Recruitment page generation device

Abstract

To provide a recruitment page generation device for reducing a burden on recruitment staff relating to generation of recruitment pages by providing the recruitment staff with a model for recruitment pages, etc.SOLUTION: A generation device 1 includes: a requirement receiving part 111 for receiving recruitment requirements; a manuscript generation part 113 for providing input including the recruitment requirements to a large-scale language model 121 and causing the large-scale language model 121 to generate a first recruitment manuscript, the recruitment manuscript reflecting the recruitment requirements; a proofread manuscript receiving part 114 for receiving a proofread manuscript, the proofread first recruitment manuscript, etc.; and a machine learning part 116 using learning data containing the recruitment requirements and the first recruitment manuscript as explanatory variables and the proofread manuscript as an objective variable, to train the large-scale language model 121 by machine learning to generate the recruitment manuscript with the recruitment requirements as input, reflecting the recruitment requirements.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a recruitment page generating device. [Background technology]

[0002] There are workplaces where the working patterns are complex and the personnel turnover is relatively high, such as nursing and care work. In workplaces where the personnel turnover is relatively high, recruitment is frequent and the work related to recruitment can be a burden on the workplace. Therefore, there is a demand for a means to support recruitment.

[0003] (Prior Art) As an example of a conventional technique for supporting job hunting, Patent Document 1 discloses an employment support device that calculates the similarity between behavioral characteristic information (first behavioral characteristic information) generated by inputting aggregate information on the feature amounts of email sending and receiving history of persons belonging to a company into a predetermined mathematical model, and behavioral characteristic information (second behavioral characteristic information) generated by inputting aggregate information on the feature amounts of email sending and receiving history of a job seeker into the above-mentioned predetermined mathematical model. The technique in Patent Document 1 allows job seekers to obtain information on suitable companies without any conscious effort, and companies can collect information on job seekers that is useful for deciding whether to hire or reject the job seeker. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2010-231685 A Summary of the Invention [Problem to be solved by the invention]

[0005] Incidentally, in workplaces with complex working patterns, there can be a wide variety of job-seeking conditions (job conditions), such as employment type, working hours, and required skills. In order to create a job-seeking page that reflects the wide variety of job-seeking conditions, a high level of skill is required to create the job-seeking page. In addition to the wide variety of job-seeking conditions, when the turnover of human resources is relatively high and job offers are made frequently, a high level of skill is required, and therefore an enormous burden for creating the job-seeking page can be concentrated on advertising personnel who have few replacement personnel. The technology of Patent Document 1 is merely capable of providing convenience to job seekers and companies through the calculation of similarity, and there is room for further improvement in terms of reducing the burden on advertising personnel.

[0006] The present invention has been made in consideration of the above circumstances, and its purpose is to reduce the burden on recruiters in creating recruitment pages, for example by providing recruiters with model recruitment pages. [Means for solving the problem]

[0007] As a result of intensive research into solving the above problem, the inventors have found that the above object can be achieved by generating job postings using a large-scale language model that has been subjected to machine learning using proofread job postings, etc. The inventors have then completed the present invention. Specifically, the present invention provides the following:

[0008] The present invention provides a recruitment page generation device comprising: a conditions receiving unit that receives job conditions; a manuscript generation unit that provides an input including the job conditions to a large-scale language model and causes the large-scale language model to generate a first job condition, which is a job condition reflecting the job conditions; a proofread manuscript receiving unit that receives a proofread manuscript, which is the first job condition that has been proofread; and a machine learning unit that uses learning data that includes the job conditions and the first job condition as explanatory variables and the proofread manuscript as a target variable to machine learn the large-scale language model to generate a job condition that reflects the job conditions, taking the job conditions as input.

[0009] If there is a device that can automatically generate a recruitment page, such a device can reduce the burden on recruiters by using the generated recruitment page as is, or by using the generated recruitment page as a model for recruiters. Regarding the automatic generation of a recruitment page, for example, a procedure for generating a recruitment page by combining standard phrases corresponding to each individual condition included in the recruitment conditions is possible. However, when creating a recruitment page that reflects a variety of recruitment conditions, a huge amount of standard phrases must be prepared in advance, which may increase the burden on recruiters.

[0010] Incidentally, a large-scale language model is a language model that has been pre-trained using a large amount of text including recruitment pages, etc. Through such pre-training, the large-scale language model has acquired the emergent ability to generate a recruitment page that reflects recruitment conditions when the recruitment conditions are input. However, if an attempt is made to generate a recruitment page that reflects various recruitment conditions using only the above-mentioned emergent ability, there may be a deficiency in any of the various requirements exemplified by the match between the recruitment conditions and the recruitment page, the ability to appeal to job seekers, etc., and a recruitment page that is not exemplary may be generated.

