Intelligent recruitment system and method based on human-computer interaction and resume mining

Abstract

The invention discloses an intelligent recruitment system and method based on human-computer interaction and resume mining, and the method comprises the steps: giving a recruitment position demand by an HR, analyzing the position demand by a robot, and generating an HR demand label; the robot searches a resume on the network, analyzes the resume and extracts a corresponding resume label; resumes are screened in combination with HR demand labels, and the resumes are sorted; the entrusting robot and the job seeker of the resume chat online, verifying the authenticity of the label of the resume, and mining the temporary label of the job seeker; evaluating the importance of the temporary label, and constructing a job seeker evaluation model according to the HR demand label, the resume label and the temporary label; the robot recommends the job seekers meeting the HR requirements to the HR, and an evaluation report is generated based on a job seeker evaluation model; through resume screening, labeling processing, position knowledge graph construction and intelligent dialogue verification, the efficiency and accuracy of the recruitment process are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information and communication technology specifically suitable for administrative, commercial, financial, management or supervisory purposes, and in particular to the technical field of administration, specifically to an intelligent recruitment system and method based on human-computer interaction and resume mining. Background Art

[0002] As enterprises expand in size and their businesses become more complex, recruitment needs are increasing, and job requirements are becoming more detailed. The popularity of the Internet and social media has made personal information more widely available, and resume data is growing exponentially. Enterprises are increasingly relying on automated and intelligent systems to handle large amounts of recruitment work. Intelligent recruitment systems help companies quickly screen out suitable candidates through resume analysis, online assessments, and preliminary interviews. The rapid development of artificial intelligence technologies such as natural language processing (NLP), machine learning, and knowledge graphs has provided strong technical support for the development of intelligent recruitment systems. These technologies make it possible to automatically extract key information from text data, build knowledge graphs, and make intelligent recommendations.

[0003] Most of the intelligent recruitment systems in the existing technology remain at the simple keyword matching level and are unable to deeply understand and analyze the content of resumes, resulting in inaccurate screening results; the system lacks effective means to verify the authenticity of resumes, and it is difficult to judge the true ability and experience of job seekers based solely on the text descriptions on the resumes, which increases the uncertainty and risk in the recruitment process; most of the existing systems can only screen and sort resumes according to preset rules, and are unable to make accurate recommendations based on the specific needs of HR and the personalized characteristics of job seekers.

[0004] Therefore, it is necessary to improve the intelligent recruitment system and method based on human-computer interaction and resume mining in the prior art to solve the above problems. Summary of the invention

[0005] The present invention overcomes the shortcomings of the prior art and provides an intelligent recruitment method based on human-computer interaction and resume mining, aiming to solve the problems of insufficient depth of resume content understanding, lack of resume authenticity verification and insufficient personalized recommendation capabilities in the prior art.

[0006] To achieve the above object, the technical solution adopted by the present invention is: an intelligent recruitment method based on human-computer interaction and resume mining, comprising:

[0007] S1. HR gives the job requirements, the robot analyzes the job requirements and generates HR requirement tags;

[0008] S2, the robot searches for resumes on the Internet, parses the resumes, and extracts the corresponding resume tags;

[0009] S3. Filter resumes based on HR requirement tags and sort them;

[0010] S4, entrust the robot to have online chats with job seekers with resumes, verify the authenticity of resume labels, and mine temporary labels of job seekers;

[0011] S5. Evaluate the importance of temporary tags and build a job applicant evaluation model based on HR demand tags, resume tags, and temporary tags;

[0012] S6. The robot recommends job seekers who meet HR requirements to HR and generates an evaluation report based on the job seeker evaluation model.

[0013] In a preferred embodiment of the present invention, in step S2, natural language processing technology is used to split the resume text, divide the continuous text into separate words or phrases, identify the resume key information, and calculate the information importance in the resume key information: TF-IDF(t,d)=TF(t,d)×IDF(t), Among them, TF represents the frequency of the key information in the resume, IDF represents the inverse of the prevalence of the key information in all resumes, N is the total number of resumes, n t It represents the number of documents containing resume key information t. The higher the TF-IDF value, the more the resume key information can represent the uniqueness of the resume. The top 10-20 resume key information are selected based on the TF-IDF value.

