Enterprise man-post matching informatization management method and system based on competency big data model

Abstract

The invention discloses an enterprise person and post matching informatization management method and system based on a competency big data model, which extract a deep association relationship between job seekers and post demands in a recruitment process through an algorithm model, and successfully solve many challenges faced by enterprises in the recruitment process. And an efficient, accurate and economic recruitment solution is provided for enterprises.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of human resource management technology, and in particular to an enterprise person-job matching information management method and system based on a competency big data model. Background Art

[0002] With the rapid development of information technology, traditional recruitment methods can no longer meet the needs of enterprises for accurate and efficient recruitment of talents. In the recruitment process, there are many challenges in accurately assessing job requirements, accurately matching talents, effectively analyzing resumes, scientifically designing questionnaires, and efficiently managing talent pools. Summary of the invention

[0003] In order to improve the above problems, the present invention proposes an enterprise person-job matching information management method based on a competency big data model, the method comprising the following steps:

[0004] Step 1: Obtain job seeker information and job requirement information. The job seeker's personal information is collected by the job seeker filling out forms online, uploading resumes, etc.; the company's job requirement information is collected through the recruitment unit, the company's human resources management system or manual input;

[0005] Step 2: extract keywords from the job seeker information, conduct preliminary screening based on the basic conditions set for the job requirements, and obtain candidate job seeker information;

[0006] Step 3: extract the personal information of the candidate job applicant and parse it to obtain at least one talent evaluation factor of the candidate job applicant, assign corresponding weights to the talent evaluation factor, and construct a talent evaluation embedding vector of the candidate job applicant;

[0007] In the process of constructing the talent evaluation embedding vector of the job seeker, the personal information title and the keywords corresponding to the personal information title are obtained through word segmentation operation, the personal information is clustered to obtain the personal information label, and the talent evaluation feature elements are constructed according to the personal information title, the keywords corresponding to the personal information title, and the personal information label; the talent feature evaluation elements are input into the pre-constructed BERT model to obtain the talent evaluation embedding vector E x ;

[0008] E x ={e,(A x1 , A x2 , ..., A xi ), (B x1 , B x2 , ..., B xi ), (C x1 , C x2 , ..., C xi )};

[0009] Among them, e represents the identity of the talent evaluation embedding vector, A xi Represents the personal information title vector, B xi represents the keyword vector corresponding to the personal information title, C xi represents the personal information label vector, i represents the number after word segmentation, which is an integer greater than 0;

[0010] Extracting the job description and job requirements of the job requirement information and parsing to obtain at least one competency element of the job requirement, assigning corresponding weights to the competency elements of the job requirement, and constructing a competency embedding vector of the job requirement information;

[0011] In the process of constructing the competency embedding vector of the job requirement information, the job requirement information is segmented to obtain the job requirement information title, the keywords corresponding to the job requirement information title, and the job requirement information tag, and the competency elements of the job requirement are constructed according to the job requirement information title, the keywords corresponding to the job requirement information title, and the job requirement information tag;

[0012] Input the competency elements of the job requirements into the pre-built BERT model to obtain the competency embedding vector E of the job requirements y ,Right now

[0013] E y ={e′,(A y1 , A y2 , ..., A yi ), (B y1 , B y2 , ..., B yi ), (C y1 , C y2 , ..., C yi )}, where e′ represents the identity of the competency embedding vector, A yi represents the job title vector, B yi represents the keyword vector corresponding to the job title, C yi Represents the label vector of job demand information, i represents the number after word segmentation, which is an integer greater than 0;

[0014] Calculate E x and E y The similarity of is used to obtain the matching degree between talent information and job requirement information;

[0015] Step 4: sort the candidate job seekers from high to low according to the matching degree between the talent information and the job requirement information, and display the matching degree between the candidate job seekers and the job requirement information in the form of a list, and obtain a preset number of candidate job seekers;

[0016] Combine the characteristics of candidates who have successfully applied for or been recruited in historical recruitment data, and use association rule mining to find the correlation between the successful matching of the talent evaluation embedding vector and the competency embedding vector of the candidates who have successfully applied for or been recruited, so as to generate a template talent evaluation embedding vector, and calculate the attribute similarity between the candidate job seeker and the template talent evaluation embedding vector;

[0017] Design personalized questionnaires for candidate job seekers and dynamically adjust the questionnaire content according to different positions and assessment purposes;

[0018] Step 5: Screen the candidate applicants based on the attribute similarity between the candidate applicants and the template talent evaluation embedding vector and the evaluation report of the personalized questionnaire.

[0019] Specifically, in step 3, the talent evaluation embedding vector E extracted by the BERT model is x and competency embedding vector E y A pooling operation is performed. During the pooling operation, the extracted feature vector is reduced in dimension through the pooling layer, and the mean of all embedded vectors is used as the embedded vector after pooling.

[0020] Specifically, the maximum value in the pooled embedding vector is selected as the pooled talent evaluation embedding vector and competency embedding vector.

[0021] Specifically, in the process of using the BERT model to calculate the matching degree, the specific loss function is:

[0022] ,

[0023] Among them, P is the talent evaluation embedding vector E x and competency embedding vector E y If there is a real relationship between the personal information and the recruitment demand information, then P = 1, otherwise P = 0, and m is a preset constant.

