Recruitment support system, recruitment support method, and program

Abstract

To provide a recruitment support system, method, and program that present combinations of job openings and candidates to a recruiter.SOLUTION: In a recruitment support system 1, a job extraction step executed by a control unit of a server apparatus extracts at least one recommended job opening based on a contract establishment likelihood value calculated from job opening content and first reference information from among job openings registered in a database by a recruiter. The first reference information includes a correlation between job opening content and the contract establishment likelihood value that is a probability of contract establishment for the job opening. A candidate extraction step extracts one or more candidates for the recommended job opening based on a job-seeker matching index calculated from the content of the recommended job opening and second reference information from among job seekers registered in the database. The second reference information includes a correlation between job opening content and the job-seeker matching index indicating a degree of suitability of a job seeker for the job opening. A presentation step associates the recommended job opening and the candidate with each other and presents them to the recruiter.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a recruitment support system, a recruitment support method, and a program. [Background technology]

[0002] As disclosed in Patent Document 1, a technique is known in which a recruiter searches for job seekers who meet desired conditions based on job seeker information registered by the job seeker. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2002-269220 Summary of the Invention [Problem to be solved by the invention]

[0004] In such technology, employers are required to input keywords and other search criteria to search for the job seekers they are looking for.

[0005] In view of the above circumstances, the present invention provides a recruitment support system and the like that can present combinations of job offers and candidates to recruiters. [Means for solving the problem]

[0006] According to one aspect of the present invention, a recruitment support system is provided. The recruitment support system includes a processor. The processor is configured to execute the following steps: In the job offer extraction step, at least one recommended job offer is extracted from job offers registered in a database by a recruiter based on the content of the job offer and an expected value for closing calculated based on first reference information. The first reference information includes a correlation between the content of the job offer and the expected value for closing, which is the probability of closing for the job offer. In the candidate extraction step, at least one candidate for the recommended job offer is extracted from job seekers registered in the database based on the content of the recommended job offer and a job seeker matching index calculated based on second reference information. The second reference information includes a correlation between the content of the job offer and a job seeker matching index indicating the degree of suitability of the job seeker for the job offer. In the presentation step, the recommended job offer and the candidate are associated and presented to the recruiter.

[0007] According to this embodiment, recommended job offers with a high expected success rate can be presented to the recruiter together with candidates who are highly suitable for the recommended job offers. Therefore, by the user, the recruiter, taking action based on the recommended job offers (for example, sending a scout document), the possibility of the job offer being concluded can be increased. Furthermore, by presenting candidates for the recommended job offers, the recruiter can identify job seekers who are likely to conclude a job offer. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a configuration diagram showing a recruitment support system 1. FIG. [Figure 2] 2 is a block diagram showing the hardware configuration of the server device 10. FIG. [Figure 3] FIG. 2 is a block diagram showing the hardware configuration of a recruiting party terminal 20 and a job seeker terminal 30. [Figure 4] 1 is a block diagram showing functions realized by a server device 10 (control unit 11), a recruiting party terminal 20 (control unit 21), and a job seeker terminal 30 (control unit 31). [Figure 5] FIG. 10 is an explanatory diagram showing an example of a procedure for acquiring a second vector. [Figure 6] 10 is a diagram showing an example of a recommended job offer presentation screen RD displayed on the recruiter terminal 20. FIG. [Figure 7] 10 is a diagram showing an example of a job listing screen LD displayed on the recruiter terminal 20. FIG. [Figure 8] FIG. 2 is an activity diagram showing the flow of information processing (recommended job offers and candidate presentation processing) executed by the recruitment support system 1. DETAILED DESCRIPTION OF THE INVENTION

[0009] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the accompanying drawings. Various features shown in the following embodiments can be combined with each other.

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

[0011] In this embodiment, the term "unit" may also include, for example, a combination of hardware resources implemented by a circuit in the broad sense and software information processing that can be specifically realized by these hardware resources. In addition, various types of information are handled in this embodiment, and this information may be represented by, for example, physical values ​​of signal values ​​representing voltages and currents, high and low signal values ​​as a binary bit set consisting of 0 or 1, or quantum superposition (so-called quantum bits), and communication and calculations may be performed on a circuit in the broad sense.

[0012] In addition, a circuit in the broad sense is a circuit realized by at least appropriately combining a circuit, circuitry, a processor, a memory, etc. That is, it includes an application specific integrated circuit (ASIC), a programmable logic device (e.g., a simple programmable logic device (SPLD), a complex programmable logic device (CPLD), and a field programmable gate array (FPGA)), etc.

[0013] 1. Hardware Configuration This section explains the hardware configuration.

[0014] <Recruitment Support System 1> FIG. 1 is a configuration diagram showing a hiring support system 1. The hiring support system 1 comprises a communication line 2, a server device 10, a plurality of recruiter terminals 20, and a plurality of job seeker terminals 30. The server device 10, the recruiter terminals 20, and the job seeker terminals 30 are configured to be able to communicate with each other via the communication line 2. The connection between the server device 10, the recruiter terminals 20, and the job seeker terminals 30 may be wired or wireless.

[0015] The hiring support system 1 constitutes part of a recruitment and job search system used by multiple recruiters (first recruiter U1 and second recruiter U2) and multiple job seekers (first job seeker U3 and third job seeker U4). The hiring support system 1 mainly manages job seeker registration information and job postings. In one embodiment, the hiring support system 1 is comprised of one or more devices or components. These components are described below.

[0016] <Server device 10> 2 is a block diagram showing the hardware configuration of server device 10. Server device 10 includes a control unit 11, a storage unit 12, a communication unit 13, and a communication bus 14. Control unit 11, storage unit 12, and communication unit 13 are electrically connected within server device 10 via communication bus 14.

[0017] <Control unit 11> The control unit 11 processes and controls the overall operations 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 out predetermined programs stored in the storage unit 12. In other words, 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 multiple control units 11 may be provided for each function. A combination of these may also be used.

[0018] <Storage section 12> The memory unit 12 stores various pieces of information defined 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 required information (arguments, arrays, etc.) related to the program operations. The memory unit 12 stores various programs, variables, etc. related to the server device 10 executed by the control unit 11.

[0019] <Communications Department 13> The communication unit 13 is preferably a wired communication means such as USB, IEEE1394, Thunderbolt (registered trademark), or wired LAN network communication, but may also include wireless LAN network communication, mobile communication such as 3G / LTE / 5G, or BLUETOOTH (registered trademark) communication as needed. That is, it is more preferable to implement it as a collection of multiple communication means. That is, the server device 10 may communicate various information from the outside via the communication unit 13 and the network.

[0020] The server device 10 may be an on-premise server or a cloud server. The cloud server device 10 may provide the above-described functions and processes in the form of, for example, SaaS (Software as a Service) or cloud computing.

[0021] <Recruiter Terminal 20> Fig. 3 is a block diagram showing the hardware configuration of the recruiting party terminal 20 and the job seeker terminal 30. As shown in Fig. 3A, the recruiting party terminal 20 comprises a control unit 21, a memory unit 22, a communication unit 23, an input unit 24, an output unit 25, and a communication bus 26. The control unit 21, the memory unit 22, the communication unit 23, the input unit 24, and the output unit 25 are electrically connected within the recruiting party terminal 20 via the communication bus 26. The explanation of the control unit 21, the memory unit 22, and the communication unit 23 is the same as the explanation of each unit in the server device 10, and will therefore be omitted. Note that the recruiting party terminal 20 may also be a terminal operated by a recruitment agency that interacts with job seekers on behalf of the recruiter.

[0022] <Input section 24> The input unit 24 accepts operation inputs made by the user. The operation inputs are transferred as command signals to the control unit 21 via the communication bus 26. The control unit 21 can execute predetermined control or calculations based on the transferred command signals as necessary. The input unit 24 may be included in the housing of the recruiter terminal 20 or may be attached externally. 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, a switch button, a mouse, a trackpad, a QWERTY keyboard, etc. can be used as the input unit 24.

[0023] <Output section 25> The output unit 25 displays a screen of a graphical user interface (GUI) that can be operated by the user. The output unit 25 may be included in the housing of the recruiter terminal 20, or may be attached externally. Specifically, the output unit 25 may be implemented as a display device such as a CRT display, a liquid crystal display, an organic EL display, or a plasma display. It is preferable that these display devices are implemented by selectively using them depending on the type of recruiter terminal 20.

[0024] <Job Seeker Terminal 30> 3B, the job seeker terminal 30 includes a control unit 31, a memory unit 32, a communication unit 33, an input unit 34, an output unit 35, and a communication bus 36. The control unit 31, the memory unit 32, the communication unit 33, the input unit 34, and the output unit 35 are electrically connected via the communication bus 36 inside the job seeker terminal 30. The explanation of the control unit 31, the memory unit 32, the communication unit 33, the input unit 34, and the output unit 35 will be omitted as they are the same as the explanation of each unit in the recruiter terminal 20.

[0025] 2. Functional configuration In this section, the functional configuration of this embodiment will be described. Information processing by the 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 (a processor provided in the recruitment support system 1).

[0026] FIG. 4 is a block diagram showing functions realized by the server device 10 (controller 11), the recruiter terminal 20 (controller 21), and the job seeker terminal 30 (controller 31).

[0027] As shown in Fig. 4A, server device 10 (control unit 11) includes a basic display control unit 111, a job offer extraction unit 112, a candidate extraction unit 113, a presentation unit 114, a scout document management unit 115, and an artificial intelligence unit 120. As shown in Fig. 4B, recruiter terminal 20 (control unit 21) includes a display unit 211 and an operation acquisition unit 212. As shown in Fig. 4C, job seeker terminal 30 (control unit 31) includes a display unit 311 and an operation reception unit 312.

[0028] <Basic display control unit 111> The basic display control unit 111 is configured to display various information on the recruiter terminal 20 and the job seeker terminal 30. For example, the basic display control unit 111 displays a resume and curriculum vitae prepared by the job seeker, a job advertisement and scouting document prepared by the recruiter, etc. on the display unit 211 of the recruiter terminal 20 or the display unit 311 of the job seeker terminal 30.

[0029] Recruiters include organizations such as for-profit corporations (such as companies), non-profit corporations (such as cooperatives and foundations), and public corporations (such as local governments). Recruiters also include recruitment agencies that act as agents of organizations to mediate between job seekers and organizations. Recruitment agencies are also called headhunters or agents.

[0030] <Job extraction section 112> The job offer extraction unit 112 is configured to extract at least one recommended job offer from the job offers registered in the database by the recruiter based on the content of the job offer (job offer advertisement data) and an expected value of contract calculated based on the first reference information. The expected value of contract is a numerical value indicating the likelihood of contract, and it is estimated that the higher the expected value of contract for a job offer, the higher the likelihood of contract.

[0031] A job posting, which is the content of a job offer, lists job information for multiple items such as the name of the position being recruited, job content and working conditions (annual salary, job type, industry, work location, work style, work environment, etc.), application qualifications (skills), desired personality, and appealing points. The job posting may also include items such as the job title, headline, and information about the employer (company size (sales, number of employees, etc.), industry, etc.). Job postings are registered in a recruiter database stored in, for example, memory unit 12. The recruiter database stores the registration information (organizational information) of the user, who is the recruiter, and the job posting created by the recruiter.

[0032] The first reference information includes a correlation between the content of the job offer and an expected closing value, which is the probability of closing a job offer for that job offer. The first reference information is stored, for example, in the memory unit 12. The first reference information is an estimator constructed so that it can input a job posting and output an expected closing value for the job posting. The first reference information may be, for example, a table, function, simple algorithm, etc., that indicates the correlation between feature values ​​(vector data) extracted from the job posting and closing results. The correlation included in the first reference information is constructed, for example, by statistically analyzing data recording closing results for actual job postings.

