task execution model. Server device 10 first acquires employer training data and job seeker training data (Activity A101). Subsequently, server device 10 uses this data to construct the employer task execution model and the job seeker task execution model (Activity A102). Activities A101 and A102 are performed as needed at any time.

[0176] The employer or job seeker inputs a task (for example, a question about the employer or job seeker) to be executed by the constructed employer task execution model or job seeker task execution model at the employer terminal 20 or job seeker terminal 30 (Activity A201). The server device 10 converts the task execution instructions received from the employer terminal 20 or job seeker terminal 30 into the employer task execution model or job seeker task execution model (Activity A202).

[0177] Next, the server device 10 outputs the task execution results (for example, answers to questions) from the recruiter task execution model or the job seeker task execution model to the recruiter terminal 20 or the job seeker terminal 30 (Activity A203). As a result, the task execution results are displayed on the recruiter terminal 20 or the job seeker terminal 30 (Activity A204).

[0178] 4. Effect The operation of this embodiment can be summarized as follows: Specifically, job seekers and employers can be efficiently matched by task execution models (AI agents) individually constructed for each party.

[0179] Although embodiments of the present invention have been described above, the present invention is not limited thereto and can be modified as appropriate without departing from the technical spirit of the invention.

[0180] 5. Others In the above embodiment, the server device 10 performed various storage and control functions, but instead of the server device 10, multiple external devices may be used. That is, various information and programs may be stored in a distributed manner across multiple external devices using blockchain technology or the like. In particular, the artificial intelligence unit 120 may be an external configuration of the server device 10. In that case, the external artificial intelligence unit 120 may be provided by, for example, an artificial intelligence service server, and is configured to receive input from each functional unit of the server device 10, receive requests to execute artificial intelligence services, and return the instructed output as a processing result to the server device 10. The artificial intelligence service server may be a server that provides services using a language model as a learning model, or a server that executes language processing tasks using a language model. The artificial intelligence service server may be constructed using an LLM. The artificial intelligence service server receives prompt input in the form of text, images, audio, etc., and generates and responds with answers to the prompts.

[0181] The information processing system 1 does not necessarily have a platform agent construction unit 117. In other words, the information processing system 1 does not necessarily have to construct a platform task execution model. Furthermore, the information processing system 1 does not necessarily have a recruiter label creation unit 113, a job seeker label creation unit 115, and a question registration unit 116.

[0182] The embodiments of this model are not limited to the information processing system 1, but may also be an information processing method or a program. The information processing method comprises each step executed by the information processing system 1. The program causes a computer to execute each step of the information processing system 1.

[0183] The product may be provided in any of the following embodiments.

[0184] (1) An information processing system comprising at least one processor, wherein the processor is configured to perform the following steps by reading a program, wherein in the recruiter agent construction step, a recruiter task execution model capable of performing tasks that handle recruiter information is constructed for each recruiter by machine learning using recruiter learning data relating to recruiters, wherein the recruiter learning data includes recruiter information registered in a recruiter database, information provided by recruiters, or publicly available information relating to recruiters on a network; in the job seeker agent construction step, a job seeker task execution model capable of performing tasks that handle job seeker information is constructed for each job seeker by machine learning using job seeker learning data relating to job seekers, wherein the job seeker learning data includes job seeker information registered in a job seeker database, information provided by job seekers, or publicly available information relating to job seekers on a network; and in the task execution step, the information processing system causes the recruiter task execution model or the job seeker task execution model to perform a task based on an instruction to perform the task on the recruiter task execution model or the job seeker task execution model.

[0185] (2) An information processing system as described in (1) above, wherein in the recruiter agent construction step, learning data is prepared for each department of the recruiter, job type of the recruiter, or job handled by the recruiter, and the recruiter task execution model is constructed for each department, job type, or job.

[0186] (3) An information processing system as described in (1) or (2) above, wherein the recruiter learning data includes, as information of recruiters registered in the recruiter database, at least one of the recruiter's personnel recruitment data used in the recruitment management system and the recruiter's evaluation of personnel within the organization used in the personnel system.

[0187] (4) An information processing system according to any one of (1) to (3) above, wherein the job seeker learning data includes, as information of a job seeker registered in the job seeker database, at least one of the job postings that the job seeker has viewed or applied for, and the recruitment emails that the job seeker has opened or replied to.