[0011] In the present invention, a proofread draft of the first job posting generated by the present invention is received by a proofread draft receiving unit, and machine learning is performed using training data that includes the job requirements and the first job posting as explanatory variables and the received proofread draft as a target variable, thereby allowing a large-scale language model to learn the abilities that are insufficient with only the above-mentioned emergent abilities. As a result, the present invention can cause the large-scale language model that has undergone machine learning to generate a model recruitment page and provide it to the recruiter. Therefore, the present invention can reduce the burden on the recruiter in generating the recruitment page. Effect of the Invention

[0012] The present invention can reduce the burden on recruiters involved in creating recruitment pages by, for example, providing recruiters with model recruitment pages. [Brief description of the drawings]

[0013] [Figure 1] FIG. 1 is a block diagram showing the hardware and software configurations of a system S according to the present embodiment. [Diagram 2] FIG. 2 is an example of the job posting script database 122. [Diagram 3] FIG. 3 is a main flowchart showing an example of a preferable flow of the generation process executed by the generation device 1 of this embodiment. [Figure 4] FIG. 4 is a continuation of FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Hereinafter, an example of an embodiment of the present invention will be described in detail with reference to the drawings.

[0015] <System S> 1 is a block diagram showing the hardware and software configurations of a system S of this embodiment. Hereinafter, an example of a preferred embodiment of the hardware and software configurations of the system S of this embodiment will be described with reference to FIG.

[0016] The system S includes a recruitment page generating device 1 and a terminal T configured to be able to communicate with the generating device 1 via a network N.

[0017] [Generation device 1] The generating device 1 includes a control unit 11, a storage unit 12, and a communication unit 13. There is no particular limitation on the type of the generating device 1. Examples of the type include a server and a cloud server.

[0018] The generation device 1 is configured to execute a process of making a large-scale language model that has been pre-trained using a large amount of text perform machine learning, and a process of making the large-scale language model that has been subjected to machine learning generate a recruitment page (job posting). The execution of these processes is realized, for example, via an API related to the use of the large-scale language model. The following is a more detailed configuration example of the generation device 1.

[0019] [Control unit 11] The control unit 11 includes a central processing unit (CPU), a random access memory (RAM), and a read only memory (ROM).

[0020] The control unit 11 cooperates with the storage unit 12 and / or the communication unit 13 as necessary. The control unit 11 then realizes software components of the program of this embodiment executed by the generation device 1, such as a condition receiving unit 111, a draft receiving unit 112, a manuscript generation unit 113, a proofread manuscript receiving unit 114, an evaluation receiving unit 115, and a machine learning unit 116. 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 generation process described later.

[0021] [Storage unit 12] The memory unit 12 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, and the like.

[0022] The storage unit 12 may have a mechanism that enables connection via the network N to a storage device or storage system such as a network attached storage (NAS), a storage area network (SAN), cloud storage, a file server, and / or a distributed file system.

[0023] The storage unit 12 stores a program executed by the microcomputer, a large-scale language model 121, a job advertisement script database 122, and the like.

[0024] (Large-scale language model 121) The large-scale language model 121 (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 / sentence is to occur as natural language. The generation device 1 inputs a sentence of generation data described below to the large-scale language model 121, predicts a sentence of an adoption page that is likely to occur following the sentence of the generation data, and generates the predicted adoption page sentence.

[0025] The large-scale language model 121 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 121, or may be stored in the storage unit 12.

[0026] Since the large-scale language model 121 is executed via an API, the generation device 1 has a simpler configuration and is more cost-effective than a configuration in which the large-scale language model 121 is stored in the storage unit 12.

[0027] Since the large-scale language model 121 is stored in the storage unit 12, the generation device 1 can execute processing related to the large-scale language model 121 without communicating with an external device such as a cloud server one by one. This allows the generation device 1 to reduce processing delays and information security risks related to communication. The following description focuses on the large-scale language model 121 being stored in the storage unit 12, but it will be understood by those skilled in the art that a similar embodiment can be realized in which the large-scale language model 121 is executed via an API in an external cloud server or the like.

[0028] The large-scale language model 121 is trained on a large amount of text, at least 500 GB (gigabytes) or more. Examples of such models include ChatGPT, GPT-3.5, and GPT-4 from OpenAI (registered trademark), and LLaMa from Meta AI (registered trademark).

[0029] It is believed that as the amount of training data increases, the performance of the language model increases along the broken neural scaling law (BNSL), and the model acquires emergent capabilities when transitioning between segments in the BNSL. Therefore, the large-scale language model 121 of this embodiment has an emergent capability to generate a job posting that reflects the job posting conditions when the job posting conditions are input, which is acquired according to the amount of text used in pre-training.