[0033] The first reference information may be an expected value calculation model that has been trained to input the content of a job offer and output the expected value of a successful contract. In this case, the job offer extraction unit 112 inputs the content of the job offer into the expected value calculation model of the artificial intelligence unit 120 and causes the expected value calculation model to output the expected value of a successful contract. This makes it possible to predict the expected value of a successful contract based on the success of a large number of job offers. The expected value calculation model outputs the expected value of a successful contract as a numerical value between 0 and 1, for example.

[0034] The expected value calculation model is a learning model that uses training data for job postings and data on the success of job offers in those job postings (typically, data on whether a job offer was successful and when it was successful) as training data. The training data for job postings includes master information (numeric values ​​or items selected from preset candidates) such as annual salary, job type, industry, work location, and skills, as well as text information consisting of sentences such as titles and detailed information.

[0035] The expected value calculation model is preferably trained using the content of multiple job offers and whether or not these job offers have been concluded within a predetermined first period. For example, by learning whether or not job offers have been concluded within a period close to the present, the expected value for conclusion can be output that reflects the most recent tendency for conclusion. This allows the user, the job offeror, to be presented with recommended job offers with a higher probability of conclusion.

[0036] The first specified period is, for example, a period within six months from the time of learning. Data on job offers that have been concluded but that were concluded outside the first specified period (for example, a contract concluded one year ago) is used to learn the expected value calculation model as training data for "job offers without contracts," or is excluded from the training data for the expected value calculation model. For example, the expected value calculation model may be trained using data on multiple job offers without contracts and data on multiple job offers that were concluded within the first specified period.

[0037] Furthermore, only job offers whose elapsed time since registration is within a predetermined second designated period may be used as learning targets (teaching data) for the expected value calculation model, regardless of whether the job offer has been concluded or not. In other words, job offers whose elapsed time since registration exceeds the second designated period may be excluded from learning targets for the expected value calculation model. The second designated period may be, for example, one year. The expected value calculation model is subjected to additional learning (updating of the learning model) at regular intervals (for example, once a day) by, for example, the artificial intelligence unit 120, based on data on newly registered job offers and newly concluded job offers.

[0038] The expected value calculation model may be a generative AI including a large-scale language model. In this case, the job offer extraction unit 112 inputs a job posting, inputs a prompt including an instruction to output a contract expectation value to the expected value calculation model, and causes the expected value calculation model to output the contract expectation value. Furthermore, the job offer extraction unit 112 may input, to the expected value calculation model, in addition to the instruction to output the contract expectation value and the job posting, a prompt that inserts, as input and output samples, for example, one or more job posting samples and one or more corresponding contract expectation value samples.

[0039] <Another example of how to calculate expected closing value> The expected value of contract may be calculated based on expected value information regarding the occurrence rate or number of actions for the job posting for which the expected value of contract is to be calculated (hereinafter referred to as the "target job posting"). That is, the job extraction unit 112 may be configured to generate expected value information based on the content of the target job posting and the first reference information. The "actions for the target job posting" include employer actions, which are actions taken by employers based on the target job posting, and job seeker actions, which are actions taken by job seekers based on the target job posting.

[0040] Employer actions include direct actions initiated by the target job posting (e.g., sending a scouting document based on the target job posting) and indirect actions initiated by another action (e.g., sending a scouting document and making a job offer for the job posting to which a job seeker has replied, or closing the job posting for which the scouting document was sent). Job seeker actions also include direct actions initiated by the target job posting (e.g., applying for the job posting) and indirect actions initiated by another action (e.g., replying to a scouting document sent by the employer, accepting a job offer from the employer, or closing the job posting for which a job seeker has applied).

[0041] Specifically, the employer actions that the job offer extraction unit 112 refers to when calculating expected value information should include at least one of sending a scouting document to a job seeker based on the target job posting, the job seeker's reply to the scouting document, and concluding a job offer with the job seeker. This makes it possible to obtain the expected value of each action that begins with the sending of a scouting document from the employer.

[0042] Furthermore, the job seeker actions that the job offer extraction unit 112 refers to when calculating the expected value information may include at least one of an application from a job seeker to a target job posting and a job offer contract for a job seeker who has applied. This makes it possible to obtain the expected value of each action that begins with a job seeker's application.

[0043] Furthermore, the job offer extraction unit 112 may generate expected value information regarding passing a job interview (first interview, second interview, etc.) and receiving a job offer as an action based on the target job advertisement.

[0044] The first reference information used when calculating the expected value information includes a correlation between the content of the job offer and the occurrence or occurrence number of actions in the job offer. The first reference information is an estimator constructed so that it can input the target job posting and output the expected value information (the predicted occurrence rate or predicted occurrence number of actions) for the target job posting.

[0045] The first reference information may be an expected value information output model that has been trained to be capable of inputting the content of the job offer (the target job posting) and outputting expected value information. The expected value information output model constitutes part of the expected value calculation model. In this case, the job offer extraction unit 112 inputs the target job posting into the expected value information output model of the artificial intelligence unit 120, and causes the expected value information output model to output expected value information. This makes it possible to learn and model the tendency of job postings that are likely to result in actions such as closing a deal (deciding to hire the job seeker) based on a model that has learned information from the reference job posting, which is an actually used job posting. This makes it possible to generate expected value information based on whether or not an action actually occurred in past job postings, or the number of times it occurred. This makes it possible to predict the rate and number of actions that will occur, thereby improving the accuracy of the predicted rate and number of actions that will occur.

[0046] The expected value information output model is a learning model that predicts whether or not an action will occur in a target job posting, or the number of times it will occur. The expected value information output model is a learning model that is trained using reference job postings for learning and data on the occurrence rate or number of occurrences of corresponding actions as training data. A "reference job posting" is a job posting that is registered or has been registered in the job posting database and that has actually posted or is currently posting a job. In this case, the parameters of the expected value information output model that have been calculated or tuned through learning correspond to the correlation between the content of the job posting and whether or not an action will occur in the job posting, or the number of times it will occur.

[0047] The expected value information output model may be a generative AI including a large-scale language model. In this case, the job offer extraction unit 112 inputs the target job posting, inputs a prompt including an instruction to predict and output the occurrence rate or number of occurrences of an action in the target job posting, and causes the expected value information output model to output the occurrence rate or number of occurrences of the action. Furthermore, the job offer extraction unit 112 may input, to the expected value information output model, in addition to the instruction to predict and output the occurrence rate or number of occurrences of the action and the target job posting, a prompt that includes, as input and output samples, for example, one or more job posting samples and one or more corresponding action occurrence rate or number samples.

[0048] The job offer extraction unit 112 may generate the probability of a job offer being concluded via the scout document as expected value information. This makes it possible to obtain the probability of a job offer being concluded for the target job posting. The probability of a job offer being concluded may be expressed as a probability such as a percentage, or may be obtained and expressed as a score. A job offer being concluded means that a job seeker accepts a job offer from a recruiter and the job seeker is decided to be employed for the target job posting, and may also be called a job offer decision. The probability of a job offer being concluded may also be rephrased as the ease of conclusion, the expected level of conclusion, the possibility of conclusion, the ease of decision, the expected level of decision, or the possibility of decision. In this case, the job offer extraction unit 112 uses a first expected value information output model as the expected value information output model, which predicts the probability of a job offer being concluded for the target job posting. The first expected value information output model is a learning model that receives the target job posting as input and is trained to output the probability of a job offer being concluded (i.e., the occurrence rate of the action of a job offer being concluded). The first expected value information output model is a learning model trained using reference job postings for learning and corresponding data on whether a job offer has been concluded as training data. In this case, the parameters of the first expected value information output model calculated by training correspond to the correlation between the content of the job posting and the success rate of the job posting. The data on whether a job offer has been concluded is binary data, for example, with "1" indicating a successful conclusion and "0" indicating no successful conclusion. The first expected value information output model trained using such data outputs the success probability as a number between 0 and 1.

[0049] The first expectation information output model may be a generative AI including a large-scale language model. In this case, the job offer extraction unit 112 inputs a target job posting and inputs a prompt including an instruction to predict and output the probability of a job offer being concluded for the target job posting to the first expectation information output model, causing the first expectation information output model to output the job offer conclusion probability. Furthermore, the job offer extraction unit 112 may input a prompt including, as input and output samples, one or more job posting samples and one or more corresponding job offer conclusion probability samples, in addition to the instruction to predict and output the job offer conclusion probability and the target job posting, to the first expectation information output model. When the first expectation information output model is a generative AI, the job offer extraction unit 112 may cause the first expectation information output model to output text representing the job offer conclusion probability, such as "easy to conclude" or "difficult to conclude." The output text is selected based on the relationship between a threshold and the job offer conclusion probability (whether the threshold is exceeded). The threshold is set based on a statistical value, such as the average value of the job offer conclusion probability.

[0050] The first expectation information output model may be trained using reference job postings in which the number of scout documents sent based on the reference job postings exceeds a predetermined first threshold, and the presence or absence of a job offer concluded in the reference job postings. That is, the first expectation information output model may be machine-learned using reference job postings in which a certain number of scout documents have been sent and data on the presence or absence of a job offer concluded in the reference job postings as training data. The first expectation information output model may also be a generative AI including a large-scale language model trained by fine-tuning or the like using reference job postings in which a certain number of scout documents have been sent and data on the presence or absence of a job offer concluded in the reference job postings. Since a certain number of scout documents are often required to be sent to conclude a job offer, job postings in which a small number of scout documents have been sent (or in which no scout documents have been sent) may have a low probability of concluding a job offer. Therefore, excluding reference job postings in which a small number of scout documents have been sent (or in which no scout documents have been sent) from the training data improves the prediction accuracy of the first expectation information output model.

[0051] The first threshold value for the number of scout documents sent for the reference job posting used in training the first expectation information output model is an arbitrary value, such as 5, 10, 20, etc. In other words, for example, only reference job postings for which 5, 10, or 20 or more scout documents have been sent are used in training the first expectation information output model.

[0052] The first expectation information output model may also weight each reference job posting in learning according to the number of scout documents sent. For example, the first expectation information output model may be trained using data on reference job postings that are weighted more heavily the more scout documents sent based on the reference job postings, and whether or not a job offer has been concluded for the reference job posting.

[0053] Furthermore, the first expected value information output model may be trained using reference job postings in which the number of scout documents sent based on the reference job posting exceeds a predetermined first threshold, and whether or not a job offer was concluded for the reference job posting within a predetermined period from the sending of the scout document. In other words, in training the first expected value information output model, among reference job postings in which a certain number of scout documents were sent, those in which a job offer was concluded within a certain period from the sending of the scout document are labeled as "job offer concluded." Furthermore, among reference job postings in which a certain number of scout documents were sent, those in which no job offer was concluded or only concluded after a certain period of time have passed are labeled as "no job offer concluded." These labeled reference job postings are then used as training data. If no conditions are set for the period from the sending of the scout document to the conclusion of a job offer, the number of job offers that are concluded increases over time, potentially making the probability of a job offer being concluded too high. Therefore, by setting a condition for the period from the sending of the scout document to the conclusion of a job offer in the labeling of the training data, the number of reference job postings that have resulted in a job offer can be prevented from continuing to increase, allowing the use of appropriate training data. Furthermore, the conditions for conclusion of a job offer can be aligned between reference job postings registered in the database at a relatively old time and reference job postings registered in the database at a relatively recent time. As a result, the prediction accuracy of the first expectation information output model is improved. The first expectation information output model may be trained using reference job postings that have resulted in a job offer within a predetermined period from the registration or publication of the job posting.

[0054] The period of time for which a job offer has been concluded for labeling the data used for training the first expectation information output model as "job offer concluded" is any period, such as within 6 months, 12 months, or 18 months from the date of sending the scouting document. In other words, only reference job postings for which a job offer has been concluded within, for example, 6 months, 12 months, or 18 months from the date of sending the scouting document are used as data labeled as "job offer concluded."