[0014] In a preferred embodiment of the present invention, in step S3,

[0015] S31, traverse the resumes and calculate the number of matches between resume tags and HR requirement tags;

[0016] S32. When the ratio of resume tags to HR requirement tags exceeds a certain value, the resumes are put into a list for sorting;

[0017] S33, calculating the number of matches between the resume tag and the HR requirement tag, and sorting the resumes in descending order according to the number of matches;

[0018] S34. When the number of matches between resume tags of multiple qualified resumes and HR requirement tags is the same, the TF-IDF values ​​of the resume tags are weighted and summed to obtain a comprehensive score S TF-IDF , according to the comprehensive score S TF-IDF The resumes with the same number of matches between resume tags and HR requirement tags are sorted by the high and low values.

[0019] In a preferred embodiment of the present invention, in step S4, the robot also constructs a job knowledge graph based on job requirements;

[0020] Construct a job knowledge graph G = (V, E), where V is the set of all nodes and E is the set of edges between nodes; for each node v∈V, define the attribute set D v ;

[0021] To calculate the importance of a node, let f be the frequency of occurrence of the node in the job requirements, a be the importance of the node to the success of the job, obtained by expert evaluation or historical data analysis, S 节点重要性评分 =ω f ×f+ω a ×a, where ω f ,ω a The weights for frequency and importance;

[0022] Calculate the relationship strength of the nodes, let c be the co-occurrence frequency of the two nodes in the job requirements, sim be the semantic similarity between the two nodes, S 关系强度评分 =ω c ×c+ω sim ×sim, where ω c ,ω sim is the weight of co-occurrence frequency and semantic similarity;

[0023] Calculate the attribute quantification, let d be the specific attribute value, min and max be the minimum and maximum values ​​of the attribute in all job requirements;

[0024] In a preferred embodiment of the present invention, in step S4, the step of verifying the resume tag includes:

[0025] S41, align the job seeker resume label with the job knowledge graph, and find the corresponding node in the job knowledge graph; use sim(T i ,L j ) Similarity score aligns and matches the resume tag with the corresponding HR requirement tag;

[0026] S42. Design an interview question database for each node. The question types include: open-ended questions, closed-ended questions, and situational questions; verify whether the job applicant actually possesses the skills and experience listed on the resume;

[0027] S43, classifying resume tags into skill tags, experience tags, and personality tags; the robot extracts questions corresponding to resume tags from the interview question database, and determines question templates based on the characteristics of the tags, and uses the template filling algorithm to dynamically generate questions based on the specific tags on the resume;

[0028] S44. When the job seeker can answer the question accurately and in detail, the label is considered to have been verified; the node status in the knowledge graph is updated to mark the HR requirement label corresponding to the resume label as verified.

[0029] In a preferred embodiment of the present invention, in step S4, the step of mining temporary tags of job seekers includes:

[0030] S45, the robot checks the job-related tags in the knowledge graph, confirms that the HR requirement tags are not explicitly mentioned or verified in the job seeker's resume, and marks them as missing tags;

[0031] S46. Evaluate the importance of missing labels based on job requirements and the relationship network in the knowledge graph, and determine their suitability and impact on the job;

[0032] S47. Based on the type and importance of the missing label, the robot selects questions corresponding to the missing label from the interview question database and guides the job seeker to discuss the experience, project or ability related to the missing label;

[0033] S48. Mining the information of job seekers obtained from question S47, the robot constructs temporary labels, associates the temporary labels with corresponding nodes in the knowledge graph, and corrects the knowledge graph.

