Human resource management method and electronic device thereof

Abstract

The invention discloses a human resource management method and an electronic device thereof, the human resource management method is used for assisting talent recruitment, and the method comprises the following steps: obtaining a plurality of resume data corresponding to a job shortage by an electronic device; instructing, by the electronic device, a language model module to analyze the job shortage and the plurality of resume data to determine a sequence of the plurality of resume data relative to the job shortage; and selecting a plurality of to-be-interviewed resume data from the plurality of resume data by the electronic device according to the sequence.

Description

Technical Field

[0001] The present invention relates to a human resource management method and an electronic device thereof, and in particular to a human resource management method and an electronic device thereof which can utilize a language model module to assist in talent recruitment to screen out more appropriate candidates for job vacancies. Background Art

[0002] Human resources are the core assets of an enterprise and a key factor in its competitiveness. The purpose of human resource management is to attract, develop and retain outstanding talents in order to achieve the strategic goals of the enterprise. Common human resource recruitment methods are nothing more than matchmaking and interviews, which usually include the following steps: First, the enterprise publishes recruitment advertisements or job vacancy information through the Internet, newspapers or other channels to attract applicants to submit resumes; secondly, the enterprise screens out qualified job seekers based on their resumes, conducts telephone or video interviews with human resources or employer supervisors, finds qualified candidates and notifies them to participate in interviews; finally, the enterprise evaluates the professional ability, personality traits and cultural fit of job seekers through interviews, and decides whether to hire them. Traditional matchmaking and interviews have certain advantages, such as being able to communicate directly with job seekers, observe their reactions and performance, save time and costs, etc. However, it also has many disadvantages, such as the possibility of subjective bias, the inability to cover all potential talents, and the inability to effectively test actual work ability.

[0003] Therefore, how to effectively improve the efficiency of talent recruitment, especially to screen out suitable talents from a large number of job seekers, and avoid excessive subjective cognition affecting the selection of candidates, so as to increase the chance of recruiting the most suitable candidates, has become one of the goals that the industry is striving for. Summary of the invention

[0004] Therefore, the present invention mainly provides a human resource management method and an electronic device thereof to assist in talent recruitment.

[0005] An embodiment of the present invention discloses a human resource management method for assisting in talent recruitment, comprising obtaining a plurality of resume data corresponding to a job vacancy by an electronic device; instructing a language model module by the electronic device to analyze the job vacancy and the plurality of resume data to determine a ranking of the plurality of resume data relative to the job vacancy; and selecting a plurality of resume data to be interviewed from the plurality of resume data according to the ranking by the electronic device.

[0006] An embodiment of the present invention further discloses an electronic device for assisting in talent recruitment, comprising a processing unit; and a storage unit coupled to the processing unit, storing a program code, wherein the program code is used to instruct the processing unit to execute a human resource management method, wherein the human resource management method comprises the following steps: obtaining a plurality of resume data corresponding to a job vacancy; instructing a language model module to analyze the job vacancy and the plurality of resume data to determine a ranking of the plurality of resume data relative to the job vacancy; and selecting a plurality of resume data to be interviewed from the plurality of resume data according to the ranking. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 A schematic diagram of a human resource management process according to a first embodiment of the present invention;

[0008] Figure 2 A schematic diagram of a multi-layer instruction architecture according to an embodiment of the present invention;

[0009] Figure 3 A schematic diagram of a human resource management process according to a first embodiment of the present invention;

[0010] Figure 4 A schematic diagram of a multi-layer instruction architecture according to an embodiment of the present invention;

[0011] Figure 5 This is a schematic diagram of a human resources management system according to a first embodiment of the present invention;

[0012] Fig. 6A , Figure 6B , Figure 6C , Fig. 7A , Figure 7B , Figure 7C , Fig.7D , Fig. 8A , Figure 8B , Figure 8C , Fig.8D and Fig. 8E The figure is a schematic diagram of the operation of a human-machine operation interface according to an embodiment of the present invention.