[0055] If multiple deals have been concluded for one reference job posting, i.e., if multiple job seekers are hired for one reference job posting, the first expected value information output model may treat the number of deals concluded per scouting document (i.e., a value normalized by dividing the number of scouting document transmissions by the number of job offers concluded) as the number of deals concluded for the scouting documents for that reference job posting. The normalized number of scouting document transmissions is used in the above-mentioned decision on whether to accept or reject the data for learning and in the weighting.

[0056] The conditions for the reference job posting used in training the first expectation information output model may further include the period elapsed from the date of registration in the database or the date of publication to job seekers, and the period from the registration or publication date to the sending of the scouting document. The publication date refers to the date the job posting is posted on a job posting / job search service or the date it is published online. For example, the first expectation information output model may use only reference job postings that meet the above-mentioned condition regarding the number of scouting document transmissions and that were registered between 18 months and 6 months prior to the present time and for which the scouting document was sent within 6 months of the registration or publication date as training data. The balance between the number of job seekers and the number of job openings varies depending on the time period, which may affect the likelihood of closing a deal. However, by limiting the time period in which the scouting document was sent, the variation in closing rates due to differences in the balance between the number of job seekers and the number of job openings at each time period can be reduced. Furthermore, the lead time from the registration of the job posting to closing a deal is also taken into account, improving the prediction accuracy of the first expectation information output model.

[0057] The job offer extraction unit 112 may generate the probability of a reply from a job seeker to a scouting message as the expected value information. A reply from a job seeker to a scouting message occurs in a shorter period of time from the registration of the job posting than a contract. Therefore, by obtaining the probability of a reply to a scouting message, a score can be obtained that uses more recent information than the contract conclusion of a job posting. In this case, the job offer extraction unit 112 uses a second expected value information output model that predicts the probability of a reply from a job seeker for a target job posting as the expected value information output model. The second expected value information output model is a learning model that uses the target job posting as input and is trained to output the probability of a reply from a job seeker (i.e., the occurrence rate of the action of replying to a scouting message). In other words, the second expected value information output model is a learning model that is trained using the reference job posting and the corresponding data on whether or not a reply to the scouting message was received as training data. In this case, the parameters of the second expected value information output model calculated by training correspond to the correlation between the content of the job posting and the probability of a reply from a job seeker. Furthermore, the data on whether or not there is a reply to the scouting message is binarized data, with "1" representing the presence of a reply and "0" representing the absence of a reply. The second expected value information output model trained using such data outputs the reply probability as a numerical value in the range from 0 to 1.

[0058] The second expectation information output model may be a generative AI including a large-scale language model. In this case, the job offer extraction unit 112 inputs the target job posting and inputs a prompt including an instruction to predict and output the probability of a reply to the scouting message for the target job posting to the second expectation information output model, causing the second expectation information output model to output the probability of a reply to the scouting message. Furthermore, the job offer extraction unit 112 may input, to the second expectation information output model, a prompt including, for example, one or more job posting samples and one or more corresponding samples of the probability of a reply to the scouting message, in addition to the instruction to predict and output the probability of a reply to the scouting message and the target job posting. When the second expectation information output model is a generative AI, the job offer extraction unit 112 may cause the second expectation information output model to output text representing the probability of a reply, such as "likely to receive a reply" or "unlikely to receive a reply." The output text is selected based on the relationship between a threshold and the probability of a reply (whether or not the threshold is exceeded). The threshold is set based on a statistical value such as the average value of the probability of a reply, for example.

[0059] The second expectation information output model may be trained using reference job postings in which the number of scout documents sent based on the reference job posting is greater than a predetermined second threshold, and the presence or absence of replies to the scout documents in the reference job postings. That is, the second expectation information output model may be machine-trained using reference job postings in which a certain number of scout documents have been sent or more and data on the presence or absence of replies to the scout documents in these reference job postings as training data. Furthermore, the second expectation information output model may be a generative AI including a large-scale language model trained by fine-tuning or the like using reference job postings in which a certain number of scout documents have been sent or more and data on the presence or absence of replies to the scout documents in these reference job postings. This eliminates reference job postings in which a small number of scout documents have been sent (or in which no scout documents have been sent) from the training data, thereby improving the prediction accuracy of the second expectation information output model.

[0060] The second threshold value for the number of scout documents sent for a reference job posting used in training the second expectation information output model is a number (e.g., 5) smaller than the first threshold value for the first expectation information output model. Similarly to the first expectation information output model, the second expectation information output model may weight each reference job posting in training according to the number of scout documents sent. For example, the second expectation information output model may train using data on reference job postings that are weighted more heavily the greater the number of scout documents sent based on the reference job posting, and the presence or absence of replies to the scout documents in the reference job posting.

[0061] Furthermore, the second expectation information output model may be trained using reference job postings in which the number of scout documents sent based on the reference job posting is greater than a predetermined threshold, and the presence or absence of replies to the scout documents in the reference job postings within a predetermined period from the sending of the scout documents. In other words, in training the second expectation information output model, among reference job postings in which a certain number of scout documents have been sent, those that received a reply within a certain period from the sending of the scout documents are labeled as "reply received," and those that received no reply, or only a reply after a certain period, are labeled as "no reply." These are used as training data. In this way, by setting a condition for the period from the sending of the scout document to the reply in the labeling of the training data, the number of reference job postings that received replies is prevented from continuing to increase, thereby improving the prediction accuracy of the second expectation information output model.

[0062] The reply period for labeling the data for training the second expectation information output model as "reply received" is, for example, within 14 days from the sending of the scouting document, which is shorter than the period for concluding a job offer (for example, within 6 months) for labeling the data for training the first expectation information output model as "concluded job offer." In other words, only reference job postings for which a reply is received within 14 days from the sending of the scouting document are used as data labeled as "reply received."

[0063] The conditions for the reference job postings used in training the second expectation information output model may further include the period of time elapsed since the date of registration in the database or the date of publication to job seekers, and the period of time from the registration or publication date to the sending of the scouting document. For example, in addition to satisfying the above-mentioned number of scouting documents sent, the second expectation information output model may use only reference job postings that were registered between 12 months and 14 days ago from the present time and for which a scouting document was sent within one year of the registration or publication date as training data. This allows the probability to be predicted using more recent reference job postings than the first expectation information output model, thereby improving the accuracy of the second expectation information output model.

[0064] The job offer extraction unit 112 may generate a predicted value for the number of scout document transmissions as expected value information. If the number of candidates who fit the contents of the job posting is small and the number of scout documents that the employer can send is small, it may be difficult to recruit personnel even if the probability of a job offer being concluded or the probability of replies to the scout documents is high. However, by generating a predicted value for the number of scout document transmissions, the expected value for the number of scout document transmissions based on the target job posting can be obtained as the expected value for a contract. As a result, it is possible to evaluate the job posting based on the predicted number of scout document transmissions. In this case, the job offer extraction unit 112 uses a third expected value information output model as the expected value information output model, which predicts the number of scout document transmissions for the target job posting. The third expected value information output model is a learning model trained to use the target job posting as input and to output the expected value for the number of scout document transmissions (i.e., the number of occurrences of the action of sending a scout document). In other words, the third expected value information output model is a learning model trained using the reference job posting and the data on the number of scout document transmissions corresponding to it as training data. In this case, the parameters of the third expected value information output model calculated by learning correspond to the correlation between the content of the job posting and the number of sent scouting documents.

[0065] The third expected value information output model may be a generative AI including a large-scale language model. In this case, the job offer extraction unit 112 inputs the target job posting and inputs a prompt including an instruction to predict and output the number of scouting document transmissions based on the target job posting to the third expected value information output model, causing the third expected value information output model to output the expected number of scouting document transmissions. Furthermore, the job offer extraction unit 112 may input, to the third expected value information output model, a prompt that includes, as input and output samples, for example, one or more job posting samples and one or more corresponding samples of the expected number of scouting document transmissions, in addition to the instruction to predict and output the number of scouting document transmissions and the target job posting. When the third expected value information output model is a generative AI, the job offer extraction unit 112 may cause the third expected value information output model to output text that represents the expected number of scouting document transmissions, such as "easy to send" or "difficult to send." The output text is selected based on the relationship between a threshold and the expected number of transmissions (whether the threshold is exceeded). The threshold is set based on a statistical value such as the average number of transmissions, for example.

[0066] The third expectation information output model may be trained using reference job postings that are within a predetermined period of time since their registration in the database or their disclosure to job seekers, and the number of scouting messages sent for those reference job postings. In other words, the third expectation information output model may be machine-trained using reference job postings (including those for which no scouting messages have been sent) that have not yet passed a certain period of time since their registration or disclosure, and data on the number of scouting messages sent for those reference job postings as training data. The third expectation information output model may also be a generative AI including a large-scale language model trained by fine-tuning or other methods using reference job postings that are within a certain period of time since their registration or disclosure, and data on the number of scouting messages sent for those reference job postings. This reduces variations in the number of scouting messages sent due to differences in the balance between job offers and job seekers at the time of job posting registration.

[0067] Furthermore, the third expectation information output model may be trained using a reference job posting and the number of scouting documents sent for that reference job posting within a predetermined period from the registration date or publication date of the reference job posting. That is, in training the third expectation information output model, the number of scouting documents sent within a certain period from the registration date or publication date of the reference job posting is labeled as the "number of scouting documents sent" and used as training data. Therefore, scouting documents sent after a certain period from the registration date or publication date are not included in the "number of scouting documents sent." By labeling the training data in this way, the number of scouting documents sent for each reference job posting is prevented from continuing to increase, thereby improving the prediction accuracy of the third expectation information output model.

[0068] In the data for learning the third expected value information output model, the period for sending scouting documents to be counted as the "number of scouting documents sent" is, for example, within six months from the date of registration or publication of the reference job posting. In other words, only scouting documents sent within six months from the date of registration or publication of the reference job posting are counted as the "number of scouting documents sent."

[0069] The job offer extraction unit 112 calculates a contract expectation value from the expectation value information. Specifically, the job offer extraction unit 112 generates first expectation value information, which is the probability of a job offer being concluded, second expectation value information, which is the probability of a reply from a job seeker, and third expectation value information, which is a predicted value of the number of scout documents to be sent, and calculates a contract expectation value for the job posting based on the first expectation value information, the second expectation value information, and the third expectation value information. This makes it possible to assign an overall score (contract expectation value) to the target job posting that combines the three pieces of expectation value information.

[0070] For example, the job offer extraction unit 112 may calculate the expected value of a deal for a job posting as the average of weighted values ​​for the first expectation information, the second expectation information, and the third expectation information. The weighting of the first expectation information may be smaller than the weighting of the second expectation information and the third expectation information. This allows the calculation of the expected value of a deal to be weighted by placing emphasis on the second expectation information (scout response rate) and the third expectation information (number of scouts sent), which are more recent than the first expectation information (job offer success rate) and also have a correlation with job offer success. As a result, a highly reliable expected value of a deal can be assigned to the target job posting. The job offer extraction unit 112 may also calculate the expected value of a deal from the first expectation information, the second expectation information, and the third expectation information using other methods, rather than weighting.

[0071] The job offer extraction unit 112 may normalize the expected contract value to a value between 0 and 1. Furthermore, the job offer extraction unit 112 may rank the target job postings with labels such as "S," "A," "B," or "none" according to the distribution of the normalized expected contract value. For example, "S" is assigned to a target job posting having an expected contract value that is in the top xx% or more of the distribution of expected contract values.

[0072] The distribution of expected contract values ​​used for ranking is determined based on the evaluation criteria of the target job posting set by the employer. For example, if it is desired to evaluate job postings handled by individuals or departments, the target job postings are ranked using the distribution of expected contract values ​​for job postings handled by individuals or departments. Also, for example, if it is desired to evaluate job postings handled by an organization, the target job postings are ranked using the distribution of expected contract values ​​for job postings within the same organization. Furthermore, for example, if it is desired to check the evaluation of job postings issued by an organization in society as a whole or within a particular industry, the target job postings are ranked using the distribution of expected contract values ​​for all job postings, including job postings from other organizations.