[0188] (5) An information processing system according to any one of (1) to (4) above, wherein the recruiter task execution model includes a large-scale language model, and in the recruiter agent construction step, a recruiter memory is created for each recruiter from the recruiter learning data, which the recruiter task execution model refers to when generating output; the job seeker task execution model includes a large-scale language model, and in the job seeker agent construction step, a job seeker memory is created for each job seeker from the job seeker learning data, which the job seeker task execution model refers to when generating output; and in the task execution step, the recruiter task execution model or the job seeker task execution model is input with the task along with the recruiter memory or the job seeker memory as reference information, and the recruiter task execution model or the job seeker task execution model is made to execute the task.

[0189] (6) An information processing system according to any one of (1) to (4) above, wherein in the recruiter agent construction step, a recruiter memory representing the characteristics or attributes of the recruiter is created for each recruiter from the recruiter learning data, and the recruiter task execution model is constructed by machine learning using the recruiter memory; and in the job seeker agent construction step, a job seeker memory representing the characteristics or attributes of the job seeker is created for each job seeker from the job seeker learning data, and the job seeker task execution model is constructed by machine learning using the job seeker memory.

[0190] (7) An information processing system as described in (5) or (6) above, wherein the recruiter memory contains at least information about the culture of the organization making the job, and the job seeker memory contains at least information about the personality of the job seeker.

[0191] (8) An information processing system according to any one of (5) to (7) above, wherein in the recruiter agent construction step, the memory generation model, which is a large-scale language model, is made to output the recruiter memory by inputting input information including input to the recruiter task execution model, output of the recruiter task execution model for said input, or a combination thereof, and an instruction to create the recruiter memory representing the characteristics or attributes of the recruiter based on said input information, and in the job seeker agent construction step, the memory generation model is made to output the job seeker memory by inputting input information including input to the job seeker task execution model, output of the job seeker task execution model for said input, or a combination thereof, and an instruction to create the aforementioned job seeker memory representing the characteristics or attributes of the job seeker based on said input information.

[0192] (9) Information processing system as described in (8) above, wherein the processor is configured to further perform the following steps: In the recruiter label creation step, the processor inputs the recruiter memory and an instruction to generate a recruiter label that represents the contents of the recruiter memory in natural language to the memory generation model, thereby causing the memory generation model to output the recruiter label and present all or part of the recruiter label to the corresponding recruiter; In the job seeker label creation step, the processor inputs the job seeker memory and an instruction to generate a job seeker label that represents the contents of the job seeker memory in natural language to the memory generation model, thereby causing the memory generation model to output the job seeker label and present all or part of the job seeker label to the corresponding job seeker.

[0193] (10) An information processing system as described in (9) above, wherein in the employer label creation step, the system accepts editing of the employer label from the corresponding employer, and in the job seeker label creation step, the system accepts editing of the job seeker label from the corresponding job seeker.

[0194] (11) An information processing system as described in (10) above, wherein in the recruiter agent construction step, the recruiter task execution model is constructed by machine learning using the edited recruiter labels, and in the job seeker agent construction step, the job seeker task execution model is constructed by machine learning using the edited job seeker labels.

[0195] (12) An information processing system according to any one of (1) to (11) above, wherein the recruiter task execution model is capable of answering questions concerning recruiters as tasks, the job seeker task execution model is capable of answering questions concerning job seekers as tasks, and in the task execution step, the system receives a question for the recruiter task execution model or the job seeker task execution model as an instruction to execute a task, and causes the recruiter task execution model or the job seeker task execution model to output an answer to the question.

[0196] (13) An information processing system as described in (12) above, wherein in the task execution step, the system requests the recruiter corresponding to the recruiter task execution model or the job seeker corresponding to the job seeker task execution model to have a conversation with the user, depending on the content of the conversation between the user and the recruiter task execution model or the job seeker task execution model.

[0197] (14) An information processing system as described in (12) or (13) above, wherein in the recruiter agent construction step, the recruiter task execution model is constructed by machine learning using further recorded data of interviews between the recruitment agency and the recruiter or job seeker.

[0198] (15) An information processing system described in any one of (12) to (14) above, wherein in the job seeker agent construction step, the information processing system constructs the job seeker task execution model by machine learning using further recorded data of interviews between the recruitment agency and the employer or job seeker.

[0199] (16) An information processing system according to any one of (1) to (15) above, wherein the task execution step includes a job seeker extraction process in which the job seeker task execution model executes a task to extract job seekers to be recommended to the job seeker corresponding to the job seeker task execution model, or a job search extraction process in which the job seeker task execution model executes a task to extract job postings to be recommended to job seekers.