[0034] In a preferred embodiment of the present invention, in step S46,

[0035] The importance of missing labels, combined with the node importance calculation process in the job knowledge graph, assuming that the missing label is m, the corresponding importance is S 缺失标签重要性评分 =ω f ×f+ω m ×m, where ω f ,ω m are the weights for the frequency of occurrence and the importance of missing labels;

[0036] The fitness of missing labels is evaluated by calculating the average semantic similarity between missing labels and verified labels. The set of verified HR demand labels g is G. sim(m,g) represents the average semantic similarity between missing labels and verified HR requirement labels;

[0037] The impact of missing tags is calculated by combining the importance of missing tags and the relationship strength of other verified tags with them, and the relationship strength of S4 is calculated, where d is the co-occurrence frequency of missing tags to verified HR demand tags, ω d ,ω sim(m,g) is the weight corresponding to the co-occurrence frequency and the average semantic similarity, S 影响评分 =S 缺失标签重要性评分 ·(1+S 关系强度评分 )=(ω f ×f+ω m ×m)·(1+(ω d ×d+ω sim(m,g)×sim(m,g)));

[0038] Combining the above scores, we can get a comprehensive evaluation score S of the missing label's suitability and impact on the position. 缺失标签 =α·S 缺失标签重要性评分 +β·S 缺失标签适合度评分 +γ·S 影响评分 , where α, β, γ are the corresponding weight coefficients.

[0039] In a preferred embodiment of the present invention, the step of modifying the knowledge graph includes:

[0040] S481. When the temporary label is highly correlated with an existing node in the knowledge graph, the temporary label is added as a child node of the node;

[0041] S482, when the temporary tag represents a low relevance to an existing node in the knowledge graph, creating a supplementary node in the knowledge graph, and connecting the supplementary node to the relevant node;

[0042] S483, updating the relationship strength between the temporary label and other nodes in the knowledge graph;

[0043] Let the set of temporary labels u be U, and the semantic similarity between them and nodes v∈V Among them, u k ,v k is the vector representation of temporary labels and knowledge graph nodes; update the relationship strength, S 更新关系强度 =μ·sim(u,v)+(1-μ)·S 关系强度评分 , μ is the weight of the semantic similarity of the temporary label; update importance, S 更新重要性 =μ·sim(u,v)+(1-μ)·S 节点重要性评分 .

[0044] In a preferred embodiment of the present invention, in step S5, the updated relationship strength, updated importance and attribute quantitative score of the job knowledge graph are used to recalculate the comprehensive score of the resume tag and the score of the temporary tag, and the overall evaluation score of the job seeker is generated in combination;

[0045] The score of the resume tag is expressed as

[0046] The score of the temporary label is expressed as

[0047] The overall evaluation score of the job seeker is S = ω 1 S 1 +ω 2 S 2 ,ω 1 +ω 2 =1;ω1 ,ω 2 are the weights of resume tag and temporary tag for HR requirement tag, where ω 0 is the baseline weight of the temporary label, U| is the number of sets of corresponding temporary labels u.

[0048] The present invention provides an intelligent recruitment system based on human-computer interaction and resume mining, comprising:

[0049] The job requirement parsing module is used to receive job requirement information provided by HR, parse and extract key information through natural language processing technology, and convert it into HR requirement tags;

[0050] Resume collection module, used to collect resume data from the Internet, parse and clean it;

[0051] Resume label generation module, used to extract key information from resumes and generate resume labels;

[0052] Resume screening module, used to screen and sort resumes based on the matching of HR requirement tags and resume tags;

[0053] Position knowledge graph construction module, used to construct position-related knowledge graphs;

[0054] Interview question database module, which stores and manages interview questions for different job requirements and tags;

[0055] Robot chat module, the robot extracts questions from the interview question database and chats with job seekers online;

[0056] Label verification module, the robot verifies the authenticity of resume labels;

[0057] Tag mining module: the robot mines temporary tags of job seekers;

[0058] The knowledge graph correction module is used to correct the job knowledge graph based on the information obtained in the online chat;

[0059] The job applicant evaluation model building module is used to build a job applicant evaluation model by combining the job knowledge graph, resume tags and temporary tags, and generate an overall evaluation score;

[0060] The job applicant recommendation module recommends job applicants who meet HR needs based on the job applicant evaluation model and generates an evaluation report.

