Reemployment support device

Abstract

To provide a re-employment support device for providing re-employment support reflecting diverse information exemplified by desired treatment conditions, etc. brought from background of resignation circumstances of re-employment applicants.SOLUTION: A re-employment support device 1 includes: a re-employment applicant extraction unit 114 for extracting re-employment applicant information satisfying a designated condition from a re-employment applicant database 132 for storing desired treatment conditions of the re-employment applicant in association with the re-employment applicant information; a condition receiving unit 112 for receiving working conditions; and a condition generation unit 113 that, when the working conditions are received, causes a large-scale language model 131 to generate conditions (first conditions) that the user is a re-employment applicant associated with the desired treatment conditions that do not conflict with the working conditions. When the first conditions are generated, the re-employment applicant extraction unit 114 extracts a first re-employment applicant information satisfying the first conditions from the re-employment applicant database 132.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a reemployment support device. [Background technology]

[0002] Workers, such as nurses, may resign for various reasons, including childbirth, childcare, injury or illness, and family circumstances. Reemployment for former workers not only stabilizes their lives, but also offers the advantage of providing new employers with competent human resources with work experience. However, it is not easy for former workers to find new jobs that meet their desired conditions, and for companies to find former workers who meet their desired working conditions. Therefore, there is a need for methods to support former workers in finding new jobs.

[0003] (Prior Art) As an example of prior art related to supporting the reemployment of former employees, Patent Document 1 discloses a reemployment support system that, when a former employee's desired conditions are input, adds personnel data including the former employee's career history to the desired conditions to create job-seeking data, which is stored in a job-seeking database; when an employer's conditions are input, adds work environment data about the employer's employment location to the desired conditions to create job vacancy data, which is stored in the job vacancy database; and when introducing employment, compares the conditions with the job-seeking data, and presents the job vacancy data to the former employee when the conditions and the job-seeking data nearly match. The technology in Patent Document 1 can easily build a highly reliable job-seeking database, and can provide a reemployment support system that can efficiently rehire former employees in suitable positions based on their job searches and job offers. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2003-256551 Summary of the Invention [Problem to be solved by the invention]

[0005] For example, when returning to work, workers who retired due to childbirth or childcare may need to balance work with childcare and lifestyle changes that have occurred as a result of childbirth or childcare. Similarly, when returning to work, workers who retired due to illness or injury may need to adjust their work schedule to take into account changes in their health due to illness or injury. Furthermore, when returning to work, workers who retired due to family circumstances may need to adjust their work schedule to take into account their family circumstances.

[0006] Thus, unlike workers who are hired as new graduates, former employees seeking reemployment may have diverse backgrounds related to the circumstances of their resignation. The technology of Patent Document 1 can efficiently carry out reemployment by matching working conditions with job-seeking data. However, it is considered difficult to respond to diverse backgrounds by simply matching working conditions with job-seeking data. Therefore, the technology of Patent Document 1 has room for further improvement in terms of providing reemployment support that takes diverse backgrounds into account.

[0007] One possible employment support option is to create a database of job seekers in collaboration with social networking services (SNS) and then provide a subscription-based service to introduce job seekers registered in that database to companies with job openings. Unlike other employment support services that charge a success fee for each introduction, this type of employment support allows for sustainable support. Therefore, this type of employment support is considered suitable for outplacement, as it allows for a gradual process of re-employment, for example, from part-time employment to full-time permanent employment. However, even when providing such outplacement support, consideration must be given to diverse backgrounds when matching job seekers with new employers.

[0008] The present invention has been made in consideration of the above circumstances. The purpose of the present invention is to provide outplacement support that reflects a variety of information, such as desired conditions for treatment, etc., that are derived from the background of the circumstances surrounding the resignation of the job seeker. [Means for solving the problem]

[0009] As a result of extensive research into solving the above problems, the inventors have found that the above object can be achieved by extracting information from a database via a large-scale language model, and have completed the present invention. Specifically, the present invention provides the following.

[0010] The present invention provides a re-employment support device comprising: a re-employment applicant extraction unit that extracts re-employment applicant information that satisfies specified conditions from a predetermined database in which the desired treatment conditions of re-employment applicants are stored in association with information related to the re-employment applicants (re-employment applicant information); a condition receiving unit that receives working conditions; and a condition generation unit that, when the working conditions are received, causes a large-scale language model to generate a condition (first condition) that the re-employment applicant is associated with desired treatment conditions that do not contradict the working conditions, wherein the re-employment applicant extraction unit, when the first condition is generated, extracts first re-employment applicant information that satisfies the first condition from the predetermined database.

[0011] The job seeker extraction unit of the present invention can extract information on job seekers who meet working conditions from a predetermined database in which the desired conditions for employment of job seekers are stored in association with job seeker information. Information is typically extracted from a database using processing based on data extraction commands written in a query language (database language) for manipulating data in a database, such as SQL (Structured Query Language), or commands written in a database language, such as software processing via an API (Application Programming Interface).

[0012] Unlike new graduates, former employees seeking reemployment may have diverse backgrounds regarding the circumstances that led to their resignation. Therefore, the desired conditions for employment of those seeking reemployment may include a variety of desired conditions written in natural language. Database languages, on the other hand, are computer languages ​​designed for the specific task of using databases, and have different grammars from natural languages.

[0013] Due to the nature of desired employment conditions and database languages, it is not easy to generate data extraction commands written in database languages ​​based on various working conditions and various desired employment conditions written in natural language. For example, a procedure for generating data extraction commands based on data classification using a neural network is conceivable, but creating a program that can generate data extraction commands corresponding to a huge number of classifications corresponding to various working conditions and various desired employment conditions is considered difficult to realize because it requires a great deal of effort.

[0014] Here, the large-scale language model is a language model that has been pre-trained using a large amount of text written in various languages, such as the languages ​​of various countries, database languages, etc. Through such pre-training, the large-scale language model has the emergent ability to generate text with similar content in other languages ​​used in the pre-training, based on natural language relating to various working conditions and various desired treatment conditions, exemplified by sentences expressing desires regarding balancing childcare and work.

[0015] The present invention can generate first conditions written in a database language based on various working conditions and various desired treatment conditions written in natural language by using a large-scale language model to generate conditions (first conditions) that the job seeker is associated with desired treatment conditions that do not contradict the working conditions. As a result, the present invention can simultaneously utilize a conventional database operated using a database language and extract first job seeker information from the database based on the various working conditions and various desired treatment conditions written in natural language. [Effects of the Invention]

[0016] The present invention can provide outplacement support that reflects a variety of information, such as desired conditions for treatment, etc., that are derived from the background of the circumstances surrounding the resignation of the job seeker. [Brief explanation of the drawings]

[0017] [Figure 1] FIG. 1 is a block diagram showing the hardware and software configurations of a system S according to this embodiment. [Figure 2] FIG. 2 is an example of information stored in the job seeker database 132. [Figure 3] FIG. 3 is an example of a job advertisement stored in the job advertisement database 133. [Figure 4] FIG. 4 is a main flowchart showing an example of a preferable flow of the assistance process executed by the assistance device 1 of this embodiment. [Figure 5] FIG. 5 is a continuation of FIG. [Figure 6] FIG. 6 is a continuation of FIG. DETAILED DESCRIPTION OF THE INVENTION

[0018] Hereinafter, an example of an embodiment of the present invention will be described in detail with reference to the drawings.