[0030] The large-scale language model 121 preferably has an upper limit of 30,000 or more contexts used to generate text when counting tokens, which are semantic chunks of text (words, etc.). This allows the large-scale language model 121 to generate job postings that take into account job conditions described using a large number of tokens as context to include many conditions. This also allows the large-scale language model 121 to perform machine learning that takes into account draft texts described using a large number of tokens as context.

[0031] The large-scale language model 121 is preferably trained on texts containing 10 trillion or more tokens, which is believed to give the large-scale language model 121 the emergent ability to generate job postings that reflect nuances according to the type of job at the workplace where the job is being advertised.

[0032] The large-scale language model 121 is preferably a model having a large number of parameters, at least 10 billion or more, so that the large-scale language model 121 can execute processing that fully reflects the emergent capabilities obtained from a large amount of text.

[0033] (Job Script Database 122) The job posting database 122 stores information that associates data used to generate a job posting (data for generation) with the generated job posting. The associated information is, for example, information that associates the job posting conditions with a first job posting generated based on the job posting conditions, information that associates a draft of a job posting with a second job posting generated based on the draft, etc. The job postings such as the first job posting and the second job posting are preferably stored in association with improvement information such as a proofread draft of the job posting and an evaluation of the job posting. This allows the generation device 1 to perform machine learning based on the improvement information.

[0034] The job posting database 122 is configured to extract job postings that satisfy predetermined conditions, thereby allowing the generation device 1 to extract job postings that satisfy predetermined conditions exemplified by the job posting conditions and the like.

[0035] The job postings stored in the job posting database 122 are not particularly limited. The job postings include various types of advertisements, exemplified by, for example, a text format suitable for transmission via e-mail, transmission via short message, posting to SNS, etc., a modified text format (e.g., HTML (HyperText Markup Language), etc.) that can achieve higher expressiveness than text in e-mail transmission, a format including text and images (e.g., PDF (Portable Document Format), web pages, etc.), a voice format suitable for transmission via voice calls, voice messages, etc., an image format suitable for image advertisements, an expressive video format, etc.

[0036] The job scripts stored in the job script database 122 preferably contain text. When the job script contains text, it becomes easier to extract the text based on the condition of the text generated by the large-scale language model 121. When the job script contains multimedia information such as images, audio, and video, it is preferable that the job script is 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.

[0037] FIG. 2 is an example of the job posting script database 122. In the example shown in FIG. 2, the job posting script identified by ID "A0001" reads "We are looking for a registered nurse (new graduates and those with experience welcome). Educational background and experience are not required, and the target is 64 years old or younger. There is a track record of taking childcare leave, nursing care leave, and nursing care leave, and we value work-life balance. Working hours: 9:00-18:00 (60-minute break), no overtime. Salary: △△ yen to △△ yen. Ample bonuses and allowances. You can also improve your skills through comprehensive training. Why not use your skills on our team? We look forward to your application." is generated based on the job posting conditions "1. Qualifications, etc. (1) Educational background and experience are not required (registered nurse required) (2) 64 years old or younger *Due to the retirement age of 65 2. Working conditions (1) There is a track record of taking childcare leave, nursing care leave, and nursing care leave (2) Working hours: 9:00-18:00 (60-minute break), no overtime" (3) Salary: △△ million to △△ million yen," and is linked to the revised draft as improvement information: "[Recruiting registered nurses (day shift)] We are recruiting registered nurses. No educational background or experience required, eligible for those under 64 years old. We have a track record of taking parental leave, nursing care leave, and nursing care leave, so you can balance work with family and childcare. Working hours (day shift only): 9:00-18:00 (60-minute break), no overtime. Salary: △△ million to △△ million yen. Generous bonuses and allowances. Why not put your skills to good use on our team? We look forward to your application."

[0038] Also, in the example shown in Figure 2, the job posting identified by ID "A0002" reads: "(New nurse wanted) Educational background and experience not required, able to commute to △△ city △△ town, 3-shift work. Day shift 8:45-17:00 (60-minute break), semi-night 16:30-24:00 (45-minute break), late night 24:00-32:45 (90-minute break). Can work 5 days or more per week, employment period 3-6 months or 6 months or more. 8-10 semi-night and late night shifts per month. Salary △△ yen - △△ yen. Plenty of benefits. Why not grow in a rewarding workplace?" We look forward to your application." corresponds to the proposed text as data to be generated, "[Nurses wanted] Educational background and experience are not required, and people who can commute to △△ town in △△ city. There are three shifts, with day shifts from 8:45 to 17:00 (60-minute break), semi-night shifts from 16:30 to 24:00 (45-minute break), and night shifts from 24:00 to 22:45 (90-minute break). People who can work at least 5 days a week for either within 3-6 months or for more than 6 months. There are 8-10 semi-night and night shifts per month. Monthly salary is △△ yen to △△ yen.", and is associated with the evaluation as improvement information, "[△△ city has a low number of young people, and it is important for veterans to return to work, so it would be better to write in a way that does not limit it to new nurses, even if educational background and experience are not required."