[0024] At the same time, this application also proposes an enterprise person-job matching information management system based on the competency big data model, which includes the following modules:

[0025] The person-job matching information acquisition module is used to obtain job seeker information and job requirement information. It collects job seekers' personal information through online forms and resume uploads; and collects the company's job requirement information through recruitment units, corporate human resources management systems or manual input.

[0026] A person-job matching information initial screening module is used to extract keywords from the job seeker information, conduct preliminary screening based on the basic conditions set by the job requirements, and obtain candidate job seeker information;

[0027] A person-job information intelligent matching module is used to extract the personal information of the candidate job applicant and parse it to obtain at least one talent evaluation factor of the candidate job applicant, assign corresponding weights to the talent evaluation factor, and construct a talent evaluation embedding vector of the candidate job applicant;

[0028] In the process of constructing the talent evaluation embedding vector of the job seeker, the personal information title and the keywords corresponding to the personal information title are obtained through word segmentation operation, the personal information is clustered to obtain the personal information label, and the talent evaluation feature elements are constructed according to the personal information title, the keywords corresponding to the personal information title, and the personal information label; the talent feature evaluation elements are input into the pre-constructed BERT model to obtain the talent evaluation embedding vector E x ;

[0029] E x ={e,(A x1 , A x2 , ..., A xi ), (B x1 , B x2 , ..., B xi ), (C x1 , C x2 , ..., C xi )};

[0030] Among them, e represents the identity of the talent evaluation embedding vector, A xi Represents the personal information title vector, B xi represents the keyword vector corresponding to the personal information title, C xi represents the personal information label vector, i represents the number after word segmentation, which is an integer greater than 0;

[0031] Extracting the job description and job requirements of the job requirement information and parsing to obtain at least one competency element of the job requirement, assigning corresponding weights to the competency elements of the job requirement, and constructing a competency embedding vector of the job requirement information;

[0032] In the process of constructing the competency embedding vector of the job requirement information, the job requirement information is segmented to obtain the job requirement information title, the keywords corresponding to the job requirement information title, and the job requirement information tag, and the competency elements of the job requirement are constructed according to the job requirement information title, the keywords corresponding to the job requirement information title, and the job requirement information tag;

[0033] Input the competency elements of the job requirements into the pre-built BERT model to obtain the competency embedding vector E of the job requirements y ,Right now

[0034] E y ={e′,(Ay1 , A y2 , ..., A yi ), (B y1 , B y2 , ..., B yi ), (C y1 , C y2 , ..., C yi )}, where e′ represents the identity of the competency embedding vector, A yi represents the job title vector, B yi represents the keyword vector corresponding to the job title, C yi Represents the label vector of job demand information, i represents the number after word segmentation, which is an integer greater than 0;

[0035] Calculate E x and E y The similarity of is used to obtain the matching degree between talent information and job requirement information;

[0036] The person-job information auxiliary matching module is used to sort the candidate job seekers from high to low according to the matching degree between the talent information and the job requirement information, and display the matching degree between the candidate job seekers and the job requirement information in the form of a list, and obtain a preset number of candidate job seekers;

[0037] Combine the characteristics of candidates who have successfully applied for or been recruited in historical recruitment data, and use association rule mining to find the correlation between the successful matching of the talent evaluation embedding vector and the competency embedding vector of the candidates who have successfully applied for or been recruited, so as to generate a template talent evaluation embedding vector, and calculate the attribute similarity between the candidate job seeker and the template talent evaluation embedding vector;

[0038] Design personalized questionnaires for candidate job seekers and dynamically adjust the questionnaire content according to different positions and assessment purposes;

[0039] The person-job matching screening module is used to screen candidate job applicants based on the attribute similarity between the candidate job applicants and the template talent evaluation embedding vector and the evaluation report of the personalized questionnaire.

[0040] Specifically, in the intelligent matching module for human-job information, the talent evaluation embedding vector E extracted by the BERT model is x and competency embedding vector E y A pooling operation is performed. During the pooling operation, the extracted feature vector is reduced in dimension through the pooling layer, and the mean of all embedded vectors is used as the embedded vector after pooling.

[0041] Specifically, in the intelligent matching module for human-job information, the maximum value is selected from the pooled embedding vectors as the pooled talent evaluation embedding vector and competency embedding vector.

[0042] Specifically, in the intelligent matching module for job information, in the process of using the BERT model to calculate the matching degree, the specific loss function is:

[0043] ,

[0044] Among them, P is the talent evaluation embedding vector E x and competency embedding vector E y If there is a real relationship between the personal information and the recruitment demand information, then P = 1, otherwise P = 0, and m is a preset constant.

[0045] The present application also proposes an electronic device, which includes a processor and a memory: the memory is used to store program code and transmit the program code to the processor, and the processor is used to execute a test case automatic generation method as described in any one of claims 1 to 4 according to instructions in the program code.

[0046] The present application also proposes a computer-readable storage medium, which is used to store program code and transmit the program code to a processor, and the processor is used to execute a test case automatic generation method as described in any one of claims 1 to 4 according to instructions in the program code.