[0019] <System S> 1 is a block diagram showing the hardware and software configurations of a system S according to this embodiment. An example of a preferred embodiment of the hardware and software configurations of the system S according to this embodiment will be described below with reference to FIG.

[0020] The system S is configured to include a reemployment support device 1 that provides reemployment support that reflects a variety of information, and a terminal T that is configured to be able to communicate with the support device 1 via a network N.

[0021] System S can be used for outplacement support, which works in conjunction with social networking services (SNS) to create a database of job seekers and introduces job seekers registered in the database to companies with job openings, on a subscription basis. In outplacement support provided through subscriptions, System S can provide continuous support, unlike other types of outplacement support that incur a success fee for each introduction. Specifically, System S, for example, matches job seekers with new employers through the support process described below, and then provides continuous outplacement support by managing the process of outplacement support so that job seekers can progress through stages of re-employment, from part-time employment to promotion to full-time permanent employment.

[0022] [Support device 1] The support device 1 includes a control unit 11, a storage unit 13, and a communication unit 14. There are no particular limitations on the type of the support device 1. Examples of the type include a server, a cloud server, and the like.

[0023] The support device 1, using the hardware and software configurations of the system S of this embodiment described below, causes the large-scale language model 131 to generate conditions that indicate that the job seeker is associated with desired treatment conditions that do not contradict the received working conditions, and extracts information on job seekers who satisfy these conditions from the job seeker database 132. A preferred flow of the support process performed by the support device 1 will be described after an example of a preferred aspect of the hardware and software configurations of the system S of this embodiment.

[0024] The assistance device 1 is configured to perform a process of causing a large-scale language model 131 that has been pre-trained using 500 GB or more of text to perform machine learning, and a process of causing the large-scale language model 131 that has undergone machine learning to generate advice. These processes are executed, for example, via an API related to the use of the large-scale language model 131. An example of a more detailed configuration of the assistance device 1 is given below.

[0025] [Control unit 11] The control unit 11 includes a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), and the like.

[0026] The control unit 11 cooperates with the storage unit 13 and / or the communication unit 14 as necessary. The control unit 11 then realizes the following software components of the program of this embodiment executed by the support device 1: a machine learning unit 111, a condition receiving unit 112, a condition generating unit 113, a job seeker extraction unit 114, a solicitation message generation unit 115, a job advertisement generation unit 116, an advertisement providing unit 117, an advertisement evaluation receiving unit 118, a job advertisement extraction unit 119, a workplace evaluation receiving unit 120, and a workplace evaluation analysis and generation unit 121. The functions provided by each of the software components of the program of this embodiment will be described in the description of a preferred flow of the support process, which will be described later.

[0027] [Storage section 13] The memory unit 13 is a device in which data and / or files are stored, and includes a storage unit that stores data non-temporarily using a hard disk, a semiconductor memory, a recording medium, a memory card, or the like.

[0028] The storage unit 13 may have a mechanism that enables connection to a storage device or storage system such as a NAS (Network Attached Storage), a SAN (Storage Area Network), cloud storage, a file server, and / or a distributed file system via a network N.

[0029] The memory unit 13 stores a program executed by a microcomputer, a large-scale language model 131, a job seeker database 132, a job advertisement database 133, information related to workplaces such as job seekers who have subscribed to recruitment services related to job seeker support, information related to the job seeker SNS, various data received by the support device 1, various data generated by the support device 1, etc.

[0030] (Large-scale language model 131) The large-scale language model 131 (LLM) in this embodiment is a type of probabilistic model used in natural language processing, and is a model for probabilistically predicting how likely a given word or sentence is to occur in natural language. The large-scale language model 131 in this embodiment works in cooperation with the condition generation unit 113 and the like to generate a condition (first condition) that the job seeker is associated with desired treatment conditions that do not contradict the working conditions.

[0031] The large-scale language model 131 may be stored in an external cloud server or the like and executed via an API (Application Programming Interface) related to the use of the large-scale language model 131, or may be stored in the storage unit 13. When the large-scale language model 131 is executed via an API, the assistance device 1 has a simpler configuration and is more cost-effective than a configuration in which the large-scale language model 131 is stored in the storage unit 13. When the large-scale language model 131 is stored in the storage unit 13, the assistance device 1 can execute processing related to the large-scale language model 131 without communicating point-by-point with an external device such as a cloud server. This allows the assistance device 1 to reduce processing delays related to communication, risks to information security, and the like.

[0032] The large-scale language model 131 is preferably trained on a large amount of text, at least 500 GB (gigabytes). Examples of such models include ChatGPT, GPT-3.5, and GPT-4 from OpenAI (registered trademark), and LLaMa from Meta AI (registered trademark). As the amount of training data increases, the performance of a language model increases according to the broken neural scaling law (BNSL), and it is believed that the model acquires emergent capabilities when transitioning between segments in the BNSL. Therefore, the large-scale language model 131 of this embodiment acquires emergent capabilities related to generating the first condition, which are acquired according to the amount of text used in pre-training.

[0033] The large-scale language model 131 preferably has an upper limit of 30,000 or more contexts used to generate text when counting tokens, which are semantic blocks of text (words, etc.). This enables the large-scale language model 131 to generate first conditions and a first job advertisement that reflects working conditions, which are described using a large number of tokens to include many conditions, as context.

[0034] This also enables the large-scale language model 131 to generate solicitation letters addressed to job seekers relating to job seeker information written using a large number of tokens, generate analysis results for advertising evaluation text written using a large number of tokens, and generate analysis results for multiple workplace evaluations that will be written using a large number of tokens because they are multiple.

[0035] The large-scale language model 131 is preferably trained on text containing 10 trillion or more tokens, which is believed to give the large-scale language model 131 the emergent ability to generate first conditions, solicitation texts, analysis results of advertising evaluation texts, analysis results of workplace evaluations, and the like that reflect nuances according to the job type of the workplace where the job is being recruited.

[0036] The large-scale language model 131 is preferably a model having a large number of parameters, at least 10 billion or more, so that the large-scale language model 131 can perform processing that fully reflects the emergent capabilities obtained from a large amount of text.

[0037] (Reemployment Seeker Database 132) The job seeker database 132 (predetermined database) stores the desired conditions of job seekers in association with information related to the job seekers (job seeker information). The means for realizing this association is not particularly limited. For example, this association may be realized in a manner in which the desired conditions of employment and the job seeker information are associated on the job seeker database 132, which is a relational database (RDB), or in a manner in which the desired conditions of employment are included in the job seeker information. The job seeker database 132 is configured so that the job seeker extraction unit 114 can extract information about job seekers who satisfy predetermined conditions.

[0038] In the following explanation, it is assumed that the association is realized in a manner in which various information associated with the job seeker information, such as desired conditions of treatment, is included in the job seeker information. However, it will be understood by those skilled in the art that the association can be similarly realized by association on an RDB.

[0039] The information on job seekers includes various personal information on job seekers, such as their name, age, sex, work history, and qualifications.

[0040] By including work history (for example, years of work as a registered nurse, years of work as a medical clerk, work history at a nursing care facility, etc.) as personal information in the re-employment applicant information, the support device 1 can handle working conditions including conditions related to work history. In other words, the support device 1 can extract re-employment applicants whose working conditions are consistent with their work history as well as their desired treatment conditions. Therefore, the support device 1 can provide re-employment support that reflects a wider variety of information.