[0013] Explanation of symbols

[0014] 10, 30: Human Resources Management Process

[0015] 100~108、300~306: Steps

[0016] 20, 40: Multi-layer instruction architecture

[0017] 200~204、400~402: Instruction stage

[0018] 5: Human Resources Management System

[0019] 50: Installation

[0020] 52: Language model module

[0021] 500: Processing unit

[0022] 502: Storage unit

[0023] 504: Program Code

[0024] 60: Human-machine interface DETAILED DESCRIPTION

[0025] In order to effectively improve the efficiency of talent recruitment, the present invention adopts a language model module, especially a chat robot, to assist in talent recruitment, screen out suitable candidates for interviews, analyze interview results, provide a basis for objective analysis, etc., so as to improve the efficiency of talent recruitment and meet manpower needs. As known in the art, a chat robot is a language model module based on a large language model, which can conduct natural language conversations with humans. Common chat robots such as ( ChatGPT, In one embodiment, the language model module may be a large language model, which refers to a language model containing hundreds of billions (or more) of parameters, which are trained on a large amount of text data. The large language model is built on the Transformer architecture, and processes a large amount of unlabeled text in an unsupervised manner, and its self-attention mechanism (Self Attention) for understanding context and the application of two-stage transfer learning (Transfer Learning). Specifically, the pre-training stage based on Transformer can use a pre-trained language model as a language model module for training, and a large number of general articles are used in advance for training (for example, pre-training is performed using multiple languages ​​of Wikipedia), and it has good generalization ability for different language texts. Generally speaking, the process of talent recruitment can be divided into two stages, the first stage is the resume screening stage, and the second stage is the interview screening stage. The present invention can use a chat robot to assist the talent recruitment process in these two stages to screen out more suitable candidates for vacancies. The following is a description of these two stages respectively.

[0026] 1. Resume screening stage.

[0027] Please refer to Figure 1 , Figure 1 1 is a schematic diagram of a human resource management process 10 according to an embodiment of the present invention. The human resource management process 10 is used to assist in talent recruitment, especially resume screening, which can be executed by an electronic device and includes the following steps:

[0028] Step 100: Start.

[0029] Step 102: Obtain multiple resume data corresponding to a job vacancy.

[0030] Step 104: Instruct a language model module to analyze the job vacancy and the plurality of resume data to determine a ranking of the plurality of resume data relative to the job vacancy.

[0031] Step 106: According to the sorting, a plurality of resume data to be interviewed are selected from the plurality of resume data.

[0032] Step 108: End.

[0033] According to the human resource management process 10, when the human resource management department of the enterprise starts the talent recruitment operation (step 100), for example, after publishing the information of a job vacancy on a website or newspapers and magazines, a large number of job seekers will submit resume data related to the job vacancy (step 102). In order to screen out suitable interviewers from the resume data, the present invention uses a language model module to analyze the job vacancy content and the resume data to determine the ranking of the resume data relative to the job vacancy (step 104), and then selects the resume data to be interviewed from the resume data according to the ranking result (step 106). In other words, in the resume screening stage, the present invention uses a language model module to analyze a large amount of job seeker resume data, and selects the resume data to be interviewed that may be suitable for the job vacancy, so as to assist the human resource management department or the employer in determining suitable job seekers.

[0034] In one embodiment, ChatGPT can be used as a language model module for analysis. For example, all job applicant resume data can be given to ChatGPT, and text analysis can be performed on the job applicant resume data through ChatGPT to summarize the background, work experience and autobiography of each job applicant, and analyze and integrate the advantages and disadvantages of the job applicant, thereby saving the time cost required for manual reading and integration. At the same time, the job vacancy requirements can be given to ChatGPT to screen whether the job applicant's background and professional ability meet a specific threshold based on the results of text analysis, and rank the job applicants according to the advantages and disadvantages of the resume (i.e., step 104), and then screen out job applicants whose background and major meet the job vacancy requirements, and compile a list of interviewees who meet the qualification conditions, that is, determine the resume data to be interviewed (step 106).

[0035] For example, the first job applicant's advantage is that he focuses on soft skills, the second job applicant's disadvantage is that he focuses on data engineering rather than data application, and the third job applicant's disadvantage is that he has little practical experience. In one embodiment, each job applicant may have multiple advantages and / or multiple disadvantages.

[0036] In another embodiment, when using the language model module to analyze job openings and resume data, user-defined conditions may also be referred to, such as points for adding or subtracting points, etc. Therefore, in step 104, the user-defined conditions may be provided to the language model module together with the job opening content and resume data to rank the resume data, so that the ranking of resume data relative to the job opening can cover the user's specific needs to meet the needs of the job opening.

[0037] In addition, in order to reduce the risk of personal or sensitive data being exposed, after obtaining multiple resume data in step 102 and before analyzing in step 104, sensitive data in the resume data may be hidden or de-identified. For example, in one embodiment, after obtaining the resume data, named-entity recognition (NER) may be performed through a natural language processing (NLP) tool to identify sensitive data in the resume data, such as ID number, mobile phone number, and address. In order to correctly perform named-entity recognition, a person with ordinary knowledge in the field should select an appropriate tool according to the system requirements; for example, for Chinese data, the natural language processing library "spacy" in Python and the Chinese model "zh_core_web_lg" may be used to perform named-entity recognition to prioritize the identification of sensitive data in the resume data, and the identification results are given NER tags. Then, the corresponding de-identification processing is performed according to the NER tag, such as masking the ID number and mobile phone number (such as A123456789 is masked to A******789), and generalizing the detailed address to only the county and city information. After the above processing, the sensitive data in the resume data can be hidden and can be used for the subsequent analysis process (i.e., step 104), thereby reducing the risk of personal or sensitive data exposure.