[0061] The present invention solves the defects existing in the background technology and has the following beneficial effects:

[0062] (1) The present invention improves the efficiency and accuracy of the recruitment process through resume screening, labeling, job knowledge graph construction and intelligent dialogue verification. It can not only quickly locate candidates who meet the job requirements, but also explore the potential abilities and true levels of candidates through in-depth communication, thereby helping the human resources department to make more informed recruitment decisions. At the same time, it provides personalized feedback to candidates, enhancing the transparency and fairness of the recruitment process.

[0063] (2) The present invention constructs a job knowledge graph by combining the construction of a job knowledge graph with the verification of resume labels, defines nodes, edges and attributes, and calculates node importance and relationship strength, thereby constructing a comprehensive job knowledge graph, which provides a basis for subsequent verification of resume labels and mining of temporary labels; aligns job seeker resume labels with the job knowledge graph, and designs an interview question database to verify whether job seekers actually possess the skills and experience listed on their resumes; enables robots to have a more comprehensive understanding of job requirements and verify job seekers' abilities and experience through interview questions, thereby improving the accuracy and reliability of recruitment; robots can use the knowledge graph to find missing labels, mine job seekers' information that is not mentioned in their resumes, establish temporary labels, and correct the knowledge graph, thereby realizing the update and iteration of the job knowledge graph.

[0064] (3) The present invention utilizes the updated relationship strength, updated importance, and attribute quantitative scoring of the job knowledge graph to recalculate the scores of resume tags and temporary tags, generate an overall evaluation score for job seekers, comprehensively consider the information of resume tags and temporary tags, and conduct a comprehensive evaluation of job seekers, thereby improving the accuracy and comprehensiveness of the job seeker evaluation.

[0065] (4) In the present invention, HR provides job requirements, and the robot parses the text through natural language processing technology, extracts key information, and converts it into HR requirement tags, thereby refining the specific requirements of the recruitment position into a label form that is easy to understand and operate, improving the depth of understanding, helping to clarify HR's expectations and requirements for candidates, providing a clear direction for subsequent steps, and improving the accuracy and efficiency of demand understanding. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] 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 or the description of the prior art are briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work.

[0067] Figure 1 is a flow chart of a preferred embodiment of the present invention;

[0068] Figure 2 It is a system structure diagram of a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0069] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0070] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0071] An intelligent recruitment method based on human-computer interaction and resume mining, comprising the steps of:

[0072] S1. HR gives the job requirements, the robot analyzes the job requirements and generates HR requirement tags;

[0073] S2, the robot searches for resumes on the Internet, parses the resumes, and extracts the corresponding resume tags;

[0074] S3. Filter resumes based on HR requirement tags and sort them;

[0075] S4, entrust the robot to have online chats with job seekers with resumes, verify the authenticity of resume labels, and mine temporary labels of job seekers;

[0076] S5. Evaluate the importance of temporary tags and build a job applicant evaluation model based on HR demand tags, resume tags, and temporary tags;

[0077] S6. The robot recommends job seekers who meet HR requirements to HR and generates an evaluation report based on the job seeker evaluation model.

[0078] In step S1, the job requirements refer to the specific requirements of HR for the required job positions, which can be given in various forms, including text, language and video. The robot analyzes the job requirements text through natural language processing technology, understands the job requirements analysis text, extracts key information, and converts the key information into HR requirement tags.

[0079] HR demand tags are a refinement and summary of job requirements, representing HR's expectations and requirements for candidates. HR demand tags are classified into:

[0080] Job skills: industry knowledge, software skills;

[0081] Work experience: years of work experience, project experience;

[0082] Educational background: academic qualifications, major, language;

[0083] Personality traits: communication skills, teamwork;

[0084] Other tags: job location, salary range;

[0085] HR demand tags refine the specific requirements of recruitment positions into easy-to-understand and easy-to-operate tags, which helps to clarify HR's expectations and requirements for candidates.