[0200] (17) An information processing system as described in (16) above, wherein the employer task execution model is capable of answering questions about employers as a task, and the job seeker task execution model is capable of answering questions about job seekers as a task, and in the task execution step, the employer task execution model and the job seeker task execution model engage in a conversation, and based on the content of the conversation, the information processing system determines which job seekers to recommend to employers or which job openings to recommend to job seekers.

[0201] (18) An information processing system according to any one of (1) to (17) above, wherein in the task execution step, the system determines a job seeker to recommend to an employer, or a job to recommend to a job seeker, based on the similarity calculated from the features of the employer training data and the features of the job seeker training data.

[0202] (19) An information processing system as described in (18) above, wherein the recruiter task execution model includes a large-scale language model, and in the recruiter agent construction step, a recruiter memory is created for each recruiter from the recruiter training data, which the recruiter task execution model refers to when generating output; the job seeker task execution model includes a large-scale language model, and in the job seeker agent construction step, a job seeker memory is created for each job seeker from the job seeker training data, which the job seeker task execution model refers to when generating output; and in the task execution step, an information processing system determines which job seekers to recommend to recruiters, or which job openings to recommend to job seekers, based on the similarity calculated from the features of the recruiter memory and the features of the job seeker memory.

[0203] (20) An information processing system according to any one of (1) to (19) above, wherein the recruiter task execution model is capable of answering questions concerning recruiters as tasks, the job seeker task execution model is capable of answering questions concerning job seekers as tasks, and in the task execution step, the results of a conversation between the recruiter task execution model and the job seeker task execution model are presented to at least one of the recruiter corresponding to the recruiter task execution model and the job seeker corresponding to the job seeker task execution model.

[0204] (21) An information processing system according to any one of (1) to (20) above, wherein the processor is configured to further perform the following steps: in the question registration step, register pre-questions to employers received from job seekers; and in the task execution step, cause the employer task execution model corresponding to the employer specified by the job seeker to execute the answers to the registered pre-questions as a task.

[0205] (22) An information processing system according to any one of (1) to (21) above, wherein the processor is configured to further perform the following steps: in the platform agent construction step, a platform task execution model capable of performing tasks that handle information of any employer or any job seeker is constructed by machine learning using the employer training data and the job seeker training data; and in the task execution step, when the employer task execution model or the job seeker task execution model is made to perform a task, the platform task execution model is also made to perform the same task or a task related to that task.

[0206] (23) An information processing system as described in (22) above, wherein the employer learning data used in the employer agent construction step includes information of employers registered in the employer database on one platform, the job seeker learning data used in the job seeker agent construction step includes information of job seekers registered in the job seeker database on the platform, the employer learning data used in the platform agent construction step includes information of multiple employers registered in the employer database used in the employer agent construction step, and the job seeker learning data used in the platform agent construction step includes information of multiple job seekers registered in the job seeker database used in the job seeker agent construction step.

[0207] (24) An information processing method comprising each step performed by the information processing system described in any one of (1) to (23) above.

[0208] (25) A program that causes a computer to perform each step of the information processing system described in any one of (1) to (23) above. Of course, this is not always the case.

[0209] Finally, while various embodiments relating to this disclosure have been described, these are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be implemented in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents. [Explanation of symbols]

[0210] 1: Information Processing System 2: Communication lines 10: Server device 11: Control Unit 12: Storage section 13: Communications Department 14: Communications bus 20: Job seeker terminal 21: Control Unit 22: Storage section 23: Communications Department 24: Input section 25: Output section 26: Communications bus 30: Job seeker terminal 31: Control Unit 32: Storage section 33: Communications Department 34: Input section 35: Output section 36: Communications bus 111: Basic Display Control Unit 112: Recruiter Agent Development Department 113: Job Posting Label Creation Department 114: Job Seeker Agent Development Department 115: Job Seeker Label Creation Department 116: Question Registration Section 117: Platform Agent Development Department 118: Task Execution Unit 120: Artificial Intelligence Department 211:Display section 212: Operation acquisition section 311: Representation Department 312: Operation Acquisition Department

Claims

1. An information processing system, Equipped with at least one processor, The aforementioned processor is configured to perform the following steps by reading a program: In the recruiter agent construction step, a recruiter task execution model capable of performing tasks that handle recruiter information is constructed for each recruiter using machine learning with recruiter training data about the recruiter. Here, the recruiter training data includes recruiter information registered in the recruiter database, information provided by the recruiter, or publicly available information about the recruiter on the network. In the job seeker agent construction step, a job seeker task execution model capable of performing tasks that handle job seeker information is constructed for each job seeker using machine learning with job seeker learning data about the job seeker. Here, the job seeker learning data includes job seeker information registered in the job seeker database, information provided by the job seeker, or publicly available information about the job seeker on the network. In the task execution step, an information processing system causes the recruiter task execution model or the job seeker task execution model to execute a task based on an instruction to execute the task to the recruiter task execution model or the job seeker task execution model.