[0038] On the other hand, when the resume is sorted by advantages and disadvantages in step 104, an automatic checking and correction mechanism can be added to avoid the situation that the advantages of one job seeker become the advantages of another job seeker, or the advantages of one job seeker become the disadvantages of the job seeker, etc., when the language model module summarizes the advantages and disadvantages of each job seeker. In this case, after the resume data is obtained in step 102, the name, advantages and disadvantages and other information in the resume data can be extracted synchronously and stored in a first data structure; then, after the advantages and disadvantages or sorting of the resume data are analyzed by the language model module in step 104, the analysis results can be stored in a second data structure; finally, check whether the contents of advantages and disadvantages in the first data structure and the second data structure are consistent, and make corrections. For example, if the advantages or disadvantages mentioned in the resume data (in the first data structure) are included in the content of the analysis result (in the second data structure), the corresponding part in the analysis result is retained; on the contrary, if it is not included, it means that the content of the analysis result is wrong and needs to be corrected, for example, the advantages and disadvantages of the resume data can be used to supplement the advantages and disadvantages of the analysis result. In other words, if the description of strengths and weaknesses described by the language model module is incorrect, the strengths and weaknesses in the resume data will be used to supplement the summary to keep the two consistent, ensuring that the ranking results of the language model module correctly match the actual situation of the job seeker.

[0039] Furthermore, after determining the resume data to be interviewed, the embodiment of the present invention can further use a language model module to analyze the resume data to be interviewed and generate corresponding interview questions. In other words, the language model module can be used to generate more in-depth questions for each of the advantages and disadvantages of each resume data to be interviewed, which can serve as the basis for questions asked by the competent unit during the interview or a list of questions for the interviewer, helping the competent unit to have a deeper understanding of the interviewer's ability during the interview, reduce the number of unqualified candidates with a gap in ability, and achieve more efficient talent recruitment.

[0040] Therefore, the human resource management process 10 can use the language model module to analyze a large amount of job applicant resume data during the resume screening stage to select the resume data suitable for the job vacancy and further generate corresponding interview questions to assist the human resource management department or the employer in determining suitable job applicants. In addition, as mentioned above, after obtaining the resume data, it can be de-identified to reduce the risk of personal or sensitive data exposure; during analysis, user-defined conditions can be added to cover user-specific needs; and when ranking the resumes, an automatic check and correction mechanism can be added to ensure that the ranking results of the language model module correctly match the actual situation of the job applicant.

[0041] It should be noted that the human resource management process 10 is to illustrate the operation mode of using the language model module to assist in talent recruitment during the resume screening stage, and a person with ordinary knowledge in the field can adopt an appropriate method to implement it. For example, based on the human resource management process 10 and its derivative changes, a person with ordinary knowledge in the field can develop a corresponding multi-layer instruction architecture for the language model module. Please refer to Figure 2 , Figure 2 Schematic diagram of a multi-layer instruction architecture 20 according to an embodiment of the present invention. The multi-layer instruction architecture 20 is an instruction layer related to the language model module, and is used to control the language model module to implement the human resource management process 10 and its derivative changes. Figure 2 As shown, the multi-layer instruction architecture 20 includes instruction stages 200-204. When the enterprise starts the talent recruitment operation, a large amount of resume data of job seekers can be obtained, and the resume data can be de-identified (such as masking, generalization, etc.), and the instruction stage 200 is to control the language model module to perform text analysis on the input resume data to judge and output the analysis results of the advantages and disadvantages of the job seekers; at the same time, an automatic inspection and correction mechanism can be adopted to ensure that the sorting results of the language model module are correctly matched with the actual situation of the job seekers. Next, the instruction stage 202 is to control the language model module to sort the resume data or further eliminate unsuitable resume data according to the job vacancy content, custom conditions and the analysis results of the advantages and disadvantages of the job seekers, and output the resume data to be interviewed accordingly. Finally, the instruction stage 204 is to control the language model module to generate a list of questions for each interviewer according to the resume data to be interviewed and its corresponding analysis results of the advantages and disadvantages of the job seekers, and then enter the interview screening stage.

[0042] Therefore, the human resource management process 10 or the multi-layer instruction architecture 20 can determine the appropriate resume data to be interviewed and can generate corresponding interview questions to assist the human resource management department or the employer in the interview screening stage. In one embodiment, the interview questions are generated based on the aforementioned content corresponding to the strengths and / or weaknesses of each job applicant. Further, since one or more interview questions are generated corresponding to the strengths and / or weaknesses of each job applicant, it can assist the competent unit in deeply understanding the ability of the interviewee during the interview, reduce the number of unsuitable candidates with a gap in ability, and thus achieve a more efficient talent recruitment effect.

[0043] During the interview screening stage, the present invention can also utilize a language model module to assist the interview process, as described in detail below.