[0086] In step S2, the robot determines the resume sources including: major recruitment websites, social media, company official websites and industry forums; using web crawler technology, the robot automatically visits these websites to retrieve and collect resume data; after obtaining the resumes, the robot cleans the data and removes duplicate resumes.

[0087] In step S2, natural language processing technology is used to perform text segmentation on the resume, to segment the continuous text into separate words or phrases, to identify key information of the resume, and to convert the identified key information of the resume into structured data;

[0088] The TF-IDF formula is used to calculate the importance of information in the key information of the resume: TF-IDF(t,d)=TF(t,d)×IDF(t), Among them, TF represents the frequency of the key information in the resume, IDF represents the inverse of the prevalence of the key information in all resumes, N is the total number of resumes, n t It represents the number of documents containing resume key information t. The higher the TF-IDF value, the more the resume key information can represent the uniqueness of the resume. The top 10-20 resume key information are selected based on the TF-IDF value.

[0089] Use named entity recognition technology to identify entities in resumes, including company names, school names, and skill names; generate primary labels based on the key information of the resume and the entities; when multiple preliminary labels express the same or similar meanings, merge them into one primary label to avoid duplication.

[0090] The preliminary tags are judged to see whether they meet the HR requirement tags through semantic similarity, category matching or keyword matching. The preliminary tags that meet the requirements are extracted as resume tags.

[0091] Semantic similarity: The preliminary label is very close to the HR requirement label in meaning. Specifically, "Rich knowledge in the XX industry field" is very close to the HR requirement label "industry knowledge", so it is considered that "Rich knowledge in the XX industry field" can be used as a resume label.

[0092] Category matching: The preliminary tag belongs to a category in the HR requirement tag; specifically, "5 years of work experience" matches the HR requirement tag "work experience", so "5 years of work experience" is considered to be a resume tag;

[0093] Keyword matching: Check whether the preliminary tags contain the keywords in the HR requirement tags; specifically, "good communication skills" matches the keyword "communication" in the HR requirement tag "communication ability", so "good communication skills" can be used as a resume tag;

[0094] In step S3,

[0095] S31, traverse the resumes and calculate the number of matches between resume tags and HR requirement tags;

[0096] S32. When the proportion of resume tags to HR requirement tags exceeds 60%, the resume is put into a list for sorting;

[0097] S33, calculating the number of matches between the resume tag and the HR requirement tag, and sorting the resumes in descending order according to the number of matches;

[0098] S34. When the number of matches between resume tags of multiple qualified resumes and HR requirement tags is the same, the TF-IDF values ​​of the resume tags are weighted and summed to obtain a comprehensive score S TF-IDF , according to the comprehensive score S TF-IDF Sort the resumes whose resume tags have the same number of matches with HR requirement tags by the highest or lowest value:

[0099] Suppose there are n resume labels T(t 1 ,t 2 …t n ), the TF-IDF value corresponding to the i-th resume label is TF-IDF(t i ), ω(t i ) is the relevance weight of the resume tag to the HR demand tag; j HR demand tags are L(l 1 ,l 2 ,…l j ), sim(T i ,L j ) represents the similarity score between the resume tag and the HR requirement tag; the similarity score of the present invention is calculated using cosine similarity.

[0100] In step S4, the robot also constructs a job knowledge graph based on job requirements;

[0101] Define nodes in the job knowledge graph;

[0102] Main node types: job skills, work experience, education background, personality traits, other tags;

[0103] Set up sub-nodes for each main node to refine the information;

[0104] Sub-nodes of job skills: industry knowledge, software skills; sub-nodes of work experience: years of work, project experience; sub-nodes of education background: education, major, language; sub-nodes of personality traits: communication skills, teamwork; sub-nodes of other tags: work location, salary range;

[0105] The edges are defined to represent the connection between nodes, including: "requirements", "includes", and "related"; specifically, the "position skills" main node requires "industry knowledge" and "software skills" sub-nodes, which can be expressed as "position skills->requirements->industry knowledge, software skills".