[0047] An electronic device, characterized in that the electronic device includes a processor and a memory: the memory is used to store program code and transmit the program code to the processor, and the processor is used to execute the above-mentioned enterprise person-job matching information management method based on competency big data model according to instructions in the program code.

[0048] A computer-readable storage medium, characterized in that the computer-readable storage medium is used to store program code and transmit the program code to a processor, and the processor is used to execute the above-mentioned enterprise person-job matching information management method based on competency big data model according to the instructions in the program code.

[0049] This application uses an algorithm model to extract the deep correlation between job seekers and job requirements during the recruitment process, successfully solving many challenges faced by companies in the recruitment process and providing companies with an efficient, accurate and economical recruitment solution. In the future, with the continuous advancement of technology and continuous optimization of the system, this application is expected to be widely used in more companies, further improving the talent management level of companies. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 It is a flow chart of the enterprise personnel-job matching information management method based on the competency big data model;

[0052] Figure 2 It is a module diagram of an enterprise person-job matching information management system based on a competency big data model. DETAILED DESCRIPTION

[0053] The technical content of the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0054] Specifically, Figure 1 As shown, the present invention includes an enterprise person-job matching information management method based on a competency big data model, which specifically includes the following steps:

[0055] Step 1: Obtain job seeker information and position requirement information. The job seeker information is collected by collecting job seekers' personal information through methods such as job seekers filling out forms online and uploading resumes; the position requirement information is collected by collecting the company's position requirement information through recruitment units, corporate human resource management systems or manual input.

[0056] Step 2: extract keywords from the job seeker information, conduct preliminary screening based on basic conditions set in job requirements, and obtain candidate job seeker information.

[0057] Specifically, in this step, the personal information of job seekers is preliminarily screened according to preset basic conditions, such as education background, years of work experience, etc., and job seekers who do not meet the conditions are eliminated.

[0058] It is understandable that when job seekers screen job requirement information, they can also conduct preliminary screening based on preset basic conditions, such as salary, to obtain the required job requirement information.

[0059] Step 3: extract the personal information of the candidate job applicant and parse it to obtain at least one talent evaluation factor of the candidate job applicant, assign corresponding weights to the talent evaluation factor, and construct a talent evaluation embedding vector of the candidate job applicant;

[0060] The job description and job requirements of the job requirement information are extracted and parsed to obtain at least one competency element of the job requirement, and corresponding weights are assigned to the competency elements of the job requirement to construct a competency embedding vector of the job requirement information.

[0061] Specifically, in step 3, personal information includes unstructured text sentences such as personal introduction, work experience, project experience, personal evaluation, etc., and job requirement information includes unstructured information such as job responsibilities and job requirements. The unstructured information is matched with the preset personal information template and job requirement information template to form the template information of personal information and the template information of job requirement information.

[0062] The template information of personal information is segmented to obtain the personal information title and the keywords corresponding to the personal information title, the personal information is clustered to obtain the personal information label, and the talent evaluation characteristic elements are constructed based on the personal information title, the keywords corresponding to the personal information title, and the personal information label.

[0063] For example, a word segmentation operation is performed on the personal information of "XX user" who has job-seeking, job competition and other needs to obtain personal information titles of "XX user" such as "Personal Information", "Educational Background", "Work Experience", "Project Experience", "Self-evaluation", etc., and keywords corresponding to the personal information titles such as "male", "Tsinghua University", "Artificial Intelligence", "Software Testing", "Python Script", "Telecommunication Operating System", "High-speed Toll System", "Hardworking", "Strong Learning Ability", etc., and personal information labels such as "Ambition" and "Teamwork" are obtained through k-means clustering.

[0064] Furthermore, the talent feature evaluation factors are input into the pre-built BERT model (Bidirectional Encoder Representation from Transformers) to obtain the talent evaluation embedding vector E x ,Right now

[0065] E x ={e,(A x1 , A x2 , ..., A xi ), (B x1 , B x2 , ..., B xi ), (C x1 , C x2 , ..., C xi )}, where e represents the identity of the talent evaluation embedding vector, A xi Represents the personal information title vector, B xi represents the keyword vector corresponding to the personal information title, Cxi Represents the personal information label vector, i represents the number after word segmentation, which is an integer greater than 0.

[0066] At the same time, in this step, the template information of the position requirement information must be segmented to obtain the position requirement information title, the keywords corresponding to the position requirement information title, and the position requirement information tag, and the competency elements of the position requirement are constructed based on the position requirement information title, the keywords corresponding to the position requirement information title, and the position requirement information tag.

[0067] For example, word segmentation operation is performed on the job requirement information of "XX Company"'s recruitment and competition needs to obtain the job requirement information titles of "XX Company" such as "educational requirements", "job capabilities", "job responsibilities", "job requirements", "other requirements", etc., and the keywords corresponding to the job requirement information titles are "python architecture", "PHP", "J2EE", "server backend architecture", "project manager", "requirements analysis", "R&D testing", "salary", etc., and k-means clustering is used to obtain job requirement information labels such as "strong stress resistance", "strong sense of responsibility", "work experience", and "management ability".