[0073] The job extraction unit 112 may classify each of the first expectation information, second expectation information, and third expectation information into levels such as "high" and "low," and then generate a contract expectation value that combines the level of the first expectation information, the level of the second expectation information, and the level of the third expectation information.

[0074] Alternatively, the job offer extraction unit 112 may input the target job posting to the expectation information output model and cause the expectation information output model to output the expected value of a contract, which is expected value information (for example, the average value of weighted values ​​for the first expectation information, the second expectation information, and the third expectation information). In this case, the job offer extraction unit 112 uses an integrated expected value information output model as the expected value information output model. The integrated expected value information output model is a learning model that takes the target job posting as input and is trained to output the expected value of a contract. In other words, the integrated expected value information output model is a learning model that is trained using the reference job posting and the corresponding expected value of a contract as training data. The integrated expected value information output model may be a generative AI that includes a large-scale language model. In this case, the job offer extraction unit 112 takes the target job posting as input and inputs a prompt including an instruction to calculate and output the expected value of a contract for the target job posting to the integrated expected value information output model, and causes the integrated expected value information output model to output the expected value of a contract. Furthermore, the job offer extraction unit 112 may input to the integrated expected value information output model, in addition to the calculation and output instructions for the expected value of a contract and the target job advertisement, a prompt that inserts, for example, one or more job advertisement samples and one or more corresponding examples of the expected value of a contract as input and output samples. If the integrated expected value information output model is a generative AI, the job offer extraction unit 112 may cause the integrated expected value information output model to output text that indicates the probability of a contract, such as "likely to be concluded" or "unlikely to be concluded."

[0075] The job offer extraction unit 112 may generate a predicted value of the number of applications from job seekers as expected value information. This makes it possible to obtain the expected value of the number of applications from job seekers based on the target job advertisement as the expected value of contracts. In this case, the job offer extraction unit 112 uses a fourth expected value information output model as the expected value information output model. The fourth expected value information output model is a learning model that takes the target job advertisement as input and is trained to output the expected value of the number of applications from job seekers (i.e., the number of occurrences of the action of applying from job seekers). In other words, the fourth expected value information output model is a learning model that is trained using the reference job advertisement and the corresponding data on the number of applications from job seekers as training data.

[0076] The fourth expectation information output model may be a generative AI including a large-scale language model. In this case, the job offer extraction unit 112 inputs a target job posting and inputs a prompt including an instruction to predict and output the number of applications from job seekers based on the target job posting to the fourth expectation information output model, causing the fourth expectation information output model to output the expected number of applications from job seekers. Furthermore, the job offer extraction unit 112 may input, to the fourth expectation information output model, a prompt including, for example, one or more job posting samples and one or more corresponding samples of the expected number of applications from job seekers, in addition to the instruction to predict and output the number of applications from job seekers and the target job posting. When the fourth expectation information output model is a generative AI, the job offer extraction unit 112 may cause the fourth expectation information output model to output text expressing the expected number of applications, such as "apparently likely to receive applications" or "unlikely to receive applications." The output text is selected based on the relationship between a threshold and the expected number of applications (whether the threshold is exceeded). The threshold is set based on a statistical value, such as the average number of applications.

[0077] The job offer extraction unit 112 may generate, as expected value information, the probability of a job offer being concluded after a job seeker applies independently. This makes it possible to obtain the probability of a job offer being concluded for the target job posting as the expected value of the conclusion. In this case, the job offer extraction unit 112 uses the fifth expected value information output model as the expected value information output model. The fifth expected value information output model is a learning model that takes the target job posting as input and is trained to output the probability of a job offer being concluded (i.e., the occurrence rate of the action of a job offer being concluded). In other words, the fifth expected value information output model is a learning model that is trained using reference job postings and the corresponding data on whether or not a job offer is concluded as training data.

[0078] The fifth expectation information output model may be a generative AI including a large-scale language model. In this case, the job offer extraction unit 112 inputs a target job posting and inputs a prompt including an instruction to predict and output the probability of a job offer being concluded for the target job posting to the fifth expectation information output model, causing the fifth expectation information output model to output the job offer conclusion probability. Furthermore, the job offer extraction unit 112 may input, in addition to the instruction to predict and output the job offer conclusion probability and the target job posting, a prompt including, as input and output samples, for example, one or more job posting samples and one or more corresponding job offer conclusion probability samples to the fifth expectation information output model. When the fifth expectation information output model is a generative AI, similar to the first expectation information output model, the job offer extraction unit 112 may cause the fifth expectation information output model to output text representing the job offer conclusion probability, such as "easy to conclude" or "difficult to conclude."

[0079] The fifth expectation information output model may be trained using reference job postings that have received more than a predetermined third threshold in the number of applications from job seekers and whether or not a job offer has been concluded in the reference job posting. In other words, the fifth expectation information output model may be machine-trained using reference job postings that have received a certain number of applications or more and data on whether or not a job offer has been concluded in the reference job posting as training data. The fifth expectation information output model may also be a generative AI that includes a large-scale language model that has undergone fine-tuning or other training using reference job postings that have received more than a certain number of applications and data on whether or not a job offer has been concluded in the reference job posting. This eliminates reference job postings with few applications (or no applications) from the training data, thereby improving the prediction accuracy of the fifth expectation information output model.

[0080] The third threshold value for the number of applications in the reference job posting used in training the fifth expectation information output model is, for example, 100. In other words, only reference job postings with 100 or more applications are used in training the fifth expectation information output model.

[0081] Specifically, the feature quantities of a job posting are used as input to the expected value calculation model or the expected value information output model. The feature quantities of a job posting are data derived from the content written in the job posting, regardless of the format of the job posting. As the feature quantities, for example, a feature vector obtained by converting the content of the target job posting into a vector expressing its features is used. Here, multiple feature vectors may be obtained from one target job posting, and the multiple feature vectors may be used as feature quantities input to the expected value calculation model or the expected value information output model. Note that the feature quantities of a job posting may also be extracted using methods other than vectorization.

[0082] Specifically, the job extraction unit 112 inputs feature quantities including a first vector obtained by vectorizing the sentences included in the job advertisement into the expected value calculation model or the expected value information output model, and causes the expected value calculation model or the expected value information output model to output the expected value of the contract or the expected value information. This eliminates the influence of the job advertisement format (type of items, order of items, etc.) and variations in spelling, and allows the output of the expected value of the contract or the expected value information based on the text information, such as the sentences included in the job advertisement. This improves the accuracy of the expected value of the contract or the expected value information.

[0083] Furthermore, the job extraction unit 112 may input a feature quantity including a second vector obtained by vectorizing the attribute information included in the job advertisement into the expected value calculation model or the expected value information output model, and cause the expected value calculation model or the expected value information output model to output the contract expected value or expected value information. This improves the accuracy of the contract expected value or expected value information based on the category information (annual salary, job type, industry, work location, etc.) included in the job advertisement.

[0084] Furthermore, the job offer extraction unit 112 may input a feature quantity including both the first vector and the second vector into an expected value calculation model or an expected value information output model, and cause the expected value calculation model or the expected value information output model to output a contract expected value or expected value information.

[0085] The expected value calculation model or expected value information output model is trained to take as input features including at least one (preferably both) of a first vector obtained by vectorizing the text included in the job posting and a second vector obtained by vectorizing the attribute information included in the job posting, and to output the expected value of the contract or expected value information.

[0086] The job extraction unit 112 converts the sentences included in the job posting into a first vector, for example, using the following procedure. First, the job extraction unit 112 performs morphological analysis on the text data included in the job posting data and divides the sentences into words. The divided words are then filtered to remove stop words (functional words such as particles and auxiliary verbs) and extract only nouns. Stop words are determined, for example, based on dictionary definitions, frequency of occurrence, etc. Next, the job extraction unit 112 performs word normalization (absorbing spelling variations) on the extracted nouns. Word normalization includes procedures such as standardizing character sets, replacing numbers, and standardizing words using a dictionary. Character set standardization includes standardizing uppercase letters to lowercase, and standardizing half-width kana to full-width kana, etc. Number replacement is the replacement of numbers included in words with a representative symbol representing a number (for example, "0").

[0087] Thereafter, the job offer extraction unit 112 acquires a first vector by using the text of the job advertisement and the words that have been processed as described above, using any vectorization method such as the TF-IDF method.

[0088] The TF-IDF method is a technique for quantifying importance using the frequency of occurrence of words in a sentence. The job search unit 112 calculates the number of times each word in a job posting appears in that job posting. The calculation of the number of times each word appears is performed, for example, using a text analysis model such as Bag of Words. Next, the job search unit 112 converts the number of times each word appears in a job posting into the term frequency (tf). Furthermore, based on the term frequency (tf) of each word in each job posting, the job search unit 112 calculates the inverse document frequency (idf), which represents the rarity of each word's appearance across all job postings from which the data was obtained. Finally, the job search unit 112 calculates the importance (tf-idf) of each word in each job posting from the term frequency (tf) and the inverse document frequency (idf), and generates a first vector whose components are the importance of each word. The dimension of the first vector is the number of words whose importance is being calculated.

[0089] In addition to the TF-IDF method described above, LSI and LDA methods may also be used to quantify each word. Furthermore, the job offer extraction unit 112 may generate a first vector from the text of a job posting using a model that performs vectorization based on distributed representations of words, such as Word2vec or BERT. The TF-IDF method has the advantage of being strong in keyword processing and lightweight. Since keywords that appear in job postings are characteristic, the use of the TF-IDF method is preferable. Furthermore, the TF-IDF method reduces the weight of words (such as "company" and "department") that appear in many job postings, regardless of whether or not the job has been sold, making it possible to determine the feature quantities of a job posting optimally.

[0090] The job offer extraction unit 112 converts the attribute information included in the job posting into a second vector, for example, using the following procedure. The job offer extraction unit 112 converts the attribute information, which is numerical data or categorical (qualitative) data, into a vector, for example, using a method such as one-hot encoding. In one-hot encoding, the job offer extraction unit 112 converts the attribute information, which is numerical data or categorical (qualitative) data, into a one-hot vector, each component of which is 0 or 1. The job offer extraction unit 112 generates the second vector by combining the one-hot vectors of each attribute information. "Attribute information" is information that is input by selecting from predetermined categories or numerical values, such as gender, age, annual income, industry, etc.

[0091] FIG. 5 is an explanatory diagram showing an example of a procedure for acquiring a second vector. For example, as shown in FIG. 5A, the job offer extraction unit 112 converts the industry of each job posting shown in the table on the left into a One-Hot vector shown in the table on the right. In the One-Hot vector of FIG. 5A, if the industry corresponds to the industry assigned to each component, the component is set to "1," and if not, the component is set to "0." If the industry of the job posting corresponds to multiple industries of the components of the One-Hot vector, all of the components are set to "1." Furthermore, as shown in FIG. 5B, for example, the job offer extraction unit 112 converts the annual salary of each job posting shown in the table on the left into a One-Hot vector shown in the table on the right. In the One-Hot vector of FIG. 5B, the annual salary is divided into multiple annual salary bands at certain intervals, and if the annual salary corresponds to the annual salary band assigned to each component, the component is set to "1," and if not, the component is set to "0." If the annual salary in the job posting falls within multiple annual salary ranges of the components of the One-Hot vector, multiple components will be set to "1".

[0092] The job extraction unit 112 may vectorize the job posting using a vectorization model, which is a generative AI including a large-scale language model. In this case, the job extraction unit 112 inputs the job posting, inputs a prompt including instructions to vectorize and output the job posting to the vectorization model, and causes the vectorization model to output a first vector and / or a second vector. Furthermore, in addition to the job posting and the instructions to vectorize and output the job posting, the job extraction unit 112 may input a prompt into the vectorization model that inserts, as input and output samples, for example, one or more job posting samples and one or more corresponding first vector and / or second vector samples.