[0106] Construct a job knowledge graph G = (V, E), where V is the set of all nodes and E is the set of edges between nodes; for each node v∈V, define the attribute set D v ;

[0107] To calculate the importance of a node, let f be the frequency of occurrence of the node in the job requirements, a be the importance of the node to the success of the job, obtained by expert evaluation or historical data analysis, S 节点重要性评分 =ω f ×f+ω a ×a, where ω f ,ω a The weights for frequency and importance;

[0108] Calculate the relationship strength of the nodes, let c be the co-occurrence frequency of the two nodes in the job requirements, sim be the semantic similarity between the two nodes, S 关系强度评分 =ω c ×c+ω sim ×sim, where ω c ,ω sim is the weight of co-occurrence frequency and semantic similarity;

[0109] Calculate the attribute quantification, let d be the specific attribute value, min and max be the minimum and maximum values ​​of the attribute in all job requirements;

[0110] In step S4, the step of verifying the resume tag includes:

[0111] S41, align the job seeker resume label with the job knowledge graph, and find the corresponding node in the job knowledge graph; use sim(T i ,L j ) Similarity score aligns and matches the resume tag with the corresponding HR requirement tag;

[0112] S42. Design an interview question database for each node. The question types include: open-ended questions, closed-ended questions, and situational questions; verify whether the job applicant actually possesses the skills and experience listed on the resume;

[0113] Specifically, for the resume tag under the "Software Skills" node, design the following question template:

[0114] Open-ended question: "Please describe a complex problem you solved using Resume Tags in a real-world project."

[0115] Closed question: "Do you have experience developing with the Resume Tag? (yes / no)"

[0116] Scenario question: "Suppose you need to extract specific data from a large dataset, how would you achieve that using CV Tagging?".

[0117] S43, classifying resume tags into skill tags, experience tags, and personality tags; the robot extracts questions corresponding to resume tags from the interview question database, and determines question templates based on the characteristics of the tags, and uses the template filling algorithm to dynamically generate questions based on the specific tags on the resume;

[0118] S44. When the job seeker can answer the question accurately and in detail, the tag is considered to have been verified; the node status in the knowledge graph is updated to mark the HR requirement tag corresponding to the resume tag as verified;

[0119] In step S4, the step of mining temporary tags of job seekers includes:

[0120] S45, the robot checks the job-related tags in the knowledge graph, confirms that the HR requirement tags are not explicitly mentioned or verified in the job seeker's resume, and marks them as missing tags;

[0121] S46. Evaluate the importance of missing labels based on job requirements and the relationship network in the knowledge graph, and determine their suitability and impact on the job;

[0122] The importance of missing labels, combined with the node importance calculation process in the job knowledge graph, assuming that the missing label is m, the corresponding importance is S 缺失标签重要性评分 =ω f ×f+ω m ×m, where ωf ,ω m are the weights for the frequency of occurrence and the importance of missing labels;

[0123] The fitness of missing labels is evaluated by calculating the average semantic similarity between missing labels and verified labels. The set of verified HR demand labels g is G. sim(m,g) represents the average semantic similarity between missing labels and verified HR requirement labels;

[0124] The impact of missing tags is calculated by combining the importance of missing tags and the relationship strength of other verified tags with them, and the relationship strength of S4 is calculated, where d is the co-occurrence frequency of missing tags to verified HR demand tags, ω d ,ω sim(m,g) is the weight corresponding to the co-occurrence frequency and the average semantic similarity, S 影响评分 =S 缺失标签重要性评分 ·(1+S 关系强度评分 )=(ω f ×f+ω m ×m)·(1+(ω d ×d+ω sim(m,g) ×sim(m,g)));

[0125] Combining the above scores, we can get a comprehensive evaluation score S of the missing label's suitability and impact on the position. 缺失标签 =α·S 缺失标签重要性评分 +β·S 缺失标签适合度评分 +γ·S 影响评分 , where α, β, and γ are corresponding weight coefficients, which are obtained by combining historical data and HR evaluation.