[0039] As a result, the generation device 1 can store in the job script database 122 the job script generated by the large-scale language model 121 based on an input including generation data indicating various job requirements, and provide it to the terminal T, etc. In addition, as a result, the generation device 1 can cause the large-scale language model 121 to learn by machine learning the generation of a job script based on the job script and improvement information stored in the job script database 122.

[0040] [Communications Section 13] The communication unit 13 is not particularly limited as long as it connects the generating device 1 to the network N and enables communication with the terminal T, etc. Examples of the communication unit 13 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.

[0041] [Network N] The type of the network N is not particularly limited as long as it enables communication between the generating 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.

[0042] [Terminal T] Terminal T is not particularly limited. Terminal T is, for example, any of various terminals used by people involved in job postings, such as a person who creates a job posting, such as a recruiter, or a person who reviews the job posting. Terminal T is configured to execute a process of transmitting data for generation to generation device 1, a process of receiving and displaying the generated job posting from generation device 1, a process of transmitting improvement information to generation device 1, and the like. In the process of transmitting improvement information to generation device 1, it is preferable to transmit information identifying the job posting corresponding to the improvement information (identification information) together with the improvement information.

[0043] [Main Flowchart of Generation Processing] Fig. 3 is a main flow chart showing an example of a preferred flow of the generation process executed by the generation device 1 of this embodiment. Fig. 4 is a diagram continuing from Fig. 3. The following is an example of a preferred flow of the generation process executed by the generation device 1 of this embodiment using Figs. 3 and 4.

[0044] The generation process includes a series of processes (steps S1 to S2) related to the generation of a job posting based on the received job posting conditions. This allows the generation device 1 to automatically generate a job posting for a hiring page that reflects various job posting conditions.

[0045] [Step S1: Determine whether job conditions have been received] The control unit 11 executes the condition receiving unit 111 in cooperation with the storage unit 12 and the communication unit 13. Then, the control unit 11 executes a process of determining whether the condition receiving unit 111 has received the recruitment conditions from the terminal T or the like (step S1, condition receiving step). If it is determined that the recruitment conditions have been received, the control unit 11 stores the received recruitment conditions in the storage unit 12 and moves the process to step S2. If it is determined that the recruitment conditions have not been received, the control unit 11 moves the process to step S3.

[0046] (Recruitment conditions) The recruitment conditions received in the conditions receiving step are not particularly limited as long as they can be given as input text to the large-scale language model 121. The recruitment conditions include a wide variety of conditions, exemplified by, for example, the job content, the place of work, the presence or absence of transfers and details thereof, working hours (including non-standard desired time periods such as desired time periods that vary depending on the day of the week and desired time periods 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, the presence or absence of social insurance and details thereof, the presence or absence of childcare facilities and details thereof, the presence or absence of childcare support and details thereof, the presence or absence of retraining support and details thereof, the presence or absence of 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.), the atmosphere of the workplace, the hobbies of the superior, workplace customs, etc.

[0047] The job conditions received in the condition receiving step are preferably hierarchically displayed according to semantic relationships, with the above-mentioned various conditions being divided into items. The hierarchical display of the job conditions is, for example, a hierarchical display of a plurality of major items, intermediate items provided under each of the major items, and small items provided under each of the intermediate items as necessary.

[0048] The above-mentioned hierarchical structure is explained below with reference to FIG. 2. The job posting identified by ID "A0001" in the example shown in FIG. 2 and the corresponding job posting conditions are an example of job posting conditions divided into categories and hierarchical. In the job posting conditions, "1. Qualifications, etc." and "2. Working conditions" are shown as major items corresponding to the highest level, and conditions related to educational background, etc. (1) and conditions related to age, etc. (2) are shown as medium items under the major item "1. Qualifications, etc.", and conditions related to leave, etc. (1), conditions related to working hours, etc. (2), and conditions related to salary, etc. (3) are shown as medium items under the major item "2. Working conditions." Based on this example, a person skilled in the art can easily conceive of various types of hierarchical structure, such as a hierarchy that further includes minor items based on the above example, and a hierarchy that further includes detailed items, etc. under the minor items.

[0049] By hierarchizing the job requirements, in the first draft generation step described below, the large-scale language model 121 can appropriately grasp the job requirements based on the job requirements whose semantic relationships have been organized by the hierarchization, and generate a first job draft that appropriately reflects the job requirements thus grasped.

[0050] [Step S2: Generate a job posting that reflects the received job requirements] The control unit 11 executes the manuscript generation unit 113 in cooperation with the storage unit 12 and the communication unit 13. The control unit 11 then executes a process in which the manuscript generation unit 113 generates a job posting (first job posting) reflecting the job posting conditions received in step S1 in the large-scale language model 121 (step S2, first manuscript generation step). The control unit 11 moves the process to step S5.