[0093] The job offer extraction unit 112 may extract job offers with a contract expectation value equal to or greater than a predetermined threshold as recommended job offers. This makes it possible to extract recommended job offers on the condition that the possibility of contract is at least a certain level. Therefore, job offers with a high possibility of contract can be preferentially presented to the user (employer). For example, if the contract expectation value is normalized to a value between 0 and 1, the job offer extraction unit 112 extracts job offers with a contract expectation value of 0.8 or greater as recommended job offers.

[0094] The threshold value of expected value of contract may also be defined as a percentage (top percent) of the highest expected value of contract for a specific population of job offers. The "specific population of job offers" consists of all valid job offers (i.e., job offers currently being offered) registered in the job offer database, or all valid job offers registered by users who are job seekers. For example, the job offer extraction unit 112 may extract job offers with expected value of contract within the top 30% as recommended job offers.

[0095] Furthermore, the job offer extraction unit 112 may prepare multiple thresholds for the percentage from the top. For example, the job offer extraction unit 112 may first extract job offers with a contract expectation value equal to or greater than a first threshold (e.g., the top 30%) as recommended job offers, and if the number of extracted recommended job offers does not meet a predetermined reference number, extract job offers with a contract expectation value equal to or greater than a second threshold (e.g., the top 50%) as recommended job offers. If there are no job offers with a contract expectation value equal to or greater than the second threshold, the job offer extraction unit 112 determines that there are no recommended job offers.

[0096] Furthermore, the job offer extraction unit 112 may also set as recommended jobs those jobs that have been given a rank of a certain level or higher based on the distribution of expected contract values ​​described above.

[0097] The job offer extraction unit 112 may extract recommended job offers from job offers registered in the database by job recruiters based on the expected value of the job offer and the history of sending scouting documents based on the job offer to job seekers. This allows job offers with lower utilization rates (low scouting frequency or no scouting) to be extracted from job offers with a high probability of success and presented to the user, the recruiter. As a result, it is possible to encourage job recruiters to take advantage of job offers that are not being used often but have a high probability of success (searching for job seekers, sending scouting documents, etc.) in accordance with the demands of job seekers, thereby enabling efficient recruitment activities.

[0098] Specifically, the job vacancy extraction unit 112 may extract recommended job vacancies based on a job vacancy score calculated for each job vacancy. The job vacancy score increases according to the magnitude of the expected value of closing and its weighting, and decreases according to the magnitude of the number of scout documents sent during a predetermined evaluation period and its weighting. The weighting of the expected value of closing is greater than the weighting of the number of sendings. This makes it possible to extract recommended job vacancies that emphasize the possibility of closing while also taking into account the utilization rate of the job vacancy. Therefore, compared to when emphasizing low utilization rate, when an employer conducts recruitment activities such as sending scout documents for recommended job vacancies, the possibility of closing a job can be increased.

[0099] For example, the job offer extraction unit 112 calculates a closing score by multiplying the closing expectation value, normalized to a value between 0 and 1, by a first weighting coefficient, and an operation score by multiplying the number of scout document transmissions during the evaluation period, normalized to a value between 0 and 1, by a second weighting coefficient, and determines the job offer score as a value obtained by subtracting the operation score from the closing score. Here, the first weighting coefficient is greater than the second weighting coefficient. The evaluation period can be, for example, within six months or one year from the present time. The normalized number of scout document transmissions can be obtained, for example, by setting the number of transmissions of the job offer with the largest number of scout document transmissions during the evaluation period among the job offers registered by the user (recruiter) as a reference value, and dividing the number of scout document transmissions of each job offer during the evaluation period by this reference value.

[0100] Furthermore, the job offer extraction unit 112 may count the number of scout document transmissions using weighting that changes depending on the time of year. For example, the job offer extraction unit 112 may count the number of scout document transmissions so that the further away the scout document transmission date and time is from the present time (the less recently the job offer is in operation), the smaller the count number of the scout document transmissions (for example, a scout document sent more than one month ago may be counted as 0.5 times instead of 1 time).

[0101] The job offer extraction unit 112 may also determine the performance score by normalizing the inverse of the number of scout documents sent during the evaluation period to a value between 0 and 1, and determine the job offer score by multiplying the performance score by the expected value for success. In this case, the performance score becomes closer to 0 as the number of sendings increases (the job offer is more active).

[0102] The recruitment score may be calculated using a recruitment performance parameter other than the number of scout documents sent from recruiters to job seekers, such as the number of applications from job seekers for a job posting, the number of times a job posting is viewed by job seekers, or the number of times a job posting is added to a bookmark (favorite) list by job seekers, instead of the number of scout documents sent. In other words, the recruitment score may be a score that increases according to the magnitude of the expected value of contract and its weighting, and decreases according to the magnitude of the recruiting performance parameter during a predetermined evaluation period and its weighting. The weighting of the expected value of contract is greater than the weighting of the recruiting performance parameter.

[0103] Furthermore, the job score may be a score that increases according to the magnitude of the expected value of closing and its weighting, and also according to the magnitude of the job operation parameter during a predetermined evaluation period and its weighting. This makes it possible to recommend job openings with a high expected value of closing from among job openings with high operation rates, thereby efficiently identifying job openings that should be focused on.

[0104] The job offer extraction unit 112 extracts, as recommended jobs, jobs for which the job offer score calculated from the contract score and the operation score is equal to or greater than a predetermined threshold, for example.

[0105] The job offer extraction unit 112 may calculate, for each job offer, a first job offer score and a second job offer score, which have different weightings for at least one of the expected value for closing and the number of scout documents sent during the evaluation period, and extract, as recommended job offers, at least one first recommended job offer extracted based on the first job offer score and at least one second recommended job offer extracted based on the second job offer score. This makes it possible to extract, for example, a first recommended job offer that emphasizes the possibility of closing and a second recommended job offer that emphasizes utilization rate more than the first recommended job offer. This allows recommended job offers to be presented to the user (the employer) from multiple perspectives, and appropriate job offers can be stably proposed, which will lead to job offer activity (the execution of recruiting activities), such as the sending of scout documents.

[0106] The first and second job scores may differ in both the first weighting coefficient (weighting for expected closing value) and the second weighting coefficient (weighting for the number of scout documents sent) used in calculation. Alternatively, only one of these may differ. For example, the first weighting coefficient of the first job score may be the same as the first weighting coefficient of the second job score, and the second weighting coefficient of the second job score may be greater than the second weighting coefficient of the first job score. In this case, the second job score places more emphasis on utilization rate than the first job score. Alternatively, either the second weighting coefficient of the first job score or the second weighting coefficient of the second job score may be zero. In other words, one of the first job score and the second job score may be a score that does not take utilization rate (the number of scout documents sent) into consideration.

[0107] The job offer extraction unit 112, for example, extracts at least one job offer whose first job offer score is equal to or greater than a predetermined threshold as a first recommended job offer, and extracts at least one job offer whose second job offer score is equal to or greater than a predetermined threshold as a second recommended job offer.

[0108] In addition, the job extraction unit 112 may further calculate a third job score for each of the first job score and the second job score, in which at least one of the first weighting coefficient or the second weighting coefficient is different, and extract a third recommended job based on the third job score.

[0109] The job offer extraction unit 112 extracts recommended job offers (calculates job offer scores) at predetermined intervals. The interval at which the job offer extraction unit 112 extracts recommended job offers is, for example, once a day. The first reference information (expected value calculation model) may be updated in accordance with the extraction of recommended job offers by the job offer extraction unit 112. In other words, the first reference information may be updated using the latest data (newly acquired information on job offers or job seekers) at each interval at which the extraction of recommended job offers is performed.

[0110] <Candidate extraction section 113> The candidate extraction unit 113 is configured to extract at least one candidate for a recommended job offer from among job seekers registered in the database, based on a job seeker matching index calculated based on the content of the recommended job offer (job posting data) and the second reference information. The job seeker matching index is a numerical value indicating the degree of match (matching degree) between the attributes of a job seeker and the content of the recommended job offer, and it is estimated that the larger the job seeker matching index, the higher the suitability (probability of passing the screening) for the recommended job offer.

[0111] The second reference information includes a correlation between the content of a job posting and a job seeker matching index that indicates the degree of suitability of a job seeker for the job posting. The second reference information is stored, for example, in the storage unit 12. The second reference information is an estimator constructed so as to be able to input a job posting and output a job seeker matching index for each job seeker registered in the database. The second reference information may be, for example, a table, a function, a simple algorithm, or the like that indicates a correlation between features (vector data) extracted from the job posting, features (vector data) of the job seeker's registered information, and the job seeker matching index for the job posting. The correlation included in the second reference information is constructed, for example, by statistically analyzing data on the features of the job posting, the features of the job seeker, and the job seeker matching index that are associated with each other.

[0112] The vectorization of the job seeker's registration information is performed in the same manner as the vectorization of the job posting described above. Specifically, the candidate extraction unit 113 obtains, by using a vectorization model, which is a generative AI including a large-scale language model, a first vector obtained by vectorizing sentences such as job content included in the job seeker's registration information, and a second vector obtained by vectorizing attribute information such as experienced job types, experienced industry, desired job type, desired industry, and desired annual salary.

[0113] The job seeker's registration information includes the job seeker's resume, curriculum vitae, and other profile information. A "resume" is a document that mainly describes the job seeker's profile, current situation, educational background, work history, desired working conditions, etc., while a "curriculum vitae," also known as a résumé, is a document in which the job seeker conveys to the employer his or her work history, experience, skills, qualifications, etc. related to his or her past work. The registration information may also include the job seeker's desired conditions (desired industry, desired job type, etc.).

[0114] The candidate extraction unit 113 calculates a job seeker matching index for, for example, job seekers registered in the job seeker database who can receive a scout document from one recommended job posting (who can apply for the recommended job posting). Specifically, the candidate extraction unit 113 calculates a job seeker matching index for each job seeker using the correlation of reference job seekers included in the second reference information (job seeker matching index for job postings). A reference job seeker is another job seeker with the same or similar attributes included in the registration information registered in the job seeker database, or another job seeker with a similar behavioral history regarding job postings.

[0115] Examples of "attributes included in the registered information" include items such as occupation, industry, skills, and qualifications. The similarity of attributes in the registered information is determined based on "first similarity judgment information" (information defining the similarity range of attributes) included in the second reference information. The first similarity judgment information may define at least one similar keyword for a keyword indicating an attribute, or may define a similarity criterion for a feature obtained by vectorizing a category, keyword, or sentence indicating an attribute. The similarity criterion for a feature is a threshold for the difference (vector distance) between the feature values ​​of two attributes; for example, attributes whose difference in feature value is less than the threshold (or whose cosine similarity is greater than or equal to the threshold) are determined to be similar to each other.

[0116] The first similarity determination information may also include information on pre-grouping similar attributes. In this case, attributes belonging to the same group are determined to be similar to each other. For example, in the first similarity determination information, "public relations," "marketing," "CRM," etc. are pre-defined as similar attributes (occupations), and are grouped into a group named "marketing." The candidate extraction unit 113 determines that occupations belonging to the same group are similar occupations.

[0117] The similarity of the behavioral histories of job seekers is determined based on "second similarity determination information" (information defining the range of similarity in the behavioral histories) included in the second reference information. The second similarity determination information defines, for example, a threshold for determining similarity based on the number of common job postings among the job postings that two job seekers have viewed, applied for, added to a bookmark list, or replied to a scouting message sent based on the job posting. In other words, two job seekers who have viewed a number of common job postings equal to or greater than the threshold are determined to have similar behavioral histories based on the second similarity determination information. The second similarity determination information may also define criteria for determining similarity based on a combination of multiple types of behavioral histories (for example, a similarity determination based on the number of common job postings among the job postings applied for and the number of common job postings among the job postings replied to the scouting message). For example, a determination formula that compares a value obtained by adding weights assigned to the number of common job postings for each behavioral history with a threshold is used as such a criterion.