[0068] At the same time, the competency elements of the job requirements are input into the pre-built BERT model to obtain the competency embedding vector E of the job requirements. y ,Right now

[0069] E y ={e′,(A y1 , A y2 , ..., A yi ), (B y1 , B y2 , ..., B yi ), (C y1 , C y2 , ..., C yi )}, where e′ represents the identity of the competency embedding vector, A yi represents the job title vector, B yi represents the keyword vector corresponding to the job title, C yi Represents the label vector of job requirement information. i represents the number after word segmentation, which is an integer greater than 0.

[0070] It can be understood that the present application combines the twin network model with the pre-built BERT model to form two identical BERT models, so that one BERT model extracts text features of the job applicant's personal information, and the other BERT model extracts text features in the job requirement information, thereby calculating the similarity between the two and combining the semantic similarity to make job or resume recommendations.

[0071] The talent evaluation embedding vector E extracted by the BERT model x and competency embedding vector E y Perform a pooling operation. During the pooling operation, the extracted feature vector is reduced in dimension through the pooling layer, and the mean of all embedded vectors is used as the embedded vector after pooling, that is:

[0072]

[0073]

[0074] Select the maximum value in the pooled embedding vector as the pooled talent evaluation embedding vector and competency embedding vector Right now:

[0075]

[0076] Furthermore, the calculation and Similarity In calculation In the process, it is necessary to calculate the embedded features of the maximum value selected after pooling.

[0077] MAX{MEAN{(A x1 , A x2 , ..., A xi}} and MAX{MEAN{(A y1 , A y2 , ..., A yi}}、

[0078] MAX{MEAN{(B x1 , B x2 , ..., B xi}} and MAX{MEAN{(B y1 , B y2 , ..., B yi}}、

[0079] MAX{MEAN{(C x1 , C x2 , ..., C xi}} and MAX{MEAN{(C y1 , C y2 , ..., C yi Multi-layer semantic similarity of}};

[0080] Specifically, in this process, the cosine similarity algorithm is used to calculate

[0081] MAX{MEAN{(A x1 , A x2, ..., A xi}} and MAX{MEAN{(A y1 , A y2 , ..., A yi}} similarity;

[0082] Using FastText algorithm to calculate MAX{MEAN{(B x1 , B x2 , ..., B xi}} and MAX{MEAN{(B y1 , B y2 , ..., B yi}} similarity;

[0083] Using the GloVe algorithm to calculate MAX{MEAN{(C x1 , C x2 , ..., C xi}} and MAX{MEAN{(C y1 , C y2 , ..., C yi}} similarity;

[0084] Combining the above three similarities to perform multi-layer semantic similarity fusion, we get and Similarity That is, the matching degree between talent information and job requirement information is obtained.

[0085] In addition, when using the BERT model to calculate the matching degree, there is still an error when comparing the recommendation list obtained based on the matching degree with the positions with actual relationships. Therefore, by optimizing the loss function, the BERT model learns to use the embedding vector similarity as the final similarity judgment, and completes the parameter adjustment for the BERT model. The specific loss function is:

[0086]

[0087] Among them, P is the talent evaluation embedding vector E x and competency embedding vector E y If there is a real relationship between personal information and recruitment demand information, then P = 1, otherwise P = 0. m is a preset constant. If there is no real relationship between personal information and recruitment demand information, that is, when P = 0, the BERT model determines that there is no match between personal information and recruitment demand information. In actual training, if S(E x ,E y )>m, indicating that the model prediction is accurate. At this time, the time complexity problem needs to be considered, and there is no need to optimize the similarity index. When P=0, if S(E x ,Ey )<m, it is necessary to adjust the parameters or weights to train the model similarity.

[0088] Step 4: sort the candidate job seekers from high to low according to the matching degree between the talent information and the job requirement information, and display the matching degree between the candidate job seekers and the job requirement information in the form of a list, and obtain a preset number of candidate job seekers.

[0089] Specifically, in this process, after obtaining the matching list of candidate job applicants, for companies and enterprises that have posted job requirements and have hiring needs, it is also necessary to combine the characteristics of candidates who have successfully competed or been recruited in historical recruitment data, and use association rule mining to discover the correlation between the successful matching of talent evaluation embedding vectors and competency embedding vectors.

[0090] Specifically, the information of personnel who have successfully applied for the corresponding position in the past is collected, and a template talent evaluation embedding vector is generated, and the candidate job applicants are further screened by calculating the similarity between the talent evaluation embedding vector of the candidate job applicant and the template talent evaluation embedding vector. Specifically, in the similarity calculation process, the mean combination is used to calculate the attribute similarity between the candidate job applicant and the template talent evaluation embedding vector.

[0091] At the same time, personalized questionnaires are designed for candidate job seekers, the questionnaire content is dynamically adjusted according to different positions and evaluation purposes, the collected personalized questionnaire data is analyzed, an evaluation report is generated, and statistical analysis and machine learning models are used to identify the candidates' potential abilities and development directions.

[0092] Step 5: Screen the candidate applicants based on the attribute similarity between the candidate applicants and the template talent evaluation embedding vector and the evaluation report of the personalized questionnaire.