[0051] Through the above-mentioned pre-learning in the large-scale language model 121, the large-scale language model 121 has acquired an emergent ability for understanding job requirements. Then, by using the large-scale language model 121 that has been pre-learned and has acquired the emergent ability, the generation device 1 generates a job posting that reflects the job requirements, based on the job requirements. In this way, the generation device 1 can reduce the labor of recruiters.

[0052] [Step S3: Determine whether the draft text has been received] The control unit 11 executes the draft text receiving unit 112 in cooperation with the storage unit 12 and the communication unit 13. The control unit 11 then executes a process to determine whether a draft text of a job posting has been received from the terminal T or the like by the draft text receiving unit 112 (step S3, draft text receiving step). If it is determined that the draft text has been received, the control unit 11 stores the received draft text in the storage unit 12 and moves the process to step S4. If it is not determined that the draft text has been received, the control unit 11 moves the process to step S5.

[0053] (Draft job posting text) The draft text of the job posting received in the draft text receiving step is not particularly limited as long as it can be given as an input text to the large-scale language model 121. The draft text includes draft texts of job postings used in a wide variety of recruitment pages, exemplified by the recruitment page reflecting the above-mentioned recruitment conditions.

[0054] [Step S4: Generate a job posting that reflects the job requirements shown in the draft] The control unit 11 executes the manuscript generation unit 113 in cooperation with the storage unit 12 and the communication unit 13. The control unit 11 then executes a process in which the manuscript generation unit 113 causes the large-scale language model 121 to generate a job posting (second job posting) reflecting the job requirements indicated in the draft text based on the draft text received in step S3 (step S4, second draft generation step). The control unit 11 then moves the process to step S5.

[0055] Through the above-mentioned pre-learning in the large-scale language model 121, the large-scale language model 121 has acquired emergent capabilities related to understanding the job requirements indicated in the draft text. Then, by using the large-scale language model 121 that has been pre-learned and has acquired emergent capabilities, the generation device 1 understands the job requirements indicated in the draft text based on the draft text, and generates a job manuscript that reflects the job requirements indicated in the draft text. In this way, the generation device 1 can reduce the labor of recruiters.

[0056] [Step S5: Submit job posting] The control unit 11, in cooperation with the storage unit 12 and the communication unit 13, executes a process of providing the job posting generated in step S2 or step S4 to the terminal T or the like (step S5, job posting step). The control unit 11 moves the process to step S6.

[0057] The destination to which the job posting script is provided in the job posting script providing step is not particularly limited. The destination to which the job posting script is provided is, for example, a terminal T used by a recruiter who creates a recruitment page. In this way, the generation device 1 provides the recruiter with a job posting script that serves as a reference for creating a recruitment page or a job posting script that serves as a model for creating a recruitment page, thereby reducing the labor of the recruiter. Thus, the generation device 1 can reduce the burden on the recruiter in generating a recruitment page by, for example, providing the recruiter with a model recruitment page.

[0058] The generation process includes a series of processes (steps S6 to S8) related to machine learning based on the received proofread manuscript. This allows the generation device 1 to learn based on the content and intent of the proofread manuscript and generate a job posting that is more appropriate for the job requirements.

[0059] [Step S6: Determine whether the proofread manuscript has been received] The control unit 11 executes the proofread manuscript receiving unit 114 in cooperation with the memory unit 12 and the communication unit 13. The control unit 11 then executes a process to determine whether the proofread manuscript receiving unit 114 has received a proofread manuscript, which is a job posting script that has been proofread, from the terminal T or the like (step S6, proofread manuscript receiving step). If it is determined that the proofread manuscript has been received, the control unit 11 associates the received proofread manuscript with the job posting script corresponding to the proofread manuscript, stores the proofread manuscript in the job posting script database 122, and moves the process to step S7. If it is determined that the proofread manuscript has not been received, the control unit 11 moves the process to step S9. Note that the "job posting script" in the proofread manuscript receiving step is not particularly limited as long as it is a job posting script generated by the generation device 1, and may be, for example, a first job posting script, a second job posting script, etc.

[0060] The means for identifying the job posting corresponding to the proofread draft in the proofread draft receiving step is not particularly limited. Such means may be, for example, a procedure for identifying the job posting using information related to identifying the job posting sent together with the proofread draft (identification information use procedure), or a procedure for identifying the job posting using the history of the job posting sent to terminal T that sent the proofread draft (context use procedure). When the means for identifying the job posting includes an identification information use procedure, generation device 1 can reliably identify the job posting. When the means for identifying the job posting includes a context use procedure, generation device 1 can identify the job posting without requiring a recruiter or the like to expend the effort of adding identification information.