[0118] The second reference information may be a matching index calculation model that has been trained to input the content of a job offer and output a job seeker matching index for each job seeker. In this case, the candidate extraction unit 113 inputs the content of the recommended job offer into the matching index calculation model of the artificial intelligence unit 120 and causes the matching index calculation model to output a job seeker matching index. This makes it possible to calculate a job seeker matching index based on the content of a large number of job offers and information on job seekers. The matching index calculation model outputs the job seeker matching index as a numerical value between 0 and 1, for example.

[0119] The matching index calculation model is a learning model that uses as training data at least one of the following data: data on a job posting for learning, data on information about a job seeker for learning (registration information, history of actions related to job postings, etc.), and data on the job seeker matching index of the job seeker for the job posting.

[0120] The matching index calculation model may be a learning model that learns the relationship between the contents of multiple job offers (job postings) and job seekers who have a history of actions related to the job offers. In this case, the matching index calculation model outputs a job seeker matching index as the degree of similarity between the input job offer and a job seeker who has a history of actions related to similar job offers. "Actions related to job offers" refers to actions from employers to job seekers based on the job offer. This allows the matching index calculation model to calculate a job seeker matching index that reflects the likelihood of actions based on the input recommended job offers occurring.

[0121] The "job-related action" may be receiving a scouting document or passing a screening. This makes it possible to extract as candidates job seekers who have received a scouting document from a job with content similar to the recommended job, or who have passed a screening (document screening, interview screening, etc.).

[0122] The matching index calculation model may be a generative AI including a large-scale language model. In this case, the candidate extraction unit 113 receives as input the job posting of the recommended job offer and information about the job seeker for whom the index is to be calculated (such as registration information and a history of actions related to the job offer), inputs a prompt including an instruction to output a job seeker matching index to the matching index calculation model, and causes the matching index calculation model to output the job seeker matching index. Furthermore, the candidate extraction unit 113 may input, to the matching index calculation model, in addition to the instruction to output the job seeker matching index and the job posting and job seeker information, a prompt that includes, as input and output samples, for example, one or more samples of the job posting and job seeker information and one or more corresponding samples of the job seeker matching index.

[0123] The candidate extraction unit 113 may extract, as candidates, job seekers whose job seeker matching index is equal to or greater than a predetermined threshold. This allows candidates to be extracted on the condition that the degree of suitability for the recommended job offer is equal to or greater than a certain level. Therefore, job seekers who are highly suitable for the recommended job offer can be preferentially presented to the user (employer). For example, if the job seeker matching index is normalized to a value between 0 and 1, the candidate extraction unit 113 extracts, as candidates, job seekers whose job seeker matching index is 0.8 or greater. The threshold may also be defined as a percentage (top percent) of the top job seekers among multiple job seekers whose job seeker matching index is calculated. For example, the candidate extraction unit 113 may extract, as candidates, job seekers whose job seeker matching index is within the top 30%.

[0124] The candidate extraction unit 113 may extract candidates from among the job seekers registered in the database based on the job seeker matching index of the job seeker and the number of activities of the job seeker for any job offer. This makes it possible to further extract active job seekers from among job seekers who are highly suitable for the recommended job offers and present them to the user, who is the employer.

[0125] "Any job offer" means any one or more job offers registered in the database, including recommended job offers and job offers other than recommended job offers. "Activities related to a job offer" includes not only actions related to a specific job offer, such as viewing a job offer, applying for a job offer, adding a job offer to a bookmark list, or replying to a scouting document sent based on a job offer, but also actions related to all job offers, such as logging in to the Recruitment Support System 1 and editing one's own registered information.

[0126] Specifically, the candidate extraction unit 113 may extract candidates based on a job seeker score calculated for each job seeker. The job seeker score increases according to the magnitude of the job seeker matching index and its weighting, and also according to the magnitude of the number of activities for a given job offer during a predetermined evaluation period and its weighting. The weighting of the job seeker matching index is greater than the weighting of the number of activities. This makes it possible to extract candidates that emphasize the degree of suitability while also taking into account the activity rate of the job seeker.

[0127] For example, the candidate extraction unit 113 calculates a matching score by multiplying the job seeker matching index, normalized to a value between 0 and 1, by a third weighting coefficient, and an activity score by multiplying the number of activities during the evaluation period, normalized to a value between 0 and 1, by a fourth weighting coefficient, and determines the job seeker score as the sum of the matching score and the activity score. Here, the third weighting coefficient is greater than the fourth weighting coefficient. The evaluation period may be, for example, within one month or within three months from the present time. The normalized number of activities may be obtained, for example, by setting the number of activities of the job seeker with the largest number of activities during the evaluation period among the job seekers whose job seeker matching indexes have been calculated as a reference value and dividing the number of activities of each job seeker during the evaluation period by the reference value. The candidate extraction unit 113 may also determine the job seeker score as the multiplication of the matching score and the activity score.

[0128] Furthermore, the candidate extraction unit 113 may count the number of activities of a job seeker using weighting that changes depending on the time period. For example, the candidate extraction unit 113 may count the number of activities of a job seeker so that the further away the date and time of the job seeker's activity is from the present time, the smaller the count number of the activity (for example, an activity that occurred more than one month ago may be counted as 0.5 times instead of 1 time).

[0129] The candidate extraction unit 113 extracts, for example, job seekers whose job seeker scores calculated from the matching scores and activity scores are equal to or greater than a predetermined threshold as candidates.

[0130] The candidate extraction unit 113 may calculate, for each job offer, a first job applicant score and a second job applicant score, which have different weightings for at least one of the matching index and the number of activities during the evaluation period, and extract, as candidates, at least one first candidate extracted based on the first job applicant score and at least one second candidate extracted based on the second job applicant score. This makes it possible to extract, for example, a first candidate who places emphasis on aptitude and a second candidate who places emphasis on activity more than the first candidate. This makes it possible to present candidates from multiple perspectives to the user, the employer.

[0131] The candidate extraction unit 113 extracts candidates (calculates job seeker scores) at predetermined intervals. The interval at which the candidate extraction unit 113 extracts candidates is, for example, one day. The timing or interval at which the candidate extraction unit 113 extracts candidates may be the same as or different from the timing or interval at which the job offer extraction unit 112 extracts recommended job offers. The second reference information (matching index calculation model) may be updated in accordance with the extraction of candidates by the candidate extraction unit 113. In other words, the second reference information may be updated using the latest data (newly acquired information on job offers or job seekers) at each interval at which candidate extraction is performed.

[0132] <Presentation part 114> The presentation unit 114 is configured to present at least one recommended job offer extracted by the job offer extraction unit 112 and at least one candidate extracted by the candidate extraction unit 113 in association with each other to the recruiter.

[0133] The presentation unit 114 presents the recommended job by, for example, displaying part of the information contained in the job posting of the recommended job (job title (position name), job number (ID), conditions, etc.) on the recruiter terminal 20. In addition, the presentation unit 114 presents the candidate by, for example, displaying part of the information contained in the candidate's registration information (name, organization (company name), age, current job type, job title, etc.) on the recruiter terminal 20.

[0134] "Presenting in correspondence" includes, for example, displaying information about recommended job offers and information about corresponding candidates side by side, vertically or horizontally, preparing an object (e.g., a frame) for each recommended job offer and displaying the corresponding candidate within this object, and displaying information about the corresponding recommended job offer (e.g., characters, labels, etc. indicating the job offer name, job offer number, etc.) attached to the candidate.

[0135] The presentation unit 114 may present all of the multiple recommended job offers extracted by the job offer extraction unit 112, or may present some of the multiple recommended job offers. Furthermore, when the job offer extraction unit 112 extracts first recommended job offers and second recommended job offers, the presentation unit 114 may present at least one first recommended job offer and at least one second recommended job offer to the employer. Furthermore, the presentation unit 114 may highlight, by coloring, changing the font, or the like, those recommended job offers with a high expected value for closing (for example, those with an expected value for closing equal to or greater than a predetermined value, those ranked high in expected value for closing) among the presented job offers.

[0136] The presentation unit 114 may present the recommended job offers by assigning a label according to the expected value of contract. This makes it possible to indicate to the user, the recruiter, the level of demand for each recommended job offer. The "label" may include, for example, a numerical value indicating the magnitude of the expected value of contract or the job offer score, a graphic representation of the numerical value in the form of a graph, bar, or the like, a character or graphic indicating a rank (e.g., "Rank A," "Rank B," etc.) assigned according to the magnitude of the expected value of contract or the job offer score, or a character or graphic indicating that the expected value of contract or the job offer score is equal to or greater than a predetermined value (e.g., "recommended job offer," etc.).

[0137] The presentation unit 114 may present information indicating the candidate's activity status as candidate information. This allows the user, the employer, to know the candidate's activity status. The "activity status" may include, for example, the most recent login date and time, the number of views of the job posting, the number of replies to the scouting document, the number of applications for the job posting, etc., during a predetermined period.

[0138] FIG. 6 is a diagram showing an example of a recommended job posting screen RD displayed on the recruiter terminal 20. On the recommended job posting screen RD, a plurality of recommended job postings RI and a plurality of candidate information CI are arranged as card-shaped objects. In the example of FIG. 6, three recommended job postings RI are displayed, and two candidate information CIs are associated with each recommended job posting RI. Specifically, the candidate information CIs are arranged to the right of the corresponding recommended job postings RI. In addition, some candidate information CIs are assigned a label LB indicating that the expected value of closing is equal to or greater than a predetermined value.

[0139] By selecting any candidate information CI displayed on the recommended job posting screen RD, the user (employer) can view the registered information of the corresponding candidate (job seeker), create a scout document for the candidate, etc. In other words, by operating and inputting the object of the candidate information CI, a screen for viewing the registered information of the candidate, a screen for creating a scout document, etc. are displayed on the employer terminal 20.

[0140] The presentation unit 114 may accept input of an instruction to present recommended job offers from a user (recruiter) and present the recommended job offers and candidates on the recruiter terminal 20, or may present the recommended job offers and candidates on the recruiter terminal 20 in response to the occurrence of a predetermined event or at a predetermined timing. Examples of predetermined events include the user (recruiter) logging in to the recruitment support system 1, and the display of a job offer management screen on the recruiter terminal 20.

[0141] The presentation unit 114 may select and present some of the recommended job offers from the extracted plurality of recommended job offers every predetermined period. This allows new recommended job offers to be presented to the user (employer) every predetermined period, thereby increasing motivation to send scout documents. The period for switching recommended job offers is, for example, one day. In other words, the presentation unit 114 may present recommended job offers on a daily basis. This can encourage the user (employer) to log in to the recruitment support system 1 every day.

[0142] Specifically, the presentation unit 114 selects the number of recommended job offers (e.g., three) as job offers to be presented from the multiple recommended job offers extracted by the job offer extraction unit 112, and displays the job offers to be presented on the recruiter terminal 20. The selection of job offers to be presented is performed, for example, based on the expected closing value or job offer score of each job offer. For example, the presentation unit 114 may select job offers to be presented from the multiple recommended job offers in descending order of expected closing value or job offer score, and after a predetermined period of time, may repeat the process of selecting job offers to be presented from the remaining recommended job offers in descending order of expected closing value or job offer score so as not to select the same job offer to be presented multiple times. For example, the presentation unit 114 may present recommended job offers with the first, second, and third highest expected closing values ​​or job offer scores on the first day, and present recommended job offers with the fourth, fifth, and sixth highest expected closing values ​​or job offer scores on the second day.

[0143] The presentation unit 114 may also assign a probability (selection probability) to each of the multiple recommended job offers that it will be selected as a job offer to be presented, randomly select a job offer to be presented based on the selection probability, and after a predetermined period of time, repeatedly select a job offer to be presented from the remaining recommended job offers based on the selection probability. The selection probability is set according to the expected value of contract or the job offer score, and is set, for example, to 40% for a recommended job offer with the highest expected value of contract or job offer score, and 30% for a recommended job offer with the second highest expected value of contract or job offer score. The selection probability of each recommended job offer may also be set randomly.