[0093] It is understandable that the embodiments of the present application use a relational database (such as MySQL) and a NoSQL database (such as MongoDB) to store candidate information in a mixed manner to ensure the scalability and high availability of data.

[0094] At the same time, this application adopts a distributed storage architecture (such as Hadoop HDFS) and builds a full-text search engine based on Elasticsearch to ensure efficient storage and fast retrieval of data, so that job seekers and companies can quickly retrieve and filter recruitment and job search information. It also adopts an event stream processing framework (such as Apache Kafka) to ensure the timeliness and consistency of data updates, and through an event-driven architecture, it can achieve real-time tracking and updating of recruitment status.

[0095] Specifically, Figure 2As shown, the present invention includes an enterprise person-job matching information management system based on a competency big data model, and the system specifically includes the following modules:

[0096] The person-job matching information acquisition module is used to obtain job seeker information and job requirement information. The job seeker information is collected by the job seeker filling out forms online, uploading resumes, etc. The job requirement information is collected by the company's job requirement information through the recruitment unit, the company's human resources management system or manual input.

[0097] The person-job matching information initial screening module is used to extract keywords from the job seeker information, perform initial screening based on the basic conditions set by the job requirements, and obtain candidate job seeker information.

[0098] Specifically, in this module, the personal information of job seekers is preliminarily screened according to preset basic conditions, such as education background, years of work experience, etc., and job seekers who do not meet the conditions are eliminated.

[0099] It is understandable that when job seekers screen job demand information, they can also conduct preliminary screening based on preset basic conditions, such as salary, to obtain the required job demand information.

[0100] A person-job information intelligent matching module is used to extract the personal information of the candidate job applicant and parse it to obtain at least one talent evaluation factor of the candidate job applicant, assign corresponding weights to the talent evaluation factor, and construct a talent evaluation embedding vector of the candidate job applicant;

[0101] The job description and job requirements of the job requirement information are extracted and parsed to obtain at least one competency element of the job requirement, and corresponding weights are assigned to the competency elements of the job requirement to construct a competency embedding vector of the job requirement information.

[0102] Specifically, in this module, personal information includes unstructured text sentences such as personal introduction, work experience, project experience, personal evaluation, etc., and job requirement information includes unstructured information such as job responsibilities and job requirements. The unstructured information is matched with the preset personal information template and job requirement information template to form the template information of personal information and the template information of job requirement information.

[0103] The template information of personal information is segmented to obtain the personal information title and the keywords corresponding to the personal information title, the personal information is clustered to obtain the personal information label, and the talent evaluation characteristic elements are constructed based on the personal information title, the keywords corresponding to the personal information title, and the personal information label.

[0104] For example, a word segmentation operation is performed on the personal information of "XX user" who has job-seeking, job competition and other needs to obtain personal information titles of "XX user" such as "Personal Information", "Educational Background", "Work Experience", "Project Experience", "Self-evaluation", etc., and keywords corresponding to the personal information titles such as "male", "Tsinghua University", "Artificial Intelligence", "Software Testing", "Python Script", "Telecommunication Operating System", "High-speed Toll System", "Hardworking", "Strong Learning Ability", etc., and personal information labels such as "Ambition" and "Teamwork" are obtained through k-means clustering.

[0105] Furthermore, the talent feature evaluation factors are input into the pre-built BERT model (Bidirectional Encoder Representation from Transformers) to obtain the talent evaluation embedding vector E x ,Right now

[0106] E x ={e,(A x1 , A x2 , ..., A xi ), (B x1 , B x2 , ..., B xi ), (C x1 , C x2 , ..., C xi )}, where e represents the identity of the talent evaluation embedding vector, A xi Represents the personal information title vector, B xi represents the keyword vector corresponding to the personal information title, C xi Represents the personal information label vector, i represents the number after word segmentation, which is an integer greater than 0.

[0107] At the same time, in this module, the template information of the job requirement information must be segmented to obtain the job requirement information title, the keywords corresponding to the job requirement information title, and the job requirement information label, and the competency elements of the job requirement are constructed based on the job requirement information title, the keywords corresponding to the job requirement information title, and the job requirement information label.

[0108] For example, word segmentation operation is performed on the job requirement information of "XX Company"'s recruitment and competition needs to obtain the job requirement information titles of "XX Company" such as "educational requirements", "job capabilities", "job responsibilities", "job requirements", "other requirements", etc., and the keywords corresponding to the job requirement information titles are "python architecture", "PHP", "J2EE", "server backend architecture", "project manager", "requirements analysis", "R&D testing", "salary", etc., and k-means clustering is used to obtain job requirement information labels such as "strong stress resistance", "strong sense of responsibility", "work experience", and "management ability".

[0109] At the same time, the competency elements of the job requirements are input into the pre-built BERT model to obtain the competency embedding vector E of the job requirements. y ,Right now

[0110] E y ={e′,(A y1 , A y2 , ..., A yi ), (B y1 , B y2 , ..., B yi ), (C y1 , C y2 , ..., C yi )}, where e′ represents the identity of the competency embedding vector, A yi represents the job title vector, B yi represents the keyword vector corresponding to the job title, C yi Represents the label vector of job requirement information. i represents the number after word segmentation, which is an integer greater than 0.