[0041] By including the qualifications held (e.g., registered nurse, licensed practical nurse, care worker, registered dietitian, etc.) and information incidental to the qualifications held (e.g., a nurse's history of attending specific practice training) as personal information in the re-employment applicant information, the support device 1 can handle working conditions including conditions related to the qualifications held. In other words, the support device 1 can extract re-employment applicants whose working conditions are consistent with their desired treatment and qualifications. Therefore, the support device 1 can provide re-employment support that reflects a wider variety of information.

[0042] Desired conditions for treatment include a wide variety of conditions, exemplified by, for example, job content, work location, whether or not there is a possibility of transfer and details thereof, working hours (including non-standard desired hours such as desired hours that vary depending on the day of the week or desired hours that vary only on specific days), working days (days of the week, etc.), contract period, employment type (contract employee, full-time employee, etc.), salary and salary increase system, whether or not there is social insurance and details thereof, whether or not there is childcare facility and details thereof, whether or not there is childcare support and details thereof, whether or not there is retraining support and details thereof, whether or not there is support for obtaining qualifications and details thereof, and compliance efforts (measures against industrial accidents, measures against sexual harassment, measures against power harassment, health management support, etc.).

[0043] FIG. 2 is an example of information stored in the re-employment applicant database 132. In the example shown in FIG. 2, the re-employment applicant information for a re-employment applicant with the personal information name "Yamada Hanako" includes the following personal information: age and gender "32-year-old female," qualification "registered nurse (has completed specific practice training (adjusting insulin dosage))," and work history "worked in the nephrology ward of △△ Hospital from △△ year to △△ year." The re-employment applicant information further includes the following desired conditions for treatment: desired work content "nursing position," desired work location, etc. "preferably around △△ City, no transfers allowed," desired working hours, etc. "Monday-Thursday 9:00-16:00, Friday-Saturday 9:00-13:00, no work on the 10th to 25th of each month," desired employment type "no questions," childcare support preference "it would be helpful if you could take care of my child after 15:00," and compliance commitment "prefer a company that provides thorough sexual harassment prevention training for female part-timers."

[0044] As a result, the support device 1 executes the support processing described below based on the above-mentioned information stored in the re-employment applicant database 132, and can provide re-employment support that reflects a variety of information, such as personal information including the qualification of having received specific procedure training that is useful in a specific medical department, as well as desired irregular working hours, such as not only days of the week but also requests to take specific days off, and desired conditions for treatment that call for compliance efforts that differ from typical harassment prevention training.

[0045] (Job Advertisement Database 133) Recruitment advertisement database 133 (specific database) stores various recruitment advertisements, such as those generated by recruitment advertisement generation unit 116. Recruitment advertisement database 133 is configured so that recruitment advertisement extraction unit 119 can extract recruitment advertisements that satisfy predetermined conditions.

[0046] There is no particular limitation on the type of recruitment advertisements stored in the recruitment advertisement database 133. The recruitment advertisements include advertisements in various formats, exemplified by, for example, text formats suitable for transmission via email, transmission via short message, posting to SNS (social networking service), etc., modified text formats (e.g., HTML (HyperText Markup Language), etc.) that can achieve greater expressiveness than text when transmitted via email, formats that include text and images (e.g., PDF (Portable Document Format), web pages, etc.), audio formats suitable for transmission via voice calls, voice messages, etc., image formats suitable for image advertisements, and expressive video formats.

[0047] The job advertisements stored in the job advertisement database 133 preferably contain text. When the job advertisements contain text, it becomes easier to extract the text based on the condition of the text generated by the large-scale language model 131. When the job advertisements contain multimedia information such as images, audio, and video, it is preferable that the job advertisements are associated with text indicating the content of the multimedia information generated by various recognition processes for the multimedia information, in order to make it easier to extract the text based on the condition of the text.

[0048] FIG. 3 illustrates an example of a job advertisement stored in the job advertisement database 133. In the example shown in FIG. 3, the job advertisement identified by ID “C0001” includes a job description corresponding to the employment condition “Registered nurse available to work in the nephrology ward of △△ City Hospital, available to work three days a week, preferred experience in nephrology.” The job description reads, “Nephrology experience preferred! Registered nurse wanted, available to work three days a week. Would you like to work in the nephrology ward of △△ City Hospital?” along with an image of a heart-shaped IV drip container. The job advertisement is also associated with text indicating the content of multimedia information including the text “Image of heart-shaped IV drip container” that represents the image. In the example shown in FIG. 3, the job advertisement identified by ID “C0002” includes a job description corresponding to the employment condition “Caregiver working at △△ Special Nursing Home, available to work night shifts, preferred experience as a certified care worker.” The job description reads, “Caregiver available to work night shifts! Work at △△ Special Nursing Home, certified care workers especially welcome.”

[0049] As a result, the support device 1 can execute the support processing described below based on the above-mentioned information stored in the job advertisement database 133, and provide job reemployment support that reflects a variety of information, such as job advertisements of various types that reflect a variety of working conditions.

[0050] [Communications Section 14] The communication unit 14 is not particularly limited as long as it connects the support device 1 to the network N and enables communication with the terminal T, etc. Examples of the communication unit 14 include a wireless device compatible with a mobile phone network, a device connectable to a wireless LAN, and a network card compatible with the Ethernet standard.

[0051] [Network N] The type of the network N is not particularly limited as long as it allows communication between the support device 1 and the terminal T, etc. The type of the network N is, for example, the Internet, a mobile phone network, a wireless LAN, etc.

[0052] [Terminal T] Terminal T includes a predetermined terminal used by a job seeker, a specific terminal used by a workplace such as a recruiter, etc. The type of terminal T is not particularly limited. Terminal T is, for example, a personal computer, a laptop computer, a smartphone, a tablet terminal, etc. Terminal T is configured to perform processes such as providing the support device 1 with desired working conditions input by a job seeker, etc., providing the support device 1 with working conditions input by a recruiter, etc., displaying a solicitation message received from the support device 1, and displaying a job advertisement received from the support device 1.

[0053] [Main flowchart of support processing] Fig. 4 is a main flowchart showing an example of a preferable flow of the support processing executed by the support device 1 of this embodiment. Fig. 5 is a diagram continuing from Fig. 4. Fig. 6 is a diagram continuing from Fig. 5. The following is an example of a preferable flow of the support processing executed by the support device 1 of this embodiment using Figs. 4 to 6.

[0054] The support process preferably includes a series of processes (steps S1 to S2) related to machine learning of the large-scale language model 131. This enables the support device 1 to generate text in the large-scale language model 131 with higher accuracy, such as a condition (first condition) that the person is an employment seeker associated with desired treatment conditions that do not contradict the working conditions.

[0055] [Step S1: Determine whether to perform machine learning] The control unit 11 executes the machine learning unit 111 in cooperation with the storage unit 13 and the communication unit 14. Then, the control unit 11 executes a process of determining whether or not to perform machine learning of the large-scale language model 131 using the machine learning unit 111 (step S1, machine learning determination step). If it is determined that machine learning should be performed, the control unit 11 proceeds to step S2. If it is not determined that machine learning should be performed, the control unit 11 proceeds to step S3.

[0056] The procedure of the machine learning determination step is not particularly limited as long as it makes the above-mentioned determination. The procedure includes, for example, a procedure for determining that machine learning is to be performed when any of various types of learning data described below is stored in the storage unit 13 and the control unit 11 is notified of the storage.