[0044] 2. Interview and screening stage.

[0045] Please refer to Figure 3 , Figure 3Schematic diagram of a human resource management process 30 according to an embodiment of the present invention. The human resource management process 30 is used to assist in talent recruitment, especially for the analysis of the interview process, which can be executed by an electronic device and includes the following steps:

[0046] Step 300: Start.

[0047] Step 302: Obtain interview contents of multiple job seekers corresponding to multiple resume data to be interviewed.

[0048] Step 304: Instruct the language model module to analyze the interview contents of the plurality of job seekers to determine a suitability ranking of the plurality of job seekers relative to the job vacancy.

[0049] Step 306: End.

[0050] According to the human resource management process 30, during or after the interview, the interview content of the job seeker can be obtained (step 302), and the language model module can be used to analyze the interview content of the job seeker to determine the suitability ranking of the job seeker relative to the job vacancy (step 304), so that the employer can determine the suitable candidate accordingly. In other words, the language model module can provide objective analysis of the interview content to avoid excessive subjective cognition affecting the candidate, so as to improve the chance of recruiting the most suitable candidate. Among them, according to different needs, the interview content of the job seeker may be an unprocessed audio or video file of the interview process, or it may be a processed voice recognition result, image content analysis result, etc., and is not limited to this.

[0051] For example, in one embodiment, ChatGPT can be used as a language model module for analysis. For example, all interview processes can be recorded and converted into text files, and then the translated text content is given to ChatGPT for text analysis. Common translation methods include using The Whisper API provided is used to convert speech into text, but is not limited to this. In addition, in order to ensure that the strengths and weaknesses analysis results of the interviewee described by ChatGPT are consistent with the actual resume data, you can refer to the practice in the resume screening stage and introduce an automatic checking and correction mechanism to ensure that the strengths and weaknesses analysis results described by ChatGPT are consistent with the description of the resume data.

[0052] In addition to converting the interview recordings into text for strengths and weaknesses analysis (such as Fig. 8AAs shown, the human-machine interface 60 will enter the interview audio file upload page to import the audio file of the interview process, and the resume screening application will convert the audio file into a text file through the Whisper API). In another embodiment, sentiment analysis technology can be combined to obtain intonation-related information. For example, the "TextBlob" in Python can be used to determine the emotions expressed by the interviewee during the interview, such as the positive and negative emotions. If the interviewee uses more positive words, sentiment analysis may give a higher positive sentiment score, indicating that the interviewee expresses satisfaction and pride in his or her achievements; on the contrary, if the interviewee uses more negative words when answering questions, resulting in a higher negative sentiment score, it means that the interviewee may feel frustrated or powerless when describing challenges or failures; of course, sentiment analysis may also give a sentiment score close to zero tendency, indicating that the interviewee is more neutral or objective when expressing. These emotional information can be used as a reference for the language model module to further understand the emotional response of the interviewee in different questions or scenarios, and then evaluate their adaptability, self-confidence, enthusiasm and other traits.

[0053] In addition to text analysis and sentiment analysis, in another embodiment, the interviewee's body language and / or facial expressions may be considered to more comprehensively evaluate the interviewee's characteristics. To determine the interviewee's body language or facial expressions, it is necessary to obtain the interviewee's interview video, and then use image recognition and deep learning technology to determine the behavioral patterns represented by the interviewee's body movements and facial expressions. For example, "OpenPose" in Python can be used for posture recognition to determine the joint positions of the interviewee in the interview video, and the positions of the interviewee's hands and head can be identified accordingly, and relevant information can be extracted to analyze and obtain more accurate and useful body language behavior information. For example, when detecting gestures, the joint positions of the left and right wrists can help identify whether the left and right hands are raised or lowered, and when there is a hand-raising behavior, it may reflect that the interviewee is actively participating and showing confidence and has a high interest and participation in the interview; detecting head movement, the joint positions of the nose and neck can help identify whether there is a head-raising or head-lowering movement. Raising the head may indicate that the interviewee is confident and willing to maintain eye contact with the interviewer, while lowering the head may imply that the interviewee is hesitant or unconfident about certain questions, or that the interviewee is thinking or reflecting.

[0054] Similarly, to determine the facial expressions of the interviewee in the interview video, a deep learning facial recognition model can be used. For example, using OpenCV in Python, the Haar-cascade classifier can be used to identify the location of faces in the video, and then the facial image of each face is converted into a format acceptable to the model, and the trained facial recognition model (such as 'haarcascade_frontalface_default.xml') is used to classify the facial expressions. Finally, based on the prediction results of the model, the expression category of each face is obtained, and the emotional response of the interviewee or the feeling about a specific topic can be determined. For example, when the predicted result is a disgusted expression, it may mean that the interviewee feels uncomfortable or dislikes a certain question, topic or environment, or it may imply that the interviewee is dissatisfied or antipathetic to a certain aspect or work content; if it is a happy expression, it may imply that the interviewee has a good feeling about the interview, or is interested in the content discussed in the interview.