[0111] It can be understood that the present application combines the twin network model with the pre-built BERT model to form two identical BERT models, so that one BERT model extracts text features of the job applicant's personal information, and the other BERT model extracts text features in the job requirement information, thereby calculating the similarity between the two and combining the semantic similarity to make job or resume recommendations.

[0112] The talent evaluation embedding vector E extracted by the BERT model x and competency embedding vector E y Perform a pooling operation. During the pooling operation, the extracted feature vector is reduced in dimension through the pooling layer, and the mean of all embedded vectors is used as the embedded vector after pooling, that is:

[0113]

[0114]

[0115] Select the maximum value in the pooled embedding vector as the pooled talent evaluation embedding vector and competency embedding vector Right now:

[0116]

[0117] Furthermore, the calculation and Similarity In calculation In the process, it is necessary to calculate the embedded features of the maximum value selected after pooling.

[0118] MAX{MEAN{(A x1 , A x2 , ..., A xi}} and MAX{MEAN{(A y1 , A y2 , ..., A yi}}、

[0119] MAX{MEAN{(B x1 , B x2 , ..., B xi}} and MAX{MEAN{(B y1 , B y2 , ..., B yi}}、

[0120] MAX{MEAN{(C x1 , C x2 , ..., C xi}} and MAX{MEAN{(C y1 , C y2 , ..., C yi Multi-layer semantic similarity of}};

[0121] Specifically, in this process, the cosine similarity algorithm is used to calculate MAX{MEAN{(A x1 , A x2 , ..., A xi}} and MAX{MEAN{(A y1 , A y2 , ..., A yi}} similarity;

[0122] Using FastText algorithm to calculate MAX{MEAN{(B x1 , Bx2, ..., B xi}} and MAX{MEAN{(B y1 , B y2 , ..., Byi}} similarity;

[0123] Using the GloVe algorithm to calculate MAX{MEAN{(C x1 , C x2 , ..., C xi}} and MAX{MEAN{(C y1 , C y2 , ..., C yi}} similarity;

[0124] Combining the above three similarities to perform multi-layer semantic similarity fusion, we get and Similarity That is, the matching degree between talent information and job requirement information is obtained.

[0125] In addition, when using the BERT model to calculate the matching degree, there is still an error when comparing the recommendation list obtained based on the matching degree with the positions with actual relationships. Therefore, by optimizing the loss function, the BERT model learns to use the embedding vector similarity as the final similarity judgment, and completes the parameter adjustment for the BERT model. The specific loss function is:

[0126]

[0127] Among them, P is the talent evaluation embedding vector E x and competency embedding vector E y If there is a real relationship between personal information and recruitment demand information, then P = 1, otherwise P = 0. m is a preset constant. If there is no real relationship between personal information and recruitment demand information, that is, when P = 0, the BERT model determines that there is no match between personal information and recruitment demand information. In actual training, if S(E x , E y )>m, indicating that the model prediction is accurate. At this time, the time complexity problem needs to be considered, and there is no need to optimize the similarity index. When P=0, if S(E x , E y )<m, it is necessary to adjust the parameters or weights to train the model similarity.

[0128] The person-job information auxiliary matching module is used to sort the talent information and job requirement information from high to low according to the matching degree, and display the matching degree between the candidate job seekers and the job requirement information in the form of a list, and obtain a preset number of alternative job seekers.

[0129] Specifically, in this module, after obtaining the matching list of candidate job applicants, for companies and enterprises that have posted job requirements and have hiring needs, it is also necessary to combine the characteristics of candidates who have successfully competed or been recruited in historical recruitment data, and use association rule mining to discover the correlation between the successful matching of talent evaluation embedding vectors and competency embedding vectors.

[0130] Specifically, the information of personnel who have successfully applied for the corresponding position in the past is collected, and a template talent evaluation embedding vector is generated, and the candidate job applicants are further screened by calculating the similarity between the talent evaluation embedding vector of the candidate job applicant and the template talent evaluation embedding vector. Specifically, in the similarity calculation process, the mean combination is used to calculate the attribute similarity between the candidate job applicant and the template talent evaluation embedding vector.

[0131] At the same time, personalized questionnaires are designed for candidate job seekers, the questionnaire content is dynamically adjusted according to different positions and evaluation purposes, the collected personalized questionnaire data is analyzed, an evaluation report is generated, and statistical analysis and machine learning models are used to identify the candidates' potential abilities and development directions.

[0132] The person-job matching screening module is used to screen candidate job applicants based on the attribute similarity between the candidate job applicants and the template talent evaluation embedding vector and the evaluation report of the personalized questionnaire.

[0133] It is understandable that the embodiments of the present application use a relational database (such as MySQL) and a NoSQL database (such as MongoDB) to store candidate information in a mixed manner to ensure the scalability and high availability of data.