[0126] A higher comprehensive evaluation score indicates a greater importance of the missing labels.

[0127] S47. Based on the type and importance of the missing label, the robot selects questions corresponding to the missing label from the interview question database and guides the job seeker to discuss the experience, project or ability related to the missing label;

[0128] S48, mining the information of the job seeker obtained by the question in S57, the robot constructs a temporary label, associates the temporary label with the corresponding node in the knowledge graph, and corrects the knowledge graph;

[0129] Extract key information points from the job seeker's answers, including skills, experience, achievements and personal characteristics; organize the extracted key information according to the resume label classification to form temporary labels that describe the job seeker's characteristics.

[0130] Specifically, key information points point to specific skills or experiences, which can form temporary labels such as "project management ability" and "Python programming"; key information points point to personal traits, which can form temporary labels such as "strong problem-solving ability" and "proactive";

[0131] Align the newly formed temporary labels with the nodes in the job knowledge graph.

[0132] The steps to modify the knowledge graph include:

[0133] S481. When the temporary label is highly correlated with an existing node in the knowledge graph, the temporary label is added as a child node of the node;

[0134] S482, when the temporary tag represents a low relevance to an existing node in the knowledge graph, creating a supplementary node in the knowledge graph, and connecting the supplementary node to the relevant node;

[0135] S483, updating the relationship strength between the temporary label and other nodes in the knowledge graph;

[0136] Let the set of temporary labels u be U, and the semantic similarity between them and nodes v∈V Among them, u k ,v k is the vector representation of temporary labels and knowledge graph nodes; update the relationship strength, S 更新关系强度 =μ·sim(u,v)+(1-μ)·S 关系强度评分 , μ is the weight of the semantic similarity of the temporary label; update importance, S 更新重要性 =μ·sim(u,v)+(1-μ)·S 节点重要性评分 .

[0137] In step S5, the updated relationship strength, updated importance and attribute quantitative score of the job knowledge graph are used to recalculate the comprehensive score of the resume tag and the score of the temporary tag, and the overall evaluation score of the job seeker is generated in combination;

[0138] The score of the resume tag is expressed as

[0139] The score of the temporary label is expressed as

[0140] The overall evaluation score of the job seeker is S = ω 1 S 1 +ω 2 S 2 ,ω 1 +ω 2 =1;ω 1 ,ω 2are the weights of resume tag and temporary tag for HR requirement tag, where ω 0 is the baseline weight of the temporary label, |U| is the number of sets of corresponding temporary labels u.

[0141] In step S6, job seekers that meet HR requirements are sorted according to their overall evaluation scores, with job seekers with higher overall evaluation scores being ranked first, and a recommendation list is generated and recommended to HR;

[0142] For each recommended job seeker, the robot generates a detailed evaluation report based on the job seeker evaluation model. The evaluation report includes: the job seeker's basic information, the score of the resume label, the score of the temporary label, the overall evaluation score, and the specific evaluation and analysis for each label.

[0143] An intelligent recruitment system based on human-computer interaction and resume mining, including:

[0144] The job requirement parsing module is used to receive job requirement information provided by HR, parse and extract key information through natural language processing technology, and convert it into HR requirement tags;

[0145] Resume collection module, used to collect resume data from the Internet, parse and clean it;

[0146] Resume label generation module, used to extract key information from resumes and generate resume labels;

[0147] Resume screening module, used to screen and sort resumes based on the matching of HR requirement tags and resume tags;

[0148] Position knowledge graph construction module, used to construct position-related knowledge graphs;

[0149] Interview question database module, which stores and manages interview questions for different job requirements and tags;

[0150] Robot chat module, the robot extracts questions from the interview question database and chats with job seekers online;

[0151] Label verification module, the robot verifies the authenticity of resume labels;

[0152] Tag mining module: the robot mines temporary tags of job seekers;

[0153] The knowledge graph correction module is used to correct the job knowledge graph based on the information obtained in the online chat;

[0154] The job applicant evaluation model building module is used to build a job applicant evaluation model by combining the job knowledge graph, resume tags and temporary tags, and generate an overall evaluation score;

[0155] The job applicant recommendation module recommends job applicants who meet HR needs based on the job applicant evaluation model and generates an evaluation report.