[0061] [Step S7: Get job requirements] The control unit 11, in cooperation with the storage unit 12, executes a process of acquiring the job requirements corresponding to the proofread manuscript received in step S6 from the job posting manuscript database 122 (step S7, job requirements acquisition step). The control unit 11 moves the process to step S8.

[0062] [Step S8: Machine learning to generate job posting scripts] The control unit 11 executes the machine learning unit 116 in cooperation with the storage unit 12 and the communication unit 13. The control unit 11 then executes a process in which the machine learning unit 116 causes the large-scale language model 121 to perform machine learning to generate a job posting that reflects the job posting conditions, using learning data that includes the job posting conditions acquired in step S7 and the job posting script identified in step S6 as explanatory variables and the proofread version received in step S6 as a target variable (step S8, first machine learning step). The control unit 11 moves the process to step S9.

[0063] By automatically generating a recruitment page, the generation device 1 can reduce the burden on recruiters by using the generated recruitment page as is, or by using the generated recruitment page as a model for recruiters, etc. Meanwhile, the large-scale language model 121 is a language model that has been pre-trained using a large amount of text including a recruitment page, etc., but if the pre-training is insufficient, there is a concern that a recruitment manuscript that is not a model will be generated due to a lack of any of the various requirements exemplified by the match between the recruitment conditions and the recruitment manuscript, the ability to appeal to job seekers, etc.

[0064] The generation device 1 of this embodiment receives the proofread draft of the job posting generated by the generation device 1 in the proofread draft receiving unit 114, and performs machine learning using learning data that includes the job requirements and the job posting draft as explanatory variables and the received proofread draft as a target variable, thereby allowing the large-scale language model 121 to learn the capabilities that are insufficient only with the above-mentioned emergent capabilities. This allows the generation device 1 to perform machine learning using the proofread draft as a model. The generation device 1 can then cause the large-scale language model 121 on which the machine learning has been performed to generate a model recruitment page and provide it to the recruiter. Thus, the generation device 1 can reduce the burden on the recruiter in generating the recruitment page.

[0065] The generation process includes a series of processes (steps S9 to S11) related to machine learning based on the received evaluations. This allows the generation device 1 to learn based on the contents and intentions of the evaluations and generate a job posting that is more appropriate for the job requirements.

[0066] [Step S9: Determine whether evaluation has been received] The control unit 11 executes the evaluation receiving unit 115 in cooperation with the storage unit 12 and the communication unit 13. The control unit 11 then executes a process to determine whether an evaluation on the first job posting has been received from the terminal T or the like by the evaluation receiving unit 115 (step S9, evaluation receiving step). If it is determined that the evaluation has been received, the control unit 11 associates the received evaluation with the first job posting corresponding to the evaluation, stores the evaluation in the job posting database 122, and proceeds to step S10. If it is not determined that the evaluation has been received, the control unit 11 returns the process to step S1 and repeats steps S1 to S11.

[0067] The evaluations include, for example, evaluations showing the effects of the recruitment page based on the generated job posting for various job conditions classified by industry, age, work location, etc. (the number of applications for jobs related to the recruitment page, the number of replies to the recruitment page on SNS, the number of likes for the recruitment page on SNS, etc.), comments added by recruiters to the generated job posting, etc.

[0068] Since the evaluation includes the effect of the recruitment page based on the generated job posting, the generating device 1 can perform machine learning based on the reaction of the actual recruitment page, i.e., machine learning reflecting the actual situation of the job posting site, in the second machine learning step described later. This allows the generating device 1 to improve the generation of the job posting so as to induce more favorable reactions.

[0069] By including the recruiter's comments attached to the generated job posting in the evaluation, the generating device 1 can perform machine learning that reflects the recruiter's specialized knowledge in the second machine learning step described below. This allows the generating device 1 to improve the generation of job postings so as to elicit a more desirable response from the recruiter.

[0070] [Step S10: Get job postings] The control unit 11, in cooperation with the storage unit 12, executes a process of acquiring the job posting and job requirements corresponding to the evaluation received in step S9 from the job posting database 122 (step S10, job posting acquisition step). The control unit 11 moves the process to step S11.

[0071] [Step S11: Machine learning to generate job posting scripts] The control unit 11 executes the machine learning unit 116 in cooperation with the storage unit 12 and the communication unit 13. The control unit 11 then executes a process in which the machine learning unit 116 performs machine learning to generate a job posting that reflects the job posting conditions, using the training data including the evaluation received in step S9 and the job posting and job posting conditions acquired in step S10 as explanatory variables, in the large-scale language model 121 (step S11, second machine learning step). The control unit 11 returns the process to step S1 and repeats steps S1 to S11.

[0072] As described above, if the large-scale language model 121 does not have enough prior training, there is a concern that it will generate a job posting that is not suitable for the job because it will lack any of the various requirements, such as the match between the job requirements and the job posting, the ability to appeal to job seekers, etc.