[0057] [Step S2: Run machine learning] The control unit 11 executes a process of executing machine learning of the large-scale language model 131 based on the training data by the machine learning unit 111 (step S2, machine learning step). The control unit 11 moves the process to step S3.

[0058] (Training data) The training data in the machine learning step is not particularly limited. The training data may include, for example, one or more of the following. Below, examples of training data and machine learning performed in the machine learning step based on the training data will be described.

[0059] The learning data in the machine learning step explains working conditions. variableand the training data preferably includes training data including, as a target variable, a condition (first condition) that the job seeker is a job seeker associated with desired treatment conditions that do not contradict the working conditions. As a result, the machine learning step can use machine learning to generate the first conditions using the working conditions as input in the large-scale language model 131. The first conditions in the training data are preferably written in a database language used to extract job seeker information from the job seeker database 132. As a result, the machine learning step can use machine learning to generate the first conditions written in the database language used for extraction from the job seeker database 132 in the large-scale language model 131.

[0060] The learning data in the machine learning step explains the desired conditions for treatment. variable and the training data preferably includes, as a target variable, a solicitation message addressed to the job seeker associated with the first job seeker information. In this way, the machine learning step uses the desired working conditions associated with the first job seeker information as an input and allows the large-scale language model 131 to perform machine learning to generate a solicitation message addressed to the job seeker related to the first job seeker information.

[0061] The learning data in the machine learning step explains working conditions. variable and the training data preferably includes a job advertisement (first job advertisement) that reflects the working conditions as a target variable. In this way, the machine learning step can use large-scale language model 131 to perform machine learning to generate the first job advertisement using the working conditions as an input.

[0062] The learning data in the machine learning step is a description of the job advertisement and the evaluation of the job advertisement (ad evaluation). variableand the training data preferably includes a condition (second condition) that the job advertisement is similar in content to the job advertisement associated with the favorable advertising rating as a target variable. In this way, the machine learning step can cause the large-scale language model 131 to learn by machine learning the generation of the second condition using the advertisement rating as input. The first condition in the training data is preferably written in a database language used to extract job advertisements from the job advertisement database 133. In this way, the machine learning step can cause the large-scale language model 131 to learn by machine learning the generation of the second condition written in the database language used for extraction from the job advertisement database 133.

[0063] The learning data in the machine learning step is a set of data that describes the ad evaluation text. variable and preferably includes training data including the analysis result of the advertisement evaluation text as a target variable. Thereby, the machine learning step can use the advertisement evaluation text as an input and have the large-scale language model 131 perform machine learning to generate the analysis result of the advertisement evaluation text.

[0064] The learning data in the machine learning step explains multiple workplace evaluations. variable and the training data includes analysis results of the plurality of workplace evaluations as a dependent variable. In this way, the machine learning step uses the plurality of workplace evaluations as input and allows the large-scale language model 131 to perform machine learning to generate analysis results related to the plurality of workplace evaluations.

[0065] The support device 1 preferably includes a series of processes (steps S3 to S5) for extracting job seekers from the job seeker database 132 by, for example, generating first conditions from working conditions using the large-scale language model 131. This enables the support device 1 to simultaneously utilize a conventional database operated using a database language as the job seeker database 132 and extract first job seeker information from the job seeker database 132 based on a variety of working conditions and a variety of desired conditions for benefits written in natural language.

[0066] [Step S3: Determine whether working conditions have been received] The control unit 11 executes the condition receiving unit 112 in cooperation with the storage unit 13 and the communication unit 14. Then, the control unit 11 executes a process of determining whether the working conditions have been received from the terminal T or the like by the condition receiving unit 112 (step S3, condition receiving step). If it is determined that the working conditions have been received, the control unit 11 stores the received working conditions in the storage unit 13 and proceeds to step S4. If it is not determined that the working conditions have been received, the control unit 11 proceeds to step S11.

[0067] (Working conditions) The working conditions received in the conditions receiving step are not particularly limited as long as they can be provided as input text to the large-scale language model 131. The working conditions include a wide variety of conditions, exemplified by, for example, job content, work location, whether or not there is a transfer and details thereof, working hours (including non-standard desired hours such as desired hours that vary depending on the day of the week or desired hours that vary only on specific days), working days (days of the week, etc.), contract period, employment type (contract employee, full-time employee, etc.), salary and salary increase system, whether or not there is social insurance and details thereof, whether or not there is a childcare facility and details thereof, whether or not there is childcare support and details thereof, whether or not there is retraining support and details thereof, whether or not there is support for obtaining qualifications and details thereof, compliance efforts (measures against industrial accidents, measures against sexual harassment, measures against power harassment, health management support, etc.), workplace atmosphere, supervisor's hobbies, workplace customs, etc.

[0068] [Step S4: Generate the first condition] Control unit 11 executes condition generation unit 113 in cooperation with storage unit 13 and communication unit 14. Control unit 11 then executes a process in which condition generation unit 113 provides input including the working conditions received in step S3 to large-scale language model 131, and causes large-scale language model 131 to generate a condition (first condition) that the job seeker is associated with desired treatment conditions that do not contradict the working conditions (step S4, first condition generation step). Control unit 11 then proceeds to step S5.

[0069] As described above, the large-scale language model 131 of this embodiment is pre-trained using a large amount of text written in various languages, such as the languages ​​of various countries, database languages, etc. Through such pre-training, the large-scale language model 131 acquires the emergent ability to generate text with similar content in other languages ​​used in the pre-training, based on natural language relating to various working conditions and various desired treatment conditions, exemplified by sentences expressing desires regarding balancing childcare and work.

[0070] Therefore, the first condition generation step allows the large-scale language model 131 to generate a condition (first condition) that the person is a re-employment applicant that is associated with desired treatment conditions that do not contradict the working conditions, thereby generating the first condition written in database language based on the various working conditions and various desired treatment conditions written in natural language.

[0071] [Step S5: Extract job seekers who meet the first condition] The control unit 11 executes the job seeker extraction unit 114 in cooperation with the storage unit 13 and the communication unit 14. Then, the control unit 11 executes a process in which the job seeker extraction unit 114 extracts job seeker information (first job seeker information) that satisfies the first condition generated in step S4 from the job seeker database 132 (step S5, job seeker extraction step). The control unit 11 then moves the process to step S6.

[0072] By performing the above-mentioned pre-training in the large-scale language model 131 and generating the first conditions using the large-scale language model 131 that has undergone the pre-training, the support device 1 can simultaneously utilize a conventional database operated using a database language and extract first job seeker information from the database based on various working conditions and various desired treatment conditions written in natural language.

[0073] The support process preferably includes a series of processes related to sending solicitation messages (steps S6 to S7). This allows the support device 1 to automatically send solicitation messages addressed to job seekers, thereby reducing the burden on recruiters. At this time, the large-scale language model 131 is used to explain desired conditions for treatment. variable It is preferable that machine learning is performed using learning data including, as a target variable, a solicitation letter addressed to job seekers associated with the desired conditions of treatment.