[0134] At the same time, this application adopts a distributed storage architecture (such as Hadoop HDFS) and builds a full-text search engine based on Elasticsearch to ensure efficient storage and fast retrieval of data, so that job seekers and companies can quickly retrieve and filter recruitment and job search information. It also adopts an event stream processing framework (such as Apache Kafka) to ensure the timeliness and consistency of data updates, and through an event-driven architecture, it can achieve real-time tracking and updating of recruitment status.

[0135] The present invention also relates to an electronic device, characterized in that the electronic device includes a processor and a memory: the memory is used to store program code and transmit the program code to the processor, and the processor is used to execute the enterprise person-job matching information management method based on the competency big data model described in the above method embodiment according to the instructions in the program code.

[0136] The present invention also relates to a computer-readable storage medium, characterized in that the computer-readable storage medium is used to store program code and transmit the program code to a processor, and the processor is used to execute the enterprise person-job matching information management method based on the competency big data model described in the above method embodiment according to the instructions in the program code.

[0137] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

Claims

1. An enterprise person-job matching information management method based on competency big data model, characterized in that The following steps are involved: Step 1: Obtain job seeker information and job requirement information, and collect job seeker personal information by having job seekers fill out online forms and upload resumes; Collect the company's job demand information through recruiting units, corporate human resources management systems or manual input; Step 2: extract keywords from the job seeker information, conduct preliminary screening based on the basic conditions set for the job requirements, and obtain candidate job seeker information; Step 3: extract the personal information of the candidate job applicant and parse it to obtain at least one talent evaluation factor of the candidate job applicant, assign corresponding weights to the talent evaluation factor, and construct a talent evaluation embedding vector of the candidate job applicant; In the process of constructing the talent evaluation embedding vector of the job seeker, the personal information title and the keywords corresponding to the personal information title are obtained through word segmentation operation, the personal information is clustered to obtain the personal information label, and the talent evaluation feature elements are constructed according to the personal information title, the keywords corresponding to the personal information title, and the personal information label; the talent feature evaluation elements are input into the pre-constructed BERT model to obtain the talent evaluation embedding vector E x ; E x ={e,(A x1 ,A x2 ,...,A xi ),(B x1 ,B x2 ,...,B xi ),(C x1 ,C x2 ,...,C xi )}; Among them, e represents the identity of the talent evaluation embedding vector, A xi represents the personal information title vector, B xi represents the keyword vector corresponding to the personal information title, C xi represents the personal information label vector, i represents the number after word segmentation, which is an integer greater than 0; Extracting the job description and job requirements of the job requirement information and parsing to obtain at least one competency element of the job requirement, assigning corresponding weights to the competency elements of the job requirement, and constructing a competency embedding vector of the job requirement information; In the process of constructing the competency embedding vector of the job requirement information, the job requirement information is segmented to obtain the job requirement information title, the keywords corresponding to the job requirement information title, and the job requirement information tag, and the competency elements of the job requirement are constructed according to the job requirement information title, the keywords corresponding to the job requirement information title, and the job requirement information tag; Input the competency elements of the job requirements into the pre-built BERT model to obtain the competency embedding vector E of the job requirements y ,Right now E y ={e',(A y1 ,A y2 ,...,A yi ),(B y1 ,B y2 ,...,B yi ),(C y1 ,C y2 ,...,C yi )}, Where e' represents the identity of the competency embedding vector, A yi represents the job title vector, B yi represents the keyword vector corresponding to the job title, C yi Represents the label vector of job demand information, i represents the number after word segmentation, which is an integer greater than 0; Calculate E x and E y The similarity of is used to obtain the matching degree between talent information and job requirement information; Step 4: sort the candidate job seekers from high to low according to the matching degree between the talent information and the job requirement information, and display the matching degree between the candidate job seekers and the job requirement information in the form of a list, and obtain a preset number of candidate job seekers; Combine the characteristics of candidates who have successfully applied for or been recruited in historical recruitment data, and use association rule mining to find the correlation between the successful matching of the talent evaluation embedding vector and the competency embedding vector of the candidates who have successfully applied for or been recruited, so as to generate a template talent evaluation embedding vector, and calculate the attribute similarity between the candidate job seeker and the template talent evaluation embedding vector; Design personalized questionnaires for candidate job seekers and dynamically adjust the questionnaire content according to different positions and assessment purposes; Step 5: Screen the candidate applicants based on the attribute similarity between the candidate applicants and the template talent evaluation embedding vector and the evaluation report of the personalized questionnaire.

2. According to claim 1, a method for enterprise person-job matching information management based on competency big data model is characterized in that: The talent evaluation embedding vector E extracted by the BERT model x and competency embedding vector E y A pooling operation is performed. During the pooling operation, the extracted feature vector is reduced in dimension through the pooling layer, and the mean of all embedded vectors is used as the embedded vector after pooling.

3. According to claim 2, the enterprise person-job matching information management method based on the competency big data model is characterized in that: The maximum value is selected from the pooled embedding vector as the pooled talent evaluation embedding vector and competency embedding vector.

4. According to claim 1, the enterprise person-job matching information management method based on the competency big data model is characterized in that: In the process of using the BERT model to calculate the matching degree, the specific loss function is: Among them, P is the talent evaluation embedding vector E x and competency embedding vector E y If there is a real relationship between the personal information and the recruitment demand information, then P = 1, otherwise P = 0, and m is a preset constant.