[0073] The generation device 1 of this embodiment receives evaluations of the job posting generated by the generation device 1 in the evaluation receiving unit 115, and performs machine learning using training data including the evaluations, job conditions, and job postings as explanatory variables, thereby allowing the large-scale language model 121 to learn the capabilities that are insufficient with only the above-mentioned emergent capabilities. Since the target of machine learning is the large-scale language model 121 that has acquired the emergent capabilities to generate job postings by itself, the generation device 1 can perform machine learning using training data that does not include a target variable. This allows the generation device 1 to revise the generation policy of the job posting based on the evaluation. Then, the generation device 1 can cause the large-scale language model 121 on which the machine learning has been performed to generate a model recruitment page and provide it to the recruiter. Therefore, the generation device 1 can reduce the burden on the recruiter in generating the recruitment page.

[0074] <Usage example> The following is an example of how the generating device 1 of this embodiment is used.

[0075] [Generating job postings] The generation device 1 receives generation data such as job requirements or draft text from a recruiter or the like via a terminal T or the like. The generation device 1 generates a job posting from the generation data using a large-scale language model 121. The generation device 1 provides the generated job posting to the terminal T or the like.

[0076] [Creating improvement information] The recruiter or the like checks the job posting generated and provided by generation device 1. Then, the recruiter or the like creates improvement information (proofreads, evaluations, etc.) to improve the job posting, if necessary.

[0077] [If the improvement information is a proofread draft] The recruiter or the like proofreads the job posting and creates a proofread draft, which is a proofread job posting. The recruiter or the like then transmits the proofread draft to the generation device 1 via a terminal T or the like. The generation device 1 receives the proofread draft and performs machine learning based on the proofread draft.

[0078] The following is an explanation of the creation of a proofread draft as improvement information and machine learning based on the proofread draft, using the job script identified by ID "A0001" in the example of job script database 122 shown in Figure 2 (hereinafter also referred to as "job script No. 1") as an example.

[0079] The first job posting reads, "We are looking for a registered nurse (new graduates and those with experience welcome). Educational background and experience are not important, and applicants must be under 64 years old. We have a track record of offering parental leave, nursing care leave, and care leave, and we emphasize work-life balance. Working hours 9:00-18:00 (60-minute break), no overtime. Salary: △△ million to △△ million yen. Generous bonuses and allowances. You can also improve your skills through comprehensive training. Why not put your skills to good use on our team? We look forward to your application."

[0080] The recruiter who reviewed the No. 1 job posting thought it should be emphasized that the job was for "registered nurses working day shifts," and thought that instead of emphasizing it by enclosing "[new graduates and experienced applicants welcome]" in tortoiseshell brackets, it should be changed to "[registered nurses wanted (day shift)]." Also, due to the local circumstances of the area, where there are many nurses who have taken time off work to raise children, the recruiter thought that rather than simply saying "emphasis on work-life balance," it should be more specific, saying "able to balance work with family and childcare." In addition, the recruiter thought that the skill-up training was not particularly comprehensive compared to other workplaces, and thought that the phrase "comprehensive training allows you to improve your skills" should be deleted.

[0081] Then, the recruiter proofread the No. 1 job posting script based on these ideas, created the proofread version of the No. 1 job posting script, "[Registered nurse wanted (day shift)] We are looking for a registered nurse. No educational background or experience requirement, eligible for age 64 or younger. There is a track record of taking parental leave, nursing care leave, and care leave, so work can be balanced with family and childcare. Working hours (day shift only): 9:00-18:00 (60-minute break), no overtime. Salary: △△ yen to △△ yen. Generous bonuses and allowances. Why not put your skills to good use on our team? We look forward to your application," and sent this to the generation device 1 as improvement information.

[0082] The generation device 1 received this proofread draft and performed machine learning from this proofread draft and the job postings and job conditions stored in the job posting draft database 122.

[0083] Specifically, the generation device 1 has machine-learned that when the job requirements include "1. Qualifications, etc. (1) Educational background and experience not required (registered nurse required)" and "(2) Working hours 9:00-18:00 (60-minute break), no overtime", it is better to use "Registered nurse wanted (day shift)" as the heading to be emphasized with tortoiseshell brackets, rather than "New graduates and experienced applicants welcome". In addition, the generation device 1 has machine-learned that when the job requirements include "2. Working conditions (1) Childcare leave, nursing care leave, and nursing care leave available", it is better to specifically write "Can balance work with family and childcare" rather than simply "Prioritize work-life balance". In addition, the generation device 1 has machine-learned that when the job requirements do not include information suggesting training, it is better not to write "Skills can be improved through comprehensive training". As a result, the generation device 1 is able to generate more appropriate and exemplary job requirements.