[0074] [Step S6: Generate solicitation text] The control unit 11 executes the solicitation message generation unit 115 in cooperation with the storage unit 13 and the communication unit 14. Then, the control unit 11 executes a process in which the solicitation message generation unit 115 causes the large-scale language model 131 to generate a solicitation message addressed to the job seeker related to the first job seeker information, based on the desired conditions of treatment, etc., related to the first job seeker information extracted in step S5 (step S6, solicitation message generation step). The control unit 11 then proceeds to step S7.

[0075] By performing pre-training on the large-scale language model 131 described above and generating a solicitation message using the large-scale language model 131 that has undergone pre-training, the support device 1 generates a solicitation message that attracts the interest of the job seeker related to the first job seeker information, based on the desired working conditions, etc. The solicitation message generation step preferably includes a procedure for causing the large-scale language model 131 to generate a solicitation message that attracts the interest of the job seeker related to the first job seeker information, based on the desired working conditions described above and the working conditions related to step S3, and that does not contradict the working conditions. This allows the support device 1 to generate a solicitation message that reflects the convenience of both parties regarding job hunting.

[0076] [Step S7: Send out solicitation letters] The control unit 11, in cooperation with the storage unit 13 and the communication unit 14, executes a process of sending the solicitation message generated in step S6 to the terminal T used by the job seeker associated with the first job seeker information extracted in step S5 (step S7, solicitation message sending step). The control unit 11 then moves the process to step S8.

[0077] The support process preferably includes a series of processes (steps S8 to S10) related to the generation and provision of job advertisements. This allows the support device 1 to support reducing the burden on recruiters by automatically providing job advertisements. At this time, the large-scale language model 131 is used to describe working conditions. variable It is preferable that machine learning is performed using training data that includes, as a target variable, a job advertisement associated with the working conditions.

[0078] [Step S8: Determine whether to generate a job advertisement] The control unit 11 cooperates with the memory unit 13 and the communication unit 14 to execute the job advertisement generation unit 116. Then, the control unit 11 executes a process to determine whether or not to generate a job advertisement based on the working conditions received in step S3 using the job advertisement generation unit 116 (step S8, job advertisement generation determination step). If it is determined that a job advertisement should be generated, the control unit 11 proceeds to step S9. If it is not determined that a job advertisement should be generated, the control unit 11 proceeds to step S11.

[0079] The procedure for determining whether or not to generate a job advertisement in the job advertisement generation determination step is not particularly limited. For example, the procedure may be a procedure for determining whether or not to generate a job advertisement when data instructing the generation of an advertisement associated with the working conditions is received.

[0080] [Step S9: Generate job advertisement] Control unit 11 executes job advertisement generation unit 116 in cooperation with memory unit 13 and communication unit 14. Control unit 11 then executes a process in which job advertisement generation unit 116 generates a job advertisement (first job advertisement) that reflects the working conditions received in step S3 in large-scale language model 131, based on the working conditions (step S9, job advertisement generation step). Control unit 11 then proceeds to step S10.

[0081] By performing pre-training on the large-scale language model 131 described above and generating a job advertisement using the pre-trained large-scale language model 131, the support device 1 generates a job advertisement that reflects working conditions based on working conditions, etc. This allows the support device 1 to reduce the workload of recruiters.

[0082] The job advertisements generated in the job advertisement generation step include, for example, a landing page on a job advertisement site, a job advertisement posted on a job seeker SNS, and a job direct mail sent to job seekers via email or the like.

[0083] [Step S10: Submit a job advertisement] The control unit 11 executes the advertisement provision unit 117 in cooperation with the storage unit 13 and the communication unit 14. Then, the control unit 11 executes a process of providing the job advertisement generated in step S9 to the terminal T, etc., by the advertisement provision unit 117 (step S10, first advertisement provision step). The control unit 11 shifts the process to step S11.

[0084] The destinations to which the job advertisements are provided in the first advertisement providing step are not particularly limited. The destinations to which the job advertisements are provided preferably include various job advertisement sites as well as SNSs (job seeker SNSs) managed in the SNS management step described below. This allows the support device 1 to reduce the labor required by recruiters to provide job advertisements via SNSs. By providing the advertisements to the job seeker SNS among SNSs, the support device 1 can provide job advertisements to the job seeker SNSs that are expected to be seeking information related to reemployment. This allows the support device 1 to provide job advertisements in which the number of job seekers relative to the number and frequency of job advertisements is greater than that of a job advertisement site.

[0085] If a landing page for a job advertisement site is generated in the job advertisement generation step, the recipient of the job advertisement in the first advertisement providing step is preferably the job advertisement site.If a job advertisement post to be posted to a job seeker SNS is generated in the job advertisement generation step, the recipient of the job advertisement in the first advertisement providing step is preferably the job seeker SNS.If a job direct mail is generated in the job advertisement generation step, the recipient of the job advertisement in the first advertisement providing step is preferably job seekers.

[0086] [Step S11: Determine whether advertisement evaluation has been received] The control unit 11 cooperates with the storage unit 13 and the communication unit 14 to execute the advertisement evaluation receiving unit 118. Then, the control unit 11 executes a process of determining whether an advertisement evaluation has been received from the terminal T or the like by the advertisement evaluation receiving unit 118 (step S11, advertisement evaluation receiving step). If it is determined that the advertisement evaluation has been received, the control unit 11 stores the received advertisement evaluation in the storage unit 13 and proceeds to step S12. If it is not determined that the advertisement evaluation has been received, the control unit 11 proceeds to step S15.

[0087] [Step S12: Generate second condition] The control unit 11 executes the condition generation unit 113 in cooperation with the storage unit 13 and the communication unit 14. Then, the control unit 11 executes a process in which the condition generation unit 113 causes the large-scale language model 131 to generate a condition (second condition) that the job advertisement has similar content to a job advertisement related to a favorable advertising evaluation among the advertising evaluations received in step S11 (step S12, second condition generation step). The control unit 11 moves the process to step S13.

[0088] As described above, the large-scale language model 131 of this embodiment is pre-trained using a large amount of text written in various languages ​​such as the languages ​​of various countries, database languages, etc. Through such pre-training, the large-scale language model 131 acquires the emergent ability to understand the working conditions indicated in job advertisements that reflect a variety of evaluations and working conditions that can be expressed in text, and to generate text in another language used in the pre-training that has the same content as the working conditions indicated in job advertisements related to evaluations that have been determined to be favorable in that understanding.

[0089] Therefore, the second condition generation step can generate a second condition written in the database language by having the large-scale language model 131 generate a condition (second condition) that the job advertisement has content similar to that of a job advertisement related to a favorable advertising evaluation.

[0090] The advertisement evaluations received in the advertisement evaluation receiving step and referenced in the second condition generating step are not particularly limited as long as they can be expressed in text. Examples of such advertisement evaluations include information indicating a favorable evaluation of a job advertisement, such as a "Like" on a social networking site, a comment added to the job advertisement, or a review of the job advertisement. By having the large-scale language model 131 generate the second conditions in the second condition generating step, the support device 1 can determine whether various evaluations that can be expressed in text are favorable evaluations. Then, based on this determination, the large-scale language model 131 can generate the second condition, which is a condition that the job advertisement has content similar to that of a job advertisement related to a favorable advertisement evaluation.

[0091] [Step S13: Extract job advertisements] Control unit 11 executes job advertisement extraction unit 119 in cooperation with memory unit 13 and communication unit 14. Control unit 11 then executes a process in which job advertisement extraction unit 119 extracts job advertisements (second job advertisements) that satisfy the second condition generated in step S12 from job advertisement database 133 (step S13, job advertisement extraction step). Control unit 11 then moves the process to step S14.