[0055] It should be noted that the aforementioned interview audio / video files, or the text, emotion, body language and / or expression analysis results translated therefrom, etc. can be regarded as interview content, but not limited thereto, and all information that can be obtained from the interview process and is related to the job seeker can be used to implement the interview content of the embodiment of the present invention. By analyzing the interview content, the language model module can comprehensively consider multiple factors, and based on objective data analysis, determine the ranking of each interviewee relative to the job vacancy, thereby assisting the subjective observation of the employer to facilitate finding a more appropriate candidate, not only reducing the risk of recruiting newcomers and the low degree of fit between the job due to subjective judgment, but also helping recruiting newcomers and the team to cooperate more smoothly in future projects. In addition, when using the language model module to analyze the ranking of each interviewee relative to the job vacancy, user-defined conditions, such as points-added or points-reduced items, etc., can also be referred to. That is, in step 304, the user-defined conditions can be provided to the language model module together with the job vacancy content and the interview content to perform the suitable ranking of the job seeker, so that the suitable ranking of the job seeker relative to the job vacancy can cover the user's specific needs to meet the job vacancy.

[0056] The human resource management process 30 is to explain the operation mode of assisting talent recruitment through the language model module in the interview screening stage. A person with ordinary knowledge in the field can adopt an appropriate method to implement it. For example, according to the human resource management process 30 and its derivative changes, a person with ordinary knowledge in the field can develop a corresponding multi-layer instruction architecture for the language model module. Please refer to Figure 4 , Figure 4 Schematic diagram of a multi-layer instruction architecture 40 according to an embodiment of the present invention. The multi-layer instruction architecture 40 is an instruction layer related to the language model module, and is used to control the language model module to implement the human resource management process 40 and its derivative changes. Figure 4 As shown, the multi-layer instruction architecture 40 includes instruction stages 400-402. When the interview is over, the instruction stage 400 controls the language model module to perform one or more of text analysis, sentiment analysis, body movement analysis, and facial expression analysis on the interview content. If text analysis and sentiment analysis are performed, the voice content of the interview process needs to be obtained and then translated into a text file (for example, by The Whisper API provided by , is used to perform analysis to output the results of strengths and weaknesses analysis and sentiment analysis; at the same time, the text analysis results can be automatically checked and corrected to correctly match the actual situation of the job seeker. On the other hand, if it is body movement analysis and expression analysis, it is necessary to obtain the video or video content of the interview process, and then use image recognition and deep learning technology to determine the behavior pattern represented by the body movements and expressions of the interviewer. Next, the instruction stage 402 is to control the language model module to sort the interviewees according to the job vacancy content, custom conditions and text / emotion / body movement / expression analysis results, and output the interviewee potential ranking accordingly.

[0057] Therefore, the human resource management process 30 or the multi-layered command framework 40 can provide objective analysis based on the interview content to determine the suitability ranking of job seekers relative to job openings, thereby avoiding overly subjective cognition affecting the selection of candidates, thereby increasing the chances of recruiting the most suitable candidates.

[0058] In addition, it should be noted that the human resource management process 10 and the human resource management process 30 (or the multi-layer instruction architecture 20 and the multi-layer instruction architecture 40) can be independent or sequentially performed. In other words, the human resource management department or the employer can choose to use the human resource management process 10 only in the resume screening stage or the human resource management process 30 only in the interview screening stage. Of course, the human resource management process 10 and 30 can also be used sequentially in the resume and interview screening stages, which all fall within the scope of the present invention.

[0059] Furthermore, when implementing the human resource management process 10, 30 or the multi-layer instruction architecture 20, 40, a person skilled in the art should adjust the implementation means or patterns according to software and hardware conditions, application scenarios, etc. For example, please refer to Figure 5 , Figure 5Schematic diagram of a human resource management system 5 according to an embodiment of the present invention. The human resource management system 5 can be used to implement or execute the human resource management process 10, 30 or the multi-layer instruction architecture 20, 40 according to the embodiment of the present invention, and includes an electronic device 50 and a language model module 52. The language model module 52 can be a chat robot, such as ChatGPT, Bard, etc., or any large language model that can conduct natural language dialogue, etc., and is not limited thereto. The electronic device 50 can be a personal computer, a tablet computer, a notebook computer, a mobile phone, etc., which generally includes a processing unit 500 and a storage unit 502. The processing unit 500 can be a general-purpose processor, a microprocessor, an application specific integrated circuit (ASIC), etc., or a combination thereof. The storage unit 502 is coupled to the processing unit 500 and can be any data storage device for storing a program code 504, and reading and executing the program code 504 through the processing unit 500. For example, the storage unit 502 may be a read-only memory (ROM), a flash memory, a random access memory (RAM), a hard disk, an optical data storage device, and a non-volatile storage unit, but is not limited thereto. In addition, the electronic device 50 may further include a wired or wireless communication port for exchanging data with the language model module 52 in a wired or wireless manner. According to the human resource management process 10, 30, the electronic device 50 may receive a user's control command (e.g., setting job vacancy requirements, ability background requirements, etc.), receive resume data and / or interview content, and then connect and exchange data with the language model module 52 and output the analysis results of the language model module 52, etc. Such operations to implement the human resource management process 10, 30 and their derivative changes should be well known in the art, so they will not be repeated.