5. An enterprise personnel-job matching information management system based on competency big data model, characterized by The system includes the following modules: The person-job matching information acquisition module is used to obtain job seeker information and job requirement information, and collect job seekers' personal information by having them fill out forms online and upload resumes; Collect the company's job demand information through recruiting units, corporate human resources management systems or manual input; A person-job matching information initial screening module is used to extract keywords from the job seeker information, conduct preliminary screening based on the basic conditions set by the job requirements, and obtain candidate job seeker information; A person-job information intelligent matching module is used to extract the personal information of the candidate job applicant and parse it to obtain at least one talent evaluation factor of the candidate job applicant, assign corresponding weights to the talent evaluation factor, and construct a talent evaluation embedding vector of the candidate job applicant; In the process of constructing the talent evaluation embedding vector of the job seeker, the personal information title and the keywords corresponding to the personal information title are obtained through word segmentation operation, the personal information is clustered to obtain the personal information label, and the talent evaluation feature elements are constructed according to the personal information title, the keywords corresponding to the personal information title, and the personal information label; the talent feature evaluation elements are input into the pre-constructed BERT model to obtain the talent evaluation embedding vector E x ; E x ={e,(A x1 ,A x2 ,...,A xi ),(B x1 ,B x2 ,...,B xi ),(C x1 ,C x2 ,...,C xi )}; Among them, e represents the identity of the talent evaluation embedding vector, A xi represents the personal information title vector, B xi represents the keyword vector corresponding to the personal information title, C xi represents the personal information label vector, i represents the number after word segmentation, which is an integer greater than 0; Extracting the job description and job requirements of the job requirement information and parsing to obtain at least one competency element of the job requirement, assigning corresponding weights to the competency elements of the job requirement, and constructing a competency embedding vector of the job requirement information; In the process of constructing the competency embedding vector of the job requirement information, the job requirement information is segmented to obtain the job requirement information title, the keywords corresponding to the job requirement information title, and the job requirement information tag, and the competency elements of the job requirement are constructed according to the job requirement information title, the keywords corresponding to the job requirement information title, and the job requirement information tag; Input the competency elements of the job requirements into the pre-built BERT model to obtain the competency embedding vector E of the job requirements y ,Right now E y ={e',(A y1 ,A y2 ,...,A yi ),(B y1 ,B y2 ,...,B yi ),(C y1 ,C y2 ,...,C yi )}, Where e' represents the identity of the competency embedding vector, A yi represents the job title vector, B yi represents the keyword vector corresponding to the job title, C yi Represents the label vector of job demand information, i represents the number after word segmentation, which is an integer greater than 0; Calculate E x and E y The similarity of is used to obtain the matching degree between talent information and job requirement information; The person-job information auxiliary matching module is used to sort the candidate job seekers from high to low according to the matching degree between the talent information and the job requirement information, and display the matching degree between the candidate job seekers and the job requirement information in the form of a list, and obtain a preset number of candidate job seekers; Combine the characteristics of candidates who have successfully applied for or been recruited in historical recruitment data, and use association rule mining to find the correlation between the successful matching of the talent evaluation embedding vector and the competency embedding vector of the candidates who have successfully applied for or been recruited, so as to generate a template talent evaluation embedding vector, and calculate the attribute similarity between the candidate job seeker and the template talent evaluation embedding vector; Design personalized questionnaires for candidate job seekers and dynamically adjust the questionnaire content according to different positions and assessment purposes; The person-job matching screening module is used to screen candidate job applicants based on the attribute similarity between the candidate job applicants and the template talent evaluation embedding vector and the evaluation report of the personalized questionnaire.

6. The enterprise person-job matching information management system based on the competency big data model according to claim 5 is characterized in that: In the intelligent matching module for human-job information, the talent evaluation embedding vector E extracted by the BERT model is x and competency embedding vector E y A pooling operation is performed. During the pooling operation, the extracted feature vector is reduced in dimension through the pooling layer, and the mean of all embedded vectors is used as the embedded vector after pooling.

7. The enterprise person-job matching information management system based on the competency big data model according to claim 6 is characterized in that: In the person-job information intelligent matching module, the maximum value is selected from the pooled embedding vector as the pooled talent evaluation embedding vector and competency embedding vector.

8. The enterprise person-job matching information management system based on the competency big data model according to claim 5 is characterized in that: In the intelligent matching module for job information, when using the BERT model to calculate the matching degree, the specific loss function is: 、 Among them, P is the talent evaluation embedding vector E x and competency embedding vector E y If there is a real relationship between the personal information and the recruitment demand information, then P = 1, otherwise P = 0, and m is a preset constant.

9. An electronic device, characterized in that: The electronic device includes a processor and a memory: the memory is used to store program code and transmit the program code to the processor, and the processor is used to execute the enterprise person-job matching information management method based on the competency big data model as described in any one of claims 1 to 4 according to the instructions in the program code.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program code and transmit the program code to a processor, and the processor is used to execute the enterprise person-job matching information management method based on the competency big data model as described in any one of claims 1 to 4 according to the instructions in the program code.

Images