2. In the information processing system described in claim 1, The aforementioned recruiter agent construction step involves preparing training data for each recruiter's department, job type, or job that the recruiter handles, and constructing the aforementioned recruiter task execution model for each department, job type, or job.

3. In the information processing system described in claim 1, An information processing system in which the recruiter learning data includes, as information about recruiters registered in the recruiter database, at least one of the following: data on recruiters' personnel recruitment used in the recruitment management system, and data on recruiters' evaluations of personnel within the organization used in the human resources system.

4. In the information processing system described in claim 1, An information processing system in which the job seeker learning data includes, as information about job seekers registered in the job seeker database, at least one of the job postings viewed or applied for by the job seeker, and the recruitment emails opened or replied to.

5. In the information processing system described in claim 1, The aforementioned job seeker task execution model includes a large-scale language model, In the aforementioned recruiter agent construction step, a recruiter memory is created for each recruiter from the recruiter training data, which the recruiter task execution model references when generating output. The aforementioned job seeker task execution model includes a large-scale language model, In the job seeker agent construction step, a job seeker memory is created for each job seeker from the job seeker training data, which the job seeker task execution model refers to when generating output. An information processing system in which, in the task execution step, the recruiter task execution model or the job seeker task execution model is input with the task along with the recruiter memory or the job seeker memory as reference information, and the recruiter task execution model or the job seeker task execution model is made to execute the task.

6. In the information processing system described in claim 1, In the aforementioned recruiter agent construction step, a recruiter memory representing the characteristics or attributes of each recruiter is created from the recruiter training data, and the recruiter task execution model is constructed using machine learning with the recruiter memory. An information processing system that, in the step of building a job seeker agent, creates a job seeker memory for each job seeker representing the characteristics or attributes of the job seeker from the job seeker learning data, and builds the job seeker task execution model by machine learning using the job seeker memory.

7. In the information processing system described in claim 5, The aforementioned recruiter memory includes at least information about the culture of the organization making the job offer. An information processing system in which the aforementioned job seeker memory contains at least information about the job seeker's personality.

8. In the information processing system described in claim 5, In the recruiter agent construction step, the memory generation model, which is a large-scale language model, is given input information including the input to the recruiter task execution model, the output of the recruiter task execution model for said input, or a combination thereof, and an instruction to create the recruiter memory representing the characteristics or attributes of the recruiter based on said input information, thereby causing the memory generation model to output the recruiter memory. An information processing system that, in the step of building a job seeker agent, inputs to the memory generation model input information including input to the job seeker task execution model, output of the job seeker task execution model for said input, or a combination thereof, and instructions to create the aforementioned job seeker memory representing the characteristics or attributes of the job seeker based on said input information, thereby causing the memory generation model to output the job seeker memory.

9. In the information processing system described in claim 8, The aforementioned processor is configured to perform the following steps: In the recruiter label creation step, the memory generation model is input with the recruiter memory and an instruction to generate a recruiter label that represents the contents of the recruiter memory in natural language, thereby causing the memory generation model to output the recruiter label and present all or part of the recruiter label to the corresponding recruiter. An information processing system that, in the job seeker label creation step, inputs the job seeker memory and an instruction to generate a job seeker label that represents the contents of the job seeker memory in natural language to the memory generation model, thereby causing the memory generation model to output the job seeker label and presenting all or part of the job seeker label to the corresponding job seeker.

10. In the information processing system described in claim 9, In the aforementioned recruiter label creation step, the recruiter requests editing the recruiter label from the corresponding recruiter. In the aforementioned job seeker label creation step, an information processing system receives requests for editing of the job seeker label from the corresponding job seeker.

11. In the information processing system according to claim 10, In the aforementioned recruiter agent construction step, the recruiter task execution model is constructed by machine learning using the edited recruiter labels. The job seeker agent construction step involves an information processing system that constructs a job seeker task execution model using machine learning with the edited job seeker labels.