[0060] In one embodiment, the program code 504 may be a resume screening application, which is used to access various services, and the services may include the language model module 52 and / or the Whisper API, etc., to perform text analysis and text translation.

[0061] In one embodiment, the electronic device 50 can also communicate with an email server, and the resume screening application can automatically send the potential ranking to the supervisor's email. Specifically, the resume screening application can input the strengths and weaknesses analysis results, sentiment analysis results and job vacancy conditions into the language model module 52, instruct the language model module 52 to sort the potential rankings of all interviewees, and connect the email API in series to allow the language model module 52 to generate the interviewer potential ranking results, and then automatically send the potential ranking to the supervisor's email to provide the supervisor with an objective interview evaluation. Furthermore, in one embodiment, the resume screening application can also automatically send an invitation letter to the job seeker to be interviewed, or send a thank-you letter to the interviewee who was not accepted.

[0062] Therefore, through the human resource management system 5, the language model module 52 can provide more reasonable and objective information about job seekers to the human resource management department or the interview supervisor to assist in recruiting the most suitable candidates. The human resource management system 5 is used to represent the necessary elements required to implement the embodiment of the present invention. A person with ordinary knowledge in the field can make different modifications and adjustments accordingly, but is not limited to this. For example, if the electronic device 50 is implemented as a mobile phone, the resume screening application (or program code 504) can use the touch screen of the mobile phone to display the home page of the human-machine operation interface 60, such as Fig. 6A As shown, the user can operate the human-machine operation interface 60 to screen suitable candidates for interviews, and can analyze the interview results to provide objective analysis basis, etc., so as to improve the efficiency of talent recruitment and meet manpower needs.

[0063] Specifically, if Fig. 6A As shown, the home page of the human-machine interface 60 includes options such as "Job Vacancy Description Content", "Enter Resume Data", "Enter Interview Audio File", "Screen Out Unsuitable Candidates", "Provide Interview Questions to Supervisors", and "Sort All Interviews". When the user clicks on the "Job Vacancy Description Content" option, the human-machine interface 60 will switch to the page to set the job vacancy requirements, such as Figure 7B In this example, assuming that the job title is data scientist, the user can fill in the job title and capability background requirements in the form in order and complete the storage change, then the human-machine operation interface 60 returns to Fig. 6A Next, the user can click the "Input History Data" option, and the human-machine operation interface 60 will jump to Figure 6C The user can upload the resume data of all applicants through the resume screening application. At this stage, the resume screening application has obtained the job vacancy content and the resume data of the job seeker, so the language model module 52 can be connected in series to perform text analysis.

[0064] After the data is uploaded and the language model module 52 completes the text analysis, the user can click on the "screen out unqualified candidates" and "provide supervisor interview questions" options on the home page of the human-machine interface 60. Fig. 7A As shown in FIG. 1 , when the user clicks the “Filter out inappropriate candidates” option, the human-machine interface 60 enters the filter out inappropriate candidates page and further displays the “List”, “Analysis”, “Topic” and “Ranking” options, and the page corresponding to the “List” is displayed by default. Fig. 7A In the page, the user can click to view the data of a specific job seeker, and the "Invite" option is displayed next to some job seeker lists, indicating that they are the resume data of candidates who are more suitable for the job vacancy after analysis by the language model module 52. This option can be used to directly send a letter to invite the job seeker to the subsequent interview process. Then, the user can click Fig. 7A The "Analysis" option on the page is as follows Figure 7B As shown, the human-machine interface 60 will enter the page displaying the analysis results, that is, reviewing the strengths and weaknesses of the job seeker. Fig. 7A , Figure 7B The "Topic" option on the page or Fig. 6A If you select the "Provide supervisor interview questions" option, Figure 7C As shown, the human-machine interface 60 will enter the page that provides interview questions for the supervisor, that is, check the list of recommended questions for each expected interviewee by the language model module 52. Figure 7B and Figure 7C It can be seen that for potential job seekers, the language model module 52 can analyze the strengths and weaknesses of the job seekers, and then organize interview questions based on the strengths and weaknesses. For example, "Mr. Chen has a background in mathematics and statistics, has strong data modeling capabilities and machine learning experience, and also has strong skills in programming, but relatively lacks industry background and application scenario practice." Therefore, the interview questions for "Mr. Chen" include "Strengths: machine learning experience, problem-solving ability, teamwork; Disadvantages: lack of industry experience, programming skills, self-learning." Accordingly, the interview questions are generated based on the corresponding content of each job seeker's strengths and weaknesses, which can help the competent unit to deeply understand the interviewee's ability during the interview, reduce the number of unsuitable candidates with ability gaps, and achieve a more efficient talent recruitment effect.