[0092] By performing pre-training in the large-scale language model 131 described above and generating the second conditions using the large-scale language model 131 that has undergone pre-training, the support device 1 can simultaneously utilize a conventional database operated using a database language and extract job advertisements from the database based on various evaluations that can be expressed in text and various working conditions written in natural language.

[0093] [Step S14: Submit job advertisement] The control unit 11 executes the advertisement provision unit 117 in cooperation with the storage unit 13 and the communication unit 14. Then, the control unit 11 executes a process of providing the job advertisement extracted in step S13 to the terminal T, etc., by the advertisement provision unit 117 (step S14, second advertisement provision step). The control unit 11 shifts the process to step S15.

[0094] [Advertising evaluation analysis step] The support process preferably includes, as a process related to an advertisement evaluation analysis generation unit (not shown), an advertisement evaluation written in natural language (advertisement evaluation text) that is an evaluation of the job advertisement provided in the first advertisement providing step or the second advertisement providing step, and an advertisement evaluation analysis generation step in which an analysis result of the advertisement evaluation text is generated in the large-scale language model 131. This allows the support device 1 to provide the results of the analysis to the recruiter as material for improving the job advertisement.

[0095] [Work experience management steps] It is preferable that the support device 1 executes a work experience management step for managing work experience programs for re-employment applicants who wish to undergo work experience related to re-employment among the re-employment applicants extracted in step S5, or for re-employment applicants who wish to undergo work experience related to re-employment after seeing the job advertisement provided in step S10 or step S14.

[0096] The work experience management step preferably includes a step of generating a work experience program that does not contradict the desired working conditions of the re-employment applicant and the working conditions of the workplace where the work experience will be conducted (work experience program generation step), a step of notifying the contents of the work experience program (work experience program notification step), and a step of storing the progress status of the work experience program in memory unit 13 and updating the progress status based on information provided by the work experience applicant or the workplace (work experience program progress update step).

[0097] By including a work experience program generation step in the work experience management step, the recruiter can proceed with the work experience for the work experience applicant in accordance with the generated work experience program or a program that the recruiter has modified from the generated work experience program, without expending the effort of generating a work experience program from scratch. The work experience program generation step preferably includes a procedure for generating a work experience program in the large-scale language model 131 based on the desired treatment conditions and working conditions described above. This allows the support device 1 to generate a work experience program that reflects a variety of desired treatment conditions and working conditions.

[0098] By including the work experience management step in the work experience program notification step, the recruiter can provide the work experience applicant with a generated notification or a notification that the recruiter has modified from the generated notification, without having to expend the effort of generating the notification from scratch. The work experience program notification step preferably includes a procedure for causing the large-scale language model 131 to generate a notification related to the work experience program based on the above-mentioned work experience program. This allows the support device 1 to generate a notification about the work experience program that reflects a variety of desired treatment conditions and working conditions.

[0099] By including the work experience management step in the work experience program progress update step, the recruiter can manage the progress of the work experience program without having to manually manage the progress of the work experience program one by one. This also allows the support device 1 to generate a notification according to the progress in the work experience program notification step.

[0100] [Reemployment Management Steps] It is preferable that the support device 1 executes a re-employment management step for managing the careers, etc., of re-employment applicants who have found new employment among the re-employment applicants extracted in step S5, or who have found new employment after seeing the job advertisement provided in step S10 or step S14.

[0101] By executing the reemployment management step, the support device 1 can, for example, treat the reemployed person as a part-time employee for a given period (e.g., about three to four years) or a given number of shifts set after reemployment, and then automatically manage to promote the reemployed person to a full-time employee. Also, in the reemployment management step, the support device 1 can manage to acquire comments about reemployment from the reemployed person (word of mouth after returning to work).

[0102] The support process preferably includes a series of processes (steps S15 to S17) for receiving a workplace evaluation from a work experience applicant of the workplace where the work experience took place or a workplace evaluation from a re-employed applicant of the new employer, and analyzing the workplace evaluation. As a result, the support device 1 can provide the results of the analysis to recruiters as materials for the recruitment process and career management of the work experience applicant or re-employed applicant.

[0103] [Step S15: Determine whether workplace evaluation has been received] The control unit 11 operates the workplace evaluation receiving unit 120 in cooperation with the storage unit 13 and the communication unit 14. Then, the control unit 11 executes a process to determine whether the workplace evaluation has been received from the terminal T or the like by the workplace evaluation receiving unit 120 (step S15, workplace evaluation receiving step). If it is determined that the workplace evaluation has been received, the control unit 11 stores the received workplace evaluation in the storage unit 13 and proceeds to step S16. If it is not determined that the workplace evaluation has been received, the control unit 11 returns the process to step S1 and repeats the processes from step S1 to step S17.

[0104] [Step S16: Generate analysis results of workplace evaluation] The control unit 11, in cooperation with the storage unit 13 and the communication unit 14, executes the workplace evaluation analysis generation unit 121. Then, the control unit 11 executes a process in which the workplace evaluation analysis generation unit 121 generates analysis results of multiple workplace evaluations in the large-scale language model 131 regarding the workplace evaluations received in step S15 (step S16, workplace evaluation analysis generation step). The control unit 11 proceeds to step S17.

[0105] The workplace evaluations received in the workplace evaluation receiving step and analyzed in the workplace evaluation analysis and generation step are not particularly limited as long as they can be expressed in text. Examples of the workplace evaluations include scores indicating evaluations of the workplace, comments and reviews on the workplace, etc. The workplace evaluation analysis and generation step causes the large-scale language model 131 to generate analysis results, so that the support device 1 can generate analysis results for a variety of workplace evaluations that can be expressed in text.

[0106] The analysis results generated in the workplace evaluation analysis generation step preferably include advice on improving the working environment of the workplace related to the workplace evaluation. The workplace evaluation is expected to include text describing the working environment of the workplace, text providing hints for improving the working environment, etc. However, it may require a huge amount of effort for a human recruiter to check a large number of workplace evaluations one by one and obtain hints for improving the working environment. By including advice on improving the working environment in the analysis results generated in the workplace evaluation analysis generation step, the support device 1 can provide the advice to the recruiter.

[0107] The analysis results generated in the workplace evaluation analysis generation step preferably include analysis results regarding the personality traits of the job seeker who provided the multiple workplace evaluations. There is a view that people who post negative comments on social media and the like are likely to have a personality trait that discourages other workers from working in the workplace, etc. By including analysis results regarding the personality traits of the job seeker in the analysis results generated in the workplace evaluation analysis generation step, the support device 1 can provide recruiters with the analysis results regarding the personality traits of the job seeker as material related to the hiring process for the job seeker.

[0108] [Step S17: Provide the analysis results of the workplace evaluation] The control unit 11, in cooperation with the storage unit 13 and the communication unit 14, executes a process of providing the analysis results generated in step S16 to the terminal T, etc. (step S17, workplace evaluation analysis providing step). The control unit 11 returns the process to step S1 and repeats the processes from step S1 to step S17.

[0109] [Social Media Management Steps] The support process preferably includes an SNS management step of managing a job-seeking person SNS, which is an SNS whose main members are job-seeking people. The management may be realized, for example, by a means for executing a program that realizes the job-seeking person SNS on the support device 1, or by a means for transmitting and receiving information via an API with an external server that realizes the job-seeking person SNS.