[0144] The presentation unit 114 may select at least one job offer to be presented from each group of multiple recommended job offers extracted based on job offer scores that are weighted differently at the time of calculation. For example, the presentation unit 114 may select at least one job offer to be presented from each of a first group consisting of multiple first recommended job offers (job offers extracted using the first job offer score), a second group consisting of multiple second recommended job offers (job offers extracted using the second job offer score), and a third group consisting of multiple third recommended job offers (job offers extracted using the third job offer score). In other words, the presentation unit 114 may select and present a first recommended job offer, a second recommended job offer, and a third recommended job offer from each of the first group, the second group, and the third group for each predetermined period.

[0145] If the number of extracted recommended job offers is equal to or less than the number of presented job offers, all recommended job offers are presented, and the presented recommended job offers are not updated until new recommended job offers are extracted by the job offer extraction unit 112. Also, if the job offer extraction unit 112 determines that there are no recommended job offers (i.e., if no recommended job offers could be extracted), the presentation unit 114 does not present the recommended job offers to the employer. Also, as the displayed recommended job offers are switched, the candidates corresponding to the recommended job offers are also switched.

[0146] The presentation unit 114 may select and present some candidates from the extracted candidates at predetermined intervals. This allows new candidates to be presented to the user (employer) at predetermined intervals, thereby increasing motivation to send scouting documents. The period for switching candidates is, for example, one day. In other words, the presentation unit 114 may present candidates on a daily basis.

[0147] The period for switching candidates may be the same as or different from the period for switching recommended job offers. For example, the presentation unit 114 may change only the candidates presented in association with the recommended job offers without changing the presented recommended job offers by setting the period for switching candidates to be shorter than the period for switching recommended job offers.

[0148] Specifically, the presentation unit 114 selects the number of candidates to be presented (e.g., two) as candidates to be presented from among the multiple candidates extracted by the candidate extraction unit 113, and displays the candidates to be presented on the employer terminal 20. The selection of candidates to be presented is performed, for example, based on the job seeker matching index or job seeker score of each candidate. For example, as in the case of recommended job offers, the presentation unit 114 may select candidates to be presented from among the multiple candidates in descending order of job seeker matching index or job seeker score, and after a predetermined period, may repeat the process of selecting candidates to be presented from among the remaining candidates in descending order of job seeker matching index or job seeker score so that the same job offer to be presented is not selected multiple times. Furthermore, as in the case of recommended job offers, the presentation unit 114 may assign a probability of being selected as a candidate to be presented (selection probability) to each of the multiple candidates, randomly select a candidate to be presented based on the selection probability, and after a predetermined period, repeat the process of selecting a candidate to be presented from among the remaining candidates based on the selection probability.

[0149] The presentation unit 114 may exclude from the candidates to be presented those candidates to whom the user, the recruiter, has sent a scout document based on the corresponding recommended job offer within a predetermined period (for example, within one month from the present time) among the candidates extracted by the candidate extraction unit 113. This prevents candidates to whom a scout document has already been sent from being presented to the recruiter.

[0150] The presentation unit 114 may switch only the candidates to be presented at predetermined periods without switching the recommended job offers to be presented at predetermined periods (i.e., keeping the recommended job offers to be presented fixed until new recommended job offer extraction results are obtained).

[0151] The job offer extraction unit 112 may extract recommended job offers, and the candidate extraction unit 113 may extract candidates in accordance with the period for switching the recommended job offers and candidates presented by the presentation unit 114. In other words, the timing for switching the recommended job offers and candidates presented may be synchronized with the timing for extracting the recommended job offers and candidates. For example, if the presentation unit 114 switches the recommended job offers and candidates on a daily basis, the job offer extraction unit 112 and the candidate extraction unit 113 extract recommended job offers and candidates (calculate the job offer score and job seeker score) once a day. The presentation unit 114 selects recommended job offers to be presented and candidates to be presented from the recommended job offers and candidates extracted on a daily basis.

[0152] In this way, when switching the recommended job offers and candidates to be presented while sequentially extracting recommended job offers and candidates, the presentation unit 114 may select the recommended job offers to be presented and the candidates to be presented from among the extracted recommended job offers and candidates that have not been presented within a predetermined period (for example, the previous day or the most recent week) using the procedure described above. In other words, the presentation unit 114 may perform processing to remove recommended job offers and candidates that have been presented within a predetermined period from the extracted recommended job offers and candidates. In this case, the job offer extraction unit 112 and the candidate extraction unit 113 may additionally extract new recommended job offers and candidates to match the number of recommended job offers and candidates that have been removed.

[0153] The presentation unit 114 may also display labels according to expected contract values ​​for job vacancies registered by the user who is a job seeker in situations other than when presenting recommended job vacancies. For example, the presentation unit 114 may assign labels to a list of job vacancies registered by the user who is a job seeker (job vacancy list). This allows the user who is a job seeker to know which job vacancies are in high demand (high expected contract value) from among the job vacancies they have.

[0154] FIG. 7 is a diagram showing an example of a job listing screen LD displayed on the recruiter terminal 20. The job listing screen LD displays a list of information (status) for multiple job listings. The job listing screen LD displays all job listings registered by the recruiter, or job listings extracted based on search criteria (i.e., as search results). In the example of FIG. 7, a label LB indicating a high expected value of contract is added to the right of the position name of a job listing with a contract expectation value equal to or greater than a predetermined value.

[0155] <Scout Document Management Department 115> The scout document management unit 115 is configured to create scout documents based on job offers and send the scout documents to job seekers. The scout document management unit 115 creates scout documents by accepting selection input from the recruiter terminal 20 of the job offer from which the scout document is to be created and the job seeker to whom the scout document is to be sent. For example, when a candidate is selected on the recommended job offer presentation screen RD of FIG. 6, the scout document management unit 115 accepts the candidate as the job seeker to whom the scout document is to be sent and the recommended job offer corresponding to the candidate as the source of the scout document. The scout document may be sent to the job seeker together with the job offer advertisement by attaching the job offer advertisement for the recommended job offer. Alternatively, the contents of the recommended job offer may be inserted into the scout document.

[0156] The scouting document may be created by the user (job seeker) inputting characters, or by using a learning model that has been trained to take a job posting as input and output a scouting document, or by a combination of these.

[0157] The created scout document is sent to the job seeker as a scout email. The scout document management unit 115 manages the number of scout emails that can be sent for each recruiter. The recruiter sends the scout document to the job seeker by consuming their own stock of scout emails. If the stock of scout emails required to send the scout document is insufficient, the scout document management unit 115 accepts the purchase of scout emails (a charge to increase the stock of scout emails) from the recruiter.

[0158] The scout document management unit 115 may provide the user (recruiter) with a stock of limited scout emails (for example, one to ten) free of charge every predetermined period, which can be sent only to candidates presented by the presentation unit 114. The predetermined period for which the limited scout emails are provided is, for example, the same as the period for which the recommended job openings and / or candidates presented by the presentation unit 114 are updated. For example, by setting the predetermined period to one day, the user (recruiter) can be encouraged to log in to the recruitment support system 1 every day and send scout documents. Furthermore, the expiration date for the limited scout email is set to the time when the next limited scout email is provided; for example, if the limited scout email is provided every day, the expiration date for the limited scout email is the same as the day on which it is provided.

[0159] In addition, when a user, i.e., an employer, sends a scout email to a candidate presented by the presentation unit 114, the scout document management unit 115 may send the scout email to the candidate without consuming the employer's stock of scout emails.

[0160] <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 the server device 10 in each functional unit may be a common one, or may be prepared individually for each functional unit.

[0161] The artificial intelligence unit 120 is an AI (Artificial Intelligence) equipped with a language model such as a Transformer including GPT (Generative Pretrained Transformer, including GPT-1, GPT-2, and GPT-3), BERT (Bidirectional Encoder Representations from Transformers), and BART (Bidirectional and Auto-regressive Transformer), and a Recurrent Neural Network (RNN), and may include a generative AI.

[0162] The language model is an example of a learning model based on a machine learning algorithm. Specific examples of machine learning algorithms include nearest neighbor methods, naive Bayes methods, decision trees, support vector machines, and deep learning using neural networks. The artificial intelligence unit 120 can apply the above algorithms as appropriate.

[0163] 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). The training data consists of pairs of input data for learning and output data (correct answer data). Furthermore, the language model may not only be trained for a specific task, but also be a general-purpose model that can be used for a wide range of tasks.

[0164] The artificial intelligence unit 120 may be a general-purpose natural language processing learning model such as a large language model (LLM) that has learned a huge amount of data as artificial intelligence. Such a general-purpose learning model includes a language model that can handle various tasks without fine tuning using one-shot learning, few-shot learning, etc. Furthermore, the general-purpose learning model may also be configured to be able 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 may be a common general-purpose learning model.

[0165] The learning models included in the artificial intelligence unit 120 (learning models used in each functional unit, such as the expected value calculation model and the matching index calculation model) can undergo additional learning as transfer learning or fine tuning. For example, each time new job seeker registration information and new job postings are registered, the artificial intelligence unit 120 may perform additional learning and fine tuning using these as new training data. This improves the accuracy of the information output from the learning models.

[0166] The learning model included in the artificial intelligence unit 120 may be a learning model (distilled model) obtained by knowledge distillation using an original learning model. In knowledge distillation, a trained model such as a large-scale language model is used as a teacher model, and the parameters of the student model (distilled model) are adjusted to reduce the output loss (Soft Target Loss) of the student model (distilled model) relative to the output (Soft Target) of the teacher model, thereby training the student model, which becomes the distilled model. Alternatively, the student model may be trained to reduce the output loss (Hard Target Loss) of the student model relative to the correct label (Hard Target) of the teacher data (combination of input data and output data of the learning model). Compared to the original learning model (teacher model), the distilled model has a smaller number of parameters and a smaller processing load while maintaining performance similar to the learning model. Therefore, using a distilled model can reduce the cost of the recruitment support system 1.

[0167] For example, the expected value calculation model or the matching index calculation model may be a distilled model trained using a combination of input data and output data in a large-scale language model as training data. Furthermore, when the recruitment support system 1 is introduced, a large-scale language model may be used as the expected value calculation model or the matching index calculation model, and when training data from the large-scale language model is accumulated, a distilled model obtained by knowledge distillation using the training data may be used as the expected value calculation model or the matching index calculation model.

[0168] <Display section> The display unit 211 of the recruiting party terminal 20 and the display unit 311 of the job seeker terminal 30 each display a screen indicated by the screen data transmitted from the server device 10.

[0169] <Operation acquisition part> The operation acquisition unit 212 of the recruiting party terminal 20 accepts operations by a user (recruiter) who uses the recruiting party terminal 20. The operation acceptance unit 312 of the job seeker terminal 30 accepts operations by a user (job seeker) who uses the job seeker terminal 30.

[0170] 3. Information Processing Method This section describes an information processing method of the server device 10. This information processing method is executed by a computer, with each unit of the server device 10 acting as each step.

[0171] This information processing method includes a job offer extraction step, a candidate extraction step, and a presentation step. In the job offer extraction step, at least one recommended job offer is extracted from job offers registered in a database by a recruiter based on the content of the job offer and an expected value of contract calculated based on the first reference information. In the candidate extraction step, at least one candidate for the recommended job offer is extracted from job seekers registered in the database based on the content of the recommended job offer and a job seeker matching index calculated based on the second reference information. In the presentation step, the recommended job offer and the candidate are associated and presented to the recruiter.

[0172] 8 is an activity diagram showing the flow of information processing (recommended job offers and candidate presentation processing) executed by the recruitment support system 1. Below, the information processing will be explained along with each activity in this activity diagram.

[0173] The process of presenting recommended job offers and candidates begins with a predetermined operation by the recruiter. The predetermined operation is an operation that causes server device 10 to start the process of presenting recommended job offers, and may be an indirect operation such as the recruiter logging in to recruitment support system 1, or a direct operation such as an instruction by the recruiter to extract recommended job offers. The recruiter inputs such a predetermined operation on recruiter terminal 20 (activity A101). In response to the predetermined operation on recruiter terminal 20, server device 10 extracts recommended job offers (activity A102).