[0065] In addition, users can click Figure 7A to Figure 7C The "Ranking" option on the page is as follows Fig.7D As shown, the human-machine operation interface 60 will enter the page for sorting the ranking of the resume data. Fig.7D In the page, the user can view the potential ranking provided by the language model module 52, so that the department manager can conduct subsequent interview processes based on the analysis results.

[0066] After the interview, the interviewer or user can click Fig. 6A If you select the "Input Interview Audio File" option, Fig. 8A As shown, the human-machine interface 60 will enter the interview audio file upload page to import the audio file of the interview process. After uploading the interview audio file, the resume screening application will convert the audio file into a text file through the Whisper API and provide it as input data to the language model module 52, so that the language model module 52 can analyze the advantages and disadvantages of the interview answer content. Then, the user can click Fig. 6A If you select the "Sort all interviews" option, Figure 8B As shown, the human-machine interface 60 enters the interview pass summary page to display the interviewer rankings, and further displays the "list", "analysis", "ranking" and "notification" options, and the default display is the page corresponding to "ranking". The user can click Figure 8B Click the "List" option in the interview to view the list of candidates who have passed the interview, such as Figure 8C or click on the "Analysis" option to view the strengths and weaknesses of the interviewee, such as Fig.8D Finally, the user can click FIG. 8B to FIG. 8D If you select the Notification option in Fig. 8E As shown, the human-machine interface 60 will jump to the email sending page to notify the interview supervisor of the recommended candidates. If the send email is further clicked, the resume screening application will send the interviewee selection list, interview audio file analysis results, interview ranking and other information to the interview supervisor's mailbox via email, providing objective result analysis to assist the unit supervisor in screening out the most suitable candidate for the job vacancy.

[0067] Therefore, the user can operate the resume screening application through the human-computer operation interface 60 to utilize the language model module 52 to assist in determining suitable resume data to be interviewed and generate corresponding interview questions. At the same time, objective analysis can be provided based on the interview content to determine the suitability ranking of job seekers relative to job vacancies, thereby avoiding overly subjective cognition affecting the selection of candidates, and thereby increasing the chances of recruiting the most suitable candidate.

[0068] It should be noted that FIG. 6A to FIG. 8E The figure illustrates a human-machine interface 60 suitable for operating a resume screening application, but is not limited thereto. A person skilled in the art may appropriately adjust, add or reduce other operating interfaces, buttons, options, etc., or appropriately modify the page content, style, etc. according to system requirements.

[0069] In the prior art, the traditional matchmaking and interview process may have subjective bias, fail to cover all potential talents, fail to effectively test actual work ability, etc. In contrast, the present invention uses the language model module to analyze the strengths and weaknesses of each job seeker, selects job seekers who are more suitable for the job vacancy to enter the interview, and sorts out questions for each candidate who enters the interview stage according to the resume content, so that the employer can ask more core questions in the interview, gain a more comprehensive understanding, and better understand the fit between the interviewer and the job vacancy, thereby reducing the time and energy spent by the HR unit and the employer on pre-work, and can also help integrate the strengths and weaknesses of job seekers and screen more suitable candidates. In addition, the present invention can use the language model module to perform text analysis of advantages and disadvantages, positive and negative emotion analysis, body language information and expression information, etc. on the conversation content of the interview process to analyze the advantages and disadvantages of the interviewee, and comprehensively sort out the potential ranking by all personality, background experience, ability performance and other information. At the same time, a comprehensive analysis and ranking report can be sent to the employer as a basis for objective analysis, and finally combined with the employer's subjective observation to find a more suitable candidate. This can reduce the risk of low fit between recruiting newcomers and positions due to subjective judgment, and also help recruiting newcomers and the team to cooperate more smoothly in future projects.

[0070] In summary, the present invention can utilize the language model module to assist in talent recruitment, so as to screen out more suitable candidates for job vacancies.

[0071] The above descriptions are only preferred embodiments of the present invention. All equivalent changes and modifications made according to the claims of the present invention should fall within the scope of the present invention.

Claims

1. A human resource management method used to assist in talent recruitment, including: Acquiring a plurality of resume data corresponding to the job vacancy from an electronic device; Instructing the language model module by the electronic device to analyze the job vacancy and the plurality of resume data to determine the ranking of the plurality of resume data relative to the job vacancy; and The electronic device selects a plurality of resume data to be interviewed from the plurality of resume data according to the sorting.