[0110] The SNS management step allows the support device 1 to send a solicitation message via the job seeker SNS in the solicitation message sending step. In addition, the SNS management step allows the support device 1 to provide a job advertisement via the job seeker SNS in the advertisement providing step.

[0111] The SNS management step preferably includes a procedure for reflecting information about desired conditions of employment and personal information written in the profiles of members of the job-seeking SNS in the job-seeking database 132. This allows the support device 1 to reflect the members of the job-seeking SNS, their profiles, etc. in the job-seeking database 132.

[0112] When a job advertisement is provided via a job-seeking SNS in the advertisement providing step, the SNS management step preferably receives from the job-seeking SNS member activity related to the provided job advertisement (such as browsing history of posts of the job advertisement etc. on the SNS, posts of comments and reviews on the job advertisement, and evaluations of the job advertisement) as advertisement evaluations in the advertisement evaluation receiving step. This allows the support device 1 to obtain evaluations of the job advertisement from the job-seeking SNS member that is expected to have a high interest in job-seeking, and use the evaluations to provide support for job-seeking.

[0113] When a job advertisement is provided via a job-seeking SNS in the advertisement providing step, the SNS management step preferably includes a series of processes (matching step) in which input including activity of job-seeking individuals in response to the job advertisement is provided to the large-scale language model 131, a condition (third condition) that the job advertisement is consistent with the activity is generated in the large-scale language model 131, and job advertisements that satisfy the third condition are extracted from the job advertisement database 133.

[0114] As described above, the large-scale language model 131 of this embodiment is pre-trained using a large amount of text written in various languages ​​such as national languages, database languages, etc. Through such pre-training, the large-scale language model 131 acquires the emergent ability to generate text with similar content in other languages ​​used in pre-training, based on the activity shown in the text.

[0115] The matching step can generate a third condition written in a database language based on various activities by having the large-scale language model 131 generate a condition (third condition) that the job advertisement is consistent with the activity. This enables the support device 1 to realize a process of introducing the job seeker to the workplace related to the extracted job advertisement based on the activity on the job seeker SNS (recommendation step), a process of introducing the job seeker to the recruiter of the workplace related to the extracted job advertisement (scouting step), and the like.

[0116] [Reemployment document generation step] Job seekers may have a variety of backgrounds, exemplified by the circumstances surrounding their resignation. Therefore, job seekers may find it difficult to include information including their background in documents (job re-employment documents) that they submit to a new employer, such as a resume or curriculum vitae. Therefore, the support process preferably includes a job re-employment document generation step that automatically generates job re-employment documents. Job re-employment documents include, for example, a resume or curriculum vitae.

[0117] This automatic generation is realized by a series of processes including, for example, an interactive information gathering step in which information regarding re-employment documents entered by the re-employment applicant is used as input and large-scale language model 131 is made to generate text (information request text) prompting the provision of information necessary for creating the re-employment documents, and a re-employment document generation execution step in which information entered by the re-employment applicant in response to the information request text is used as input and large-scale language model 131 is made to generate the re-employment documents.

[0118] The large-scale language model 131 of this embodiment has the emergent ability to generate information request texts and new employment documents, acquired according to the amount of text used in pre-training. Therefore, the support device 1 can extract from the new job seeker a variety of information, such as desired conditions of compensation, derived from the background of the job seeker's circumstances surrounding his or her resignation, and generate new employment documents that reflect this information. The support device 1 can then register information about the new job seeker in the new employment seeker database 132 based on the new employment documents. This allows the support device 1 to support the new job seeker in creating new employment documents. Therefore, the support device 1 can reduce the burden on the new job seeker and provide new employment support that reflects a variety of information, such as desired conditions of compensation, derived from the background of the job seeker's circumstances surrounding his or her resignation.

[0119] [Recruitment interview support steps] The support process preferably includes a job interview support step of automatically generating questions to be asked by new employers to job seekers during job interviews. The automatic generation is realized, for example, by a procedure in which information about the job seekers stored in the job seeker database 132 is used as input and the questions are generated in the large-scale language model 131.

[0120] The large-scale language model 131 of this embodiment has an emergent ability to generate job interview questions, which is acquired according to the amount of text used in pre-training. Therefore, the support device 1 can generate questions that reflect a variety of information, such as the desired working conditions and working conditions at the new job, that are derived from the background of the job seeker's circumstances surrounding his or her resignation. This reduces the burden on the recruiter at the new job seeker's place of employment, and enables the support device 1 to provide job seeker support that reflects a variety of information, such as the desired working conditions and working conditions at the new job, that are derived from the background of the job seeker's circumstances surrounding his or her resignation.

[0121] [Automatic response step] The support processing preferably includes an automated recruitment response step of automatically responding to contact received by telephone, chat, etc. from job seekers. The automated response is realized, for example, by an interactive information collection step of converting the received contact into text using a voice recognition module as needed, generating a response text for responding to the contact in large-scale language model 131 using information about the job seeker stored in job seeker database 132 and the contact in text format as input, converting the generated response text into voice using a voice synthesis module as needed, and sending a response based on the response text in a format appropriate to the contact means used to receive the call (voice for telephone, text for chat, etc.), and a procedure of generating questions in large-scale language model 131 using information entered by the job seeker in response to the information request text as input.

[0122] The large-scale language model 131 of this embodiment has an emergent ability for generating response texts, which is acquired according to the amount of text used in pre-training. Therefore, the support device 1 can generate response texts that reflect diverse information exemplified by the desired working conditions and working conditions at the new job destination, which are derived from the background of the job seeker's circumstances surrounding his or her resignation. This reduces the burden on the recruiter at the new job destination, and enables the support device 1 to provide job re-employment support that reflects diverse information exemplified by the desired working conditions and working conditions at the new job destination, which are derived from the background of the job seeker's circumstances surrounding his or her resignation.

[0123] [Effect of support processing] In the above-mentioned support process, the re-employment applicant extraction unit 114 extracts information on re-employment applicants who meet the working conditions from a re-employment applicant database 132 (a specified database) in which the desired conditions for treatment of re-employment applicants are stored in association with re-employment applicant information.

[0124] Incidentally, information extraction from a database is usually performed using processing based on data extraction commands written in a query language (database language) for manipulating data in a database, such as SQL (Structured Query Language), or commands written in a database language, such as software processing via an API (Application Programming Interface). Therefore, it is not easy to extract data under conditions that reflect a variety of working conditions and desired treatment conditions.

[0125] The large-scale language model 131 of this embodiment is a language model that has undergone pre-training using a large amount of text written in various languages, such as the languages ​​of various countries, database languages, etc. Through such pre-training, the large-scale language model 131 has acquired the emergent ability to generate text with similar content in other languages ​​used in the pre-training, based on natural language relating to various working conditions and various desired treatment conditions, exemplified by sentences expressing desires regarding balancing childcare and work.

[0126] In the above-described support process, the condition generation unit 113 can generate first conditions written in a database language based on the various working conditions and various desired treatment conditions written in natural language by having the large-scale language model 131 generate conditions (first conditions) that the job seeker is associated with desired treatment conditions that do not contradict the working conditions. This allows the support device 1 of this embodiment to both utilize a database of the prior art that is operated using a database language and extract first job seeker information from the job seeker database 132 based on the various working conditions and various desired treatment conditions written in natural language.