[0084] In this way, since the generating device 1 of this embodiment performs machine learning based on the proofread draft, the recruiting staff can improve the generation of the job posting in the generating device 1 by intuitively creating a proofread draft of the job posting and sending it to the generating device 1, as in the above explanation using the No. 1 job posting as an example. Therefore, the generating device 1 of this embodiment can reduce the burden on the recruiting staff in generating the recruitment page by machine learning the generation of the job posting via improvement information (proofread draft) provided by intuitive and familiar operations and providing the recruiting staff with a model recruitment page.

[0085] [If improvement information is evaluation] The recruiter or the like creates an evaluation of the job posting. The recruiter or the like then transmits the evaluation to the generating device 1 via a terminal T or the like. The generating device 1 receives the evaluation and performs machine learning based on the evaluation.

[0086] The following is an explanation of the creation of a proofread draft as improvement information and machine learning based on the proofread draft, using the job script identified by ID "A0002" in the example of job script database 122 shown in Figure 2 (hereinafter also referred to as "Type 2 job script") as an example.

[0087] The second job advertisement reads, "[Recruiting new nurses] Educational background and experience not required, must be able to commute to △△ city △△ town, three-shift work. Day shift 8:45-17:00 (60 minute break), semi-night shift 16:30-24:00 (45 minute break), late night shift 24:00-32:45 (90 minute break). Must be able to work at least five days a week, employment period 3-6 months or more than 6 months. 8-10 semi-night and late night shifts per month. Salary is △△ million to △△ million yen. Plenty of allowances. Why not grow in a rewarding workplace? We look forward to your application."

[0088] The recruiter who checked the No. 2 job posting script thought that since there are few young people in △△ city and therefore few new graduate nurses, and since it is more important for veteran nurses to return to work, it would be better to write it so that it does not limit the applicant to new nurses even if educational background and experience are not required. Based on this idea, the recruiter created an evaluation of the No. 2 job posting script, "Since there are few young people in △△ city and it is more important for veteran nurses to return to work, it would be better to write it so that it does not limit the applicant to new nurses even if educational background and experience are not required." as improvement information and sent it to generation device 1.

[0089] The generation device 1 received this evaluation and performed machine learning from this evaluation and the job postings and job requirements stored in the job posting database 122.

[0090] Specifically, the generation device 1 has learned by machine learning that when the draft text as the generation data includes "Those who can commute to △△ City...." in addition to "educational background and experience not required," it is better to use a headline that is not limited to new nurses, such as "[Nurses Wanted]" or "[New graduates and those with experience welcome]," instead of using the typical headline "[New nurses wanted]" when educational background and experience are not required.

[0091] In this way, since the generating device 1 of this embodiment performs machine learning based on the evaluation, the recruiter can improve the generation of the job posting in the generating device 1 by intuitively creating an evaluation of the job posting and sending it to the generating device 1, as in the above explanation using the Type 2 job posting as an example. Therefore, the generating device 1 of this embodiment can reduce the burden on the recruiter in generating the recruitment page by performing machine learning to generate the job posting via improvement information (evaluation) provided by intuitive and familiar operations and providing the recruiter with a model recruitment page.

[0092] It should be noted that within the scope of the concept of the present invention, a person skilled in the art may come up with various modifications and alterations. 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 the design of the above-described embodiment, or adds, omits, or modifies the conditions of a process, these modifications and alterations are included in the scope of the present invention as long as they include the gist of the present invention. [Explanation of symbols]

[0093] S System 1 generator 11 Control section 111 Condition Receiving Unit 112 Draft Reception Department 113 Manuscript generation section 114 Proofreading Receiving Department 115 Evaluation Reception Department 116 Machine Learning Department 12 Storage section 121 Large-scale language models 122 Job posting database 13. Communications Department N Network T-Terminal

Claims

1. A condition receiving unit for receiving job conditions; a script generation unit that provides an input including the job requirements to a large-scale language model and causes the large-scale language model to generate a first job requirements script that is a job requirements script reflecting the job requirements; A proofreading receiving unit for receiving a proofread copy, which is the first job posting copy that has been proofread; A machine learning unit that uses learning data including the job requirements and the first job requirements as explanatory variables and the proofread version as a target variable to train a large-scale language model to generate a job requirements manuscript reflecting the job requirements as input; A recruitment page generating device comprising:

2. A rating receiving unit for receiving a rating for the first job posting script, The generating device according to claim 1 , wherein the machine learning unit performs machine learning on a large-scale language model to perform the generation using training data including the first job posting and the evaluation as explanatory variables.

3. A draft receiving unit for receiving a draft of the job posting is further provided, The manuscript generation unit provides an input including the draft text to a large-scale language model, and causes the large-scale language model to generate a second job manuscript reflecting the job requirements included in the draft text. The generating device of claim 1 .

Images