[0174] After extracting the recommended job offers, the server device 10 extracts candidates corresponding to the recommended job offers (activity A103). Next, the server device 10 selects the recommended job offers and candidates to present to the recruiter from the extracted recommended job offers and candidates (activity A104). Furthermore, the server device 10 outputs the selected recommended job offers and candidates to the recruiter terminal 20 (activity A105). As a result, the recommended job offers and candidates are displayed (presented) on the recruiter terminal 20 (activity A106). If necessary, the recruiter instructs the recruiter terminal 20 to view the registered information of the candidates, create scouting documents, etc.

[0175] 4. Effect The operation of this embodiment can be summarized as follows. That is, recommended job offers with a high expected value of success can be presented to the employer together with candidates who are highly suitable for the recommended job offers. Therefore, by the user, the employer, taking action based on the recommended job offers (for example, sending a scout document), the possibility of success in the job offer can be increased. Furthermore, by presenting candidates for the recommended job offers, the employer can identify job seekers who are likely to succeed in the job offer.

[0176] Therefore, according to this embodiment, the priority of job openings to be filled can be presented to the user (recruiter), while the ease of job selection can be visualized and provided to the recruiter. Furthermore, since the personal nature of the job openings to be filled can be reduced on the recruiter's side, the criteria for determining which job openings to fill can be standardized across individuals or departments. Furthermore, by presenting recommended job openings and candidates as a set, the recruiter's effort in searching for job seekers can be reduced and the recruiter can find job seekers who are difficult to find through their own search. For example, a recruiter with multiple job openings (especially a recruiter with a large number of job openings) must prioritize multiple job openings and fill them to find highly suitable job seekers within the constraints of internal resources such as personnel. According to this embodiment, job openings to be filled can be efficiently identified, and job seekers who are highly suitable for the job openings can also be simultaneously identified. This allows for efficient recruitment activities.

[0177] Although the embodiment of the present invention has been described above, the present invention is not limited to this and can be modified as appropriate within the scope of the technical idea of ​​the invention.

[0178] 5.Other In the above embodiment, the server device 10 performs various storage and control operations. However, multiple external devices may be used instead of the server device 10. That is, various information and programs may be distributed and stored in multiple external devices using blockchain technology or the like. In particular, the artificial intelligence unit 120 may be configured external to the server device 10. In this case, the artificial intelligence unit 120, which is an external configuration, is configured to receive input from each functional unit of the server device 10 and return instructed output to the server device 10.

[0179] The aspect of this embodiment is not limited to the hiring support system 1, and may be an information processing method or a program. The hiring support method includes steps executed by the hiring support system 1. The program causes a computer to execute the steps of the hiring support system 1.

[0180] It may be provided in the following manner.

[0181] (1) A recruitment support system comprising a processor configured to execute the following steps: in a job offer extraction step, at least one recommended job offer is extracted from job offers registered in a database by an employer based on the content of the job offer and an expected value of closing calculated based on first reference information, wherein the first reference information includes a correlation between the content of the job offer and the expected value of closing, which is the probability of closing for the job offer; in a candidate extraction step, at least one candidate for the recommended job offer is extracted from job seekers registered in the database based on the content of the recommended job offer and a job seeker matching index calculated based on second reference information, wherein the second reference information includes a correlation between the content of the job offer and the job seeker matching index, which indicates the degree of suitability of the job seeker for the job offer; and in a presentation step, the recommended job offer and the candidate are associated and presented to the employer.

[0182] (2) In the recruitment support system described in (1) above, the first reference information is an expected value calculation model that has been trained to be able to input the content of a job offer and output the expected value of the successful conclusion, and in the job offer extraction step, the content of the job offer is input into the expected value calculation model and the expected value of the successful conclusion is output from the expected value calculation model.

[0183] (3) In the recruitment support system described in (2) above, the expected value calculation model is learned using the contents of multiple job offers and whether or not these job offers have been concluded within a predetermined period.

[0184] (4) In the recruitment support system described in any one of (1) to (3) above, in the job extraction step, job offers whose expected value of closing is equal to or greater than a predetermined threshold are extracted as the recommended job offers.

[0185] (5) In the recruitment support system described in any one of (1) to (4) above, in the job offer extraction step, the recommended job offers are extracted from the job offers registered in the database by the employer based on the expected closing value of the job offer and the history of sending scouting documents based on the job offer to job seekers.

[0186] (6) In the recruitment support system described in (5) above, in the job extraction step, the recommended job offers are extracted based on a job offer score calculated for each job offer, the job offer score increases according to the magnitude of the expected value of success and its weighting, and decreases according to the magnitude of the number of scout document transmissions in a predetermined period and its weighting, and the weighting of the expected value of success is greater than the weighting of the number of transmissions.

[0187] (7) In the recruitment support system described in (6) above, in the job extraction step, a first job score and a second job score are calculated for each job as the job score, with different weightings for at least one of the expected value for closing and the number of submissions; at least one first recommended job extracted based on the first job score and at least one second recommended job extracted based on the second job score are extracted as the recommended job; and in the presentation step, the first recommended job and the second recommended job are presented to the employer.

[0188] (8) In the recruitment support system described in any one of (1) to (7) above, the second reference information is a matching index calculation model that has been trained to be able to input the content of a job offer and output the job seeker matching index for each job seeker, and in the candidate extraction step, the content of the recommended job offer is input into the matching index calculation model, and the matching index calculation model is caused to output the job seeker matching index.

[0189] (9) In the recruitment support system described in (8) above, the matching index calculation model is a learning model that learns the relationship between the contents of multiple job offers and job seekers who have a history of actions related to the job offers, and outputs the job seeker matching index as the degree of similarity between the input job offer and a job seeker who has a history of the actions related to a similar job offer.

[0190] (10) In the recruitment support system described in (9) above, the action is receiving a scouting document or passing an examination.

[0191] (11) In the recruitment support system described in any one of (1) to (9) above, in the candidate extraction step, the candidate is extracted from among job seekers registered in the database based on the job seeker matching index of the job seeker and the number of activities of the job seeker for any job opening.

[0192] (12) In the recruitment support system described in any one of (1) to (11) above, in the presentation step, a portion of the recommended job offers is selected and presented from the extracted plurality of recommended job offers for each specified period.

[0193] (13) In the recruitment support system described in any one of (1) to (12) above, in the presentation step, a portion of the candidates is selected and presented from the extracted plurality of candidates at predetermined intervals.

[0194] (14) A recruitment support system according to any one of (1) to (13) above, wherein in the presentation step, a label corresponding to the expected value of the contract is assigned to the recommended job offer and presented.

[0195] (15) A recruitment support method comprising the steps executed by the recruitment support system described in any one of (1) to (14) above.

[0196] (16) A program that causes a computer to execute each step of the recruitment support system described in any one of (1) to (14) above. Of course, this is not the case.

[0197] Finally, while various embodiments of the present disclosure have been described, they are presented as examples and are not intended to limit the scope of the invention. The novel embodiments may be embodied in various other forms, and various omissions, substitutions, and modifications may be made without departing from the spirit of the invention. Such embodiments and modifications are intended to be included within the scope and spirit of the invention, as well as within the scope of the inventions and their equivalents as defined in the claims. [Explanation of symbols]

[0198] 1: Recruitment support system 2: Communication line 10: Server device 11: Control section 12: Storage section 13: Communications Department 14: Communication bus 20: Recruiter terminal 21: Control unit 22: Storage section 23: Communications Department 24: Input section 25: Output section 26: Communication bus 30: Job seeker terminal 31: Control unit 32: Storage section 33: Communications Department 34: Input section 35: Output section 36: Communication bus 111: Basic display control section 112: Job extraction department 113: Candidate extraction department 114: Presentation part 115: Scout Document Management Department 120: Artificial Intelligence Department 211: Display section 212: Operation acquisition section 311: Display section 312: Operation reception section

Claims

1. A recruitment support system, a processor; The processor is configured to perform the following steps: In the job offer extraction step, at least one recommended job offer is extracted from the job offers registered in the database by the employer based on the content of the job offer and an expected value of contract calculated based on first reference information, wherein the first reference information includes a correlation between the content of the job offer and the expected value of contract, which is a probability of contract for the job offer; In the candidate extraction step, at least one candidate for the recommended job offer is extracted from among job seekers registered in the database based on a job seeker matching index calculated based on the content of the recommended job offer and second reference information, wherein the second reference information includes a correlation between the content of the job offer and the job seeker matching index indicating the degree of suitability of the job seeker for the job offer; In the presentation step, the recruitment support system presents the recommended job offers and the candidates in association with each other to the recruiter.

2. The recruitment support system according to claim 1, The first reference information is an expected value calculation model that has been trained to input the content of a job offer and output the expected value of the contract, In the job offer extraction step, the content of the job offer is input to the expected value calculation model, and the expected value calculation model is caused to output the contract expected value.

3. 3. The recruitment support system according to claim 2, The recruitment support system, wherein the expected value calculation model is learned using the contents of a plurality of job offers and whether or not these job offers have been concluded within a predetermined period.

4. The recruitment support system according to claim 1, In the job offer extraction step, job offers whose expected value of successful contract is equal to or greater than a predetermined threshold are extracted as the recommended job offers.

5. The recruitment support system according to claim 1, In the job offer extraction step, the recruitment support system extracts the recommended job offers from the job offers registered in the database by the employer based on the expected value of the job offer and the history of sending scouting documents based on the job offer to job seekers.

6. 6. The recruitment support system according to claim 5, In the job offer extraction step, the recommended job offers are extracted based on a job offer score calculated for each job offer; The job vacancy score increases in accordance with the magnitude of the expected value of contract and its weighting, and decreases in accordance with the magnitude of the number of scout documents sent in a predetermined period and its weighting, A recruitment support system, wherein the weighting of the expected value of closing is greater than the weighting of the number of transmissions.

7. 7. The recruitment support system according to claim 6, In the job offer extraction step, a first job offer score and a second job offer score are calculated for each job offer, with different weightings for at least one of the expected value for successful contract and the number of transmissions, and at least one first recommended job offer extracted based on the first job offer score and at least one second recommended job offer extracted based on the second job offer score are extracted as the recommended job offers; In the presenting step, the recruitment support system presents the first recommended job offer and the second recommended job offer to the recruiter.

8. The recruitment support system according to claim 1, the second reference information is a matching index calculation model that has been trained to input details of a job offer and to output the job seeker matching index for each job seeker, In the candidate extraction step, the content of the recommended job offers is input to the matching index calculation model, and the matching index calculation model is caused to output the job seeker matching index.

9. 9. The recruitment support system according to claim 8, The matching index calculation model is a learning model that learns the relationship between the contents of multiple job offers and job seekers who have a history of actions related to those job offers, and outputs the job seeker matching index as the degree of similarity between the input job offer and job seekers who have a history of actions related to similar job offers.

10. 10. The recruitment support system according to claim 9, The action is receiving a scout document or passing an examination, in the recruitment support system.

11. The recruitment support system according to claim 1, In the candidate extraction step, the recruitment support system extracts the candidates from among job seekers registered in a database based on the job seeker matching index of the job seeker and the number of activities of the job seeker for any job opening.

12. The recruitment support system according to claim 1, In the presenting step, a portion of the recommended job offers is selected and presented from the extracted plurality of recommended job offers for each predetermined period.

13. The recruitment support system according to claim 1, In the presenting step, a portion of the candidates is selected from the extracted plurality of candidates for each predetermined period and presented.

14. The recruitment support system according to claim 1, In the presentation step, the recruitment support system presents the recommended job offers by assigning a label corresponding to the expected value of contract.

15. A recruitment support method, comprising: A recruitment support method comprising the steps executed by the recruitment support system according to any one of claims 1 to 14.

16. A program, A program that causes a computer to execute each step of the recruitment support system according to any one of claims 1 to 14.

Images