[0127] Therefore, the support device 1 of this embodiment can provide reemployment support that reflects a variety of information exemplified by desired conditions of treatment and the like that are brought about by the background related to the circumstances of the resignation of the reemployment seeker.

[0128] <Usage example> The following is an example of how the support device 1 of this embodiment is used.

[0129] [Database update] The support device 1 updates the information stored in the job seeker database 132 based on the desired conditions and personal information provided by the job seekers, and the profiles of members acquired from the job seeker SNS.

[0130] [Providing a recruitment platform] The support device 1 receives working conditions from a terminal T or the like used by a recruiter at a new job who has subscribed to a fixed-price service that introduces new job seekers. The support device 1 generates in the large-scale language model 131 a condition (first condition) that the new job seeker is associated with desired treatment conditions that do not contradict the working conditions, and extracts information on first new job seekers who satisfy the first condition from the new job seeker database 132. In this way, the support device 1 can provide a recruiting platform that introduces new job seekers who meet the working conditions to recruiters or the like.

[0131] [Matching for reemployment] Based on the desired conditions of employment and the activity of the job seeker regarding the job advertisements on the job seeker SNS (for example, job advertisements and workplaces that have been "liked," job advertisements and workplaces with favorable comments, job advertisements and workplaces that have been favorably reviewed, etc.), the support device 1 generates in the large-scale language model 131 a condition (third condition) that the job advertisements are consistent with the activity, and extracts job advertisements that satisfy the third condition from the job advertisement database 133. Then, based on the job advertisements that are considered to be compatible with the job seeker related to the activity extracted based on the third condition, the support device 1 performs matching such as sending the job seeker an invitation to work at the job destination related to the job advertisement, or recommending the job seeker to the recruiter related to the job advertisement.

[0132] [Work experience registration] The support device 1 registers in the storage unit 13 a work experience applicant who has become interested in work experience at a new employer through the above-mentioned recruitment platform and matching. The support device 1 then generates a work experience program, sends notifications related to the work experience program to the work experience applicant, and manages the progress of the work experience program. The support device 1 also receives word-of-mouth about the work experience (work experience word-of-mouth) and the like as workplace evaluations, generates analysis results thereof, and provides advice on improving the work environment and analysis results related to the personality tendencies of the new employer to the recruiter.

[0133] [Management after returning to work] The support device 1 registers in the storage unit 13 reemployed individuals (returnees) who have found employment (returned to work) at a new employer through the above-mentioned recruitment platform and matching. The support device 1 then generates a work experience program and manages the careers, etc., of the reemployed individuals. The support device 1 also receives word-of-mouth reviews (word-of-mouth reviews after returning to work) and the like about the new employer as workplace evaluations, generates the results of its analysis, and provides advice on improving the work environment and analysis results on the personality tendencies of reemployment applicants to recruiters.

[0134] It should be noted that within the scope of the concept of the present invention, those skilled in the art may conceive of various modifications and alterations. Therefore, it is understood that such modifications and alterations fall within the scope of the present invention. For example, even if a person skilled in the art appropriately adds, deletes, or modifies components of the above-described embodiment, or adds, omits, or modifies the conditions of a process, such modifications are also included within the scope of the present invention as long as they maintain the gist of the present invention. [Explanation of symbols]

[0135] S System 1 Support equipment 11 Control section 111 Machine Learning Department 112 Condition receiving unit 113 Condition generator 114 Reemployment applicant extraction department 115 Solicitation Text Generation Unit 116 Job Advertisement Generation Department 117 Advertising Department 118 Advertising Evaluation Reception Department 119 Job Advertisement Extraction Unit 120 Workplace Evaluation Reception Department 121 Workplace Evaluation Analysis and Generation Department 13 Storage section 131 Large-scale Language Models 132 Reemployment Applicant Database (prescribed database) 133 Job Advertisement Database (Specific Database) 14 Communications Department N Network T-Terminal

Claims

1. a re-employment applicant extraction unit that extracts information on re-employment applicants that meets designated conditions from a predetermined database that stores information on the re-employment applicants (re-employment applicant information) that includes the desired conditions for treatment of the re-employment applicants; a condition receiving unit that receives working conditions; a condition generation unit that, when the working conditions are received, provides an input including the working conditions to a large-scale language model via an API and causes the large-scale language model to generate conditions (first conditions) for extracting job seekers whose desired treatment conditions are consistent with the working conditions in all respects; Equipped with the first condition is written in a database language used to extract the job seeker information from the predetermined database; pre-training the large-scale language model using training data including the working conditions as explanatory variables and the first condition as a target variable; the job seeker extraction unit extracts, when the first condition is generated, first job seeker information that satisfies the first condition from a predetermined database; A support device for reemployment.

2. 2. The support device according to claim 1, further comprising an invitation message generation unit that, when the first job seeker information is extracted from a predetermined database, causes the large-scale language model that has been pre-trained using training data that includes the desired working conditions as an explanatory function and the invitation message addressed to the job seeker associated with the desired working conditions as a target variable to generate, via an API, an invitation message addressed to the job seeker related to the first job seeker information based on the desired working conditions associated with the first job seeker information.

3. a job advertisement generation unit that, when the working conditions are received, generates, via an API, a first job advertisement that reflects the working conditions based on the working conditions, using the large-scale language model that has been pre-trained using training data that includes the working conditions as an explanatory function and the first job advertisement as a target variable; The support device according to claim 1 .

4. an advertisement providing unit that provides the first job advertisement to a predetermined terminal used by the job seeker; an advertisement evaluation receiving unit that receives an evaluation (advertisement evaluation) of the first job advertisement from the predetermined terminal; a job advertisement extraction unit that extracts second job advertisements that satisfy specified conditions from a specific database in which the first job advertisements are stored; Furthermore, pre-training the large-scale language model using training data including job advertisements and evaluations of the job advertisements as explanatory variables and conditions written in a database language that are similar to the working conditions shown in the job advertisements related to the evaluations determined to be favorable as objective variables; the condition generation unit provides the large-scale language model with an input including a job advertisement that reflects various evaluations expressed in text and various working conditions, and causes the large-scale language model to generate a condition (second condition) in which text similar to the working conditions shown in the job advertisement related to the evaluation determined to be favorable is written in a database language; When the second condition is generated, the job advertisement extraction unit extracts second job advertisements that satisfy the second condition from a specific database. The support device according to claim 3 .

5. The advertisement evaluation includes an advertisement evaluation written in natural language (advertisement evaluation text), The support device includes: The large-scale language model further includes an advertisement evaluation analysis generation unit that provides an input including the advertisement evaluation text to the large-scale language model that has been pre-trained using training data that includes an advertisement evaluation text as an explanatory function and an analysis result of the advertisement evaluation text as a target variable, and causes the large-scale language model to generate an analysis result of the advertisement evaluation text. The support device according to claim 4.

6. a workplace evaluation receiving unit that receives an evaluation of a workplace where the new employee will be employed (workplace evaluation) from the new employee; a workplace evaluation analysis generation unit that provides an input including the workplace evaluations to the large-scale language model that has been pre-trained using training data that includes a plurality of workplace evaluations as explanatory functions and includes analysis results of the plurality of workplace evaluations as objective variables, and causes the large-scale language model to generate analysis results of the workplace evaluations; Further provided with The support device according to claim 1 .

Images