2. The human resource management method as described in claim 1, wherein the step of instructing the language model module by the electronic device to analyze the job vacancy and the multiple resume data to determine the ranking of the multiple resume data relative to the job vacancy is a step in which the electronic device instructs the language model module to utilize text analysis to integrate the advantages and disadvantages of the multiple resume data relative to the job vacancy to determine the ranking of the multiple resume data relative to the job vacancy.

3. The human resource management method as claimed in claim 1 further comprises checking by the electronic device whether the sorting is consistent with the contents of the plurality of resume data, and adjusting the sorting when the sorting is not consistent with the contents of the plurality of resume data.

4. The human resource management method as described in claim 1, wherein the step of selecting the multiple resume data to be interviewed from the multiple resume data according to the ranking by the electronic device is that the electronic device instructs the language model module to select the resume data that meets the specific threshold corresponding to the job vacancy from the ranking as the multiple resume data to be interviewed. 5 . The human resource management method as claimed in claim 1 , further comprising hiding or de-identifying sensitive data in the plurality of resume data.

6. The human resource management method as described in claim 1 further comprises the electronic device instructing the language model module to analyze the multiple resume data to be interviewed and generate multiple interview questions corresponding to the multiple resume data to be interviewed respectively.

7. The human resource management method according to claim 1, further comprising: Obtaining, by the electronic device, interview contents of a plurality of job seekers corresponding to the plurality of resume data to be interviewed; and The electronic device instructs the language model module to analyze the interview contents of the multiple job seekers to determine the suitability ranking of the multiple job seekers relative to the job vacancy.

8. The human resource management method as claimed in claim 7, wherein the interview contents of the plurality of job seekers are selected from one or more of audio files, video files, audio files translated into text files, and image recognition results of the interview processes of the plurality of job seekers.

9. The human resource management method as described in claim 7, wherein the step of instructing the language model module by the electronic device to analyze the interview contents of the multiple job applicants is a step of instructing the language model module by the electronic device to analyze one or more of the textual content, emotional content, body movements and expressions of the interview process of the multiple job applicants.

10. The human resource management method as claimed in claim 7, further comprising sending information related to the ranking or the eligibility ranking by the electronic device through a mail server.

11. An electronic device for assisting talent recruitment, comprising: Processing unit; as well as A storage unit is coupled to the processing unit and stores a program code, wherein the program code is used to instruct the processing unit to execute a human resource management method, wherein the human resource management method comprises the following steps: Obtaining multiple resume data corresponding to the job opening; Instructing the language model module to analyze the job opening and the plurality of resume data to determine a ranking of the plurality of resume data relative to the job opening; as well as According to the sorting, a plurality of resume data to be interviewed are selected from the plurality of resume data.

12. A human resource management system as described in claim 11, wherein the step of instructing the language model module to analyze the job vacancy and the multiple resume data to determine the ranking of the multiple resume data relative to the job vacancy is to instruct the language model module to use text analysis to integrate the advantages and disadvantages of the multiple resume data relative to the job vacancy to determine the ranking of the multiple resume data relative to the job vacancy. 13 . The human resource management system as claimed in claim 11 , wherein the human resource management method further comprises checking whether the sorting is consistent with the contents of the plurality of historical data, and adjusting the sorting when the sorting is not consistent with the contents of the plurality of historical data.

14. A human resource management system as described in claim 11, wherein the step of selecting the multiple resume data to be interviewed from the multiple resume data according to the sorting is to instruct the language model module to select the resume data that meets the specific threshold corresponding to the job vacancy from the sorting as the multiple resume data to be interviewed.

15. The human resource management system as claimed in claim 11, wherein the human resource management method further comprises hiding or de-identifying sensitive data in the plurality of resume data.

16. The human resource management system as claimed in claim 11, wherein the human resource management method further comprises instructing the language model module to analyze the plurality of resume data to be interviewed and respectively generate a plurality of interview questions corresponding to the plurality of resume data to be interviewed.

17. The human resource management system as claimed in claim 11, wherein the human resource management method further comprises: Obtaining interview contents of a plurality of job seekers corresponding to the plurality of resume data to be interviewed; and The language model module is instructed to analyze the interview contents of the plurality of job seekers to determine the suitability ranking of the plurality of job seekers relative to the job vacancy.

18. The human resource management system as claimed in claim 17, wherein the interview contents of the plurality of job seekers are selected from one or more of audio files, video files, audio files translated into text files, and image recognition results of the interview processes of the plurality of job seekers.

19. A human resource management system as described in claim 17, wherein the step of instructing the language model module to analyze the interview content of the multiple job applicants is to instruct the language model module to analyze one or more of the text content, emotional content, body movements and expressions of the interview process of the multiple job applicants.

20. The human resource management system as claimed in claim 17, wherein the human resource management method further comprises sending information related to the ranking or the eligibility ranking through a mail server.

Images