12. In the information processing system described in claim 1, The aforementioned recruiter task execution model is capable of answering questions about recruiters as a task, The aforementioned job seeker task execution model is capable of answering questions about job seekers as a task, An information processing system that, in the task execution step, receives a question for the employer task execution model or the job seeker task execution model as a task execution instruction, and causes the employer task execution model or the job seeker task execution model to output an answer to the question.

13. In the information processing system according to claim 12, In the task execution step, the information processing system requests the recruiter corresponding to the recruiter task execution model or the job seeker corresponding to the job seeker task execution model to engage in conversation with the user, depending on the content of the conversation between the user and the recruiter task execution model or the job seeker task execution model.

14. In the information processing system according to claim 12, The aforementioned recruiter agent construction step involves an information processing system that constructs the recruiter task execution model by machine learning using further recorded data of interviews between recruitment agencies and recruiters or job seekers.

15. In the information processing system according to claim 12, The aforementioned job seeker agent construction step involves an information processing system that constructs the job seeker task execution model by machine learning using recorded data of interviews between recruitment agencies and employers or job seekers.

16. In the information processing system described in claim 1, The information processing system performs, in the task execution step, a job seeker extraction process in which the job seeker task execution model executes a task to extract job seekers to be recommended to the job seeker corresponding to the job seeker task execution model, or a job search extraction process in which the job seeker task execution model executes a task to extract job postings to be recommended to job seekers.

17. In the information processing system described in claim 16, The aforementioned recruiter task execution model is capable of answering questions about recruiters as a task, The aforementioned job seeker task execution model is capable of answering questions about job seekers as a task, In the task execution step, the information processing system causes the employer task execution model and the job seeker task execution model to converse, and determines which job seekers to recommend to employers or which job openings to recommend to job seekers based on the content of the conversation.

18. In the information processing system described in claim 1, The task execution step involves an information processing system that determines which job seekers to recommend to employers, or which job openings to recommend to job seekers, based on the similarity calculated from the features of the employer training data and the features of the job seeker training data.

19. In the information processing system described in claim 18, The aforementioned job seeker task execution model includes a large-scale language model, In the aforementioned recruiter agent construction step, a recruiter memory is created for each recruiter from the recruiter training data, which the recruiter task execution model references when generating output. The aforementioned job seeker task execution model includes a large-scale language model, In the job seeker agent construction step, a job seeker memory is created for each job seeker from the job seeker training data, which the job seeker task execution model refers to when generating output. In the task execution step, the information processing system determines which job seekers to recommend to employers, or which job openings to recommend to job seekers, based on the similarity calculated from the features of the employer memory and the features of the job seeker memory.

20. In the information processing system described in claim 1, The aforementioned recruiter task execution model is capable of answering questions about recruiters as a task, The aforementioned job seeker task execution model is capable of answering questions about job seekers as a task, An information processing system that, in the task execution step, presents the results of a conversation between the employer task execution model and the job seeker task execution model to at least one of the employers corresponding to the employer task execution model and the job seekers corresponding to the job seeker task execution model.

21. In the information processing system described in claim 1, The aforementioned processor is configured to perform the following steps: In the question registration step, job seekers register pre-questions they have received from employers. In the task execution step, the information processing system causes the employer task execution model, which corresponds to the employer specified by the job seeker, to perform the task of answering the registered pre-questions.

22. In the information processing system described in claim 1, The aforementioned processor is configured to perform the following steps: In the platform agent construction step, a platform task execution model capable of performing tasks that handle information of any employer or any job seeker is constructed using machine learning with the employer training data and the job seeker training data. An information processing system in which, in the task execution step, when causing the recruiter task execution model or the job seeker task execution model to execute a task, the platform task execution model also causes the same task or a task related to that task to execute.

23. In the information processing system described in claim 22, The recruiter training data used in the recruiter agent construction step includes information on recruiters registered in the recruiter database on one platform. The job seeker learning data used in the job seeker agent construction step includes information on job seekers registered in the job seeker database on the platform. The recruiter training data used in the platform agent construction step includes information on multiple recruiters registered in the recruiter database used in the recruiter agent construction step, An information processing system in which the job seeker learning data used in the platform agent construction step includes information on multiple job seekers registered in the job seeker database used in the job seeker agent construction step.

24. Information processing method, An information processing method comprising each step performed by the information processing system according to any one of claims 1 to 23.

25. It is a program, A program for causing a computer to perform each step of the information processing system described in any one of claims 1 to 23.

Images