[0156] The above is based on the ideal embodiment of the present invention. Through the above description, relevant personnel can make various changes and modifications without departing from the technical concept of the present invention. The technical scope of the present invention is not limited to the content in the specification, and the technical scope must be determined according to the scope of the claims.

Claims

1. An intelligent recruitment method based on human-computer interaction and resume mining, It is characterized in that Includes steps: S1. HR gives the job requirements, the robot analyzes the job requirements and generates HR requirement tags; S2, the robot searches for resumes on the Internet, parses the resumes, and extracts the corresponding resume tags; S3. Filter resumes based on HR requirement tags and sort them; S4, entrust the robot to have online chats with job seekers with resumes, verify the authenticity of resume labels, and mine temporary labels of job seekers; S5. Evaluate the importance of temporary tags and build a job applicant evaluation model based on HR demand tags, resume tags, and temporary tags; S6. The robot recommends job seekers who meet HR requirements to HR and generates an evaluation report based on the job seeker evaluation model.

2. According to claim 1, an intelligent recruitment method based on human-computer interaction and resume mining, Features: In step S2, natural language processing technology is used to split the resume text into separate words or phrases, identify the key information of the resume, and calculate the importance of the information in the key information of the resume: TF-IDF(t,d)=TF(t,d)×IDF(t). Among them, TF represents the frequency of the key information in the resume, IDF represents the inverse of the prevalence of the key information in all resumes, N is the total number of resumes, n t It represents the number of documents containing resume key information t. The higher the TF-IDF value, the more the resume key information can represent the uniqueness of the resume. The top 10-20 resume key information are selected based on the TF-IDF value.

3. According to claim 2, an intelligent recruitment method based on human-computer interaction and resume mining, Features: In step S3, S31, traverse the resumes and calculate the number of matches between resume tags and HR requirement tags; S32. When the ratio of resume tags to HR requirement tags exceeds a certain value, the resumes are put into a list for sorting; S33, calculating the number of matches between the resume tag and the HR requirement tag, and sorting the resumes in descending order according to the number of matches; S34. When the number of matches between resume tags of multiple qualified resumes and HR requirement tags is the same, the TF-IDF values ​​of the resume tags are weighted and summed to obtain a comprehensive score S TF-IDF , according to the comprehensive score S TF-IDF The resumes with the same number of matches between resume tags and HR requirement tags are sorted by the high and low values.

4. According to claim 1, an intelligent recruitment method based on human-computer interaction and resume mining, Features: In step S4, the robot also constructs a job knowledge graph based on job requirements; Construct a job knowledge graph G = (V, E), where V is the set of all nodes and E is the set of edges between nodes; for each node v∈V, define the attribute set D v ; To calculate the importance of a node, let f be the frequency of occurrence of the node in the job requirements, a be the importance of the node to the success of the job, obtained by expert evaluation or historical data analysis, S 节点重要性评分 =ω f ×f+ω a ×a, where ω f ,ω a The weights for frequency and importance; Calculate the relationship strength of the nodes, let c be the co-occurrence frequency of the two nodes in the job requirements, sim be the semantic similarity between the two nodes, S 关系强度评分 =ω c ×c+ω sim ×sim, where ω c ,ω sim is the weight of co-occurrence frequency and semantic similarity; Calculate the attribute quantification, let d be the specific attribute value, min and max be the minimum and maximum values ​​of the attribute in all job requirements; 5. According to claim 4, an intelligent recruitment method based on human-computer interaction and resume mining, Features: In step S4, the step of verifying the resume tag includes: S41, align the job seeker resume label with the job knowledge graph, and find the corresponding node in the job knowledge graph; use sim(T i ,L j ) Similarity score aligns and matches the resume tag with the corresponding HR requirement tag; S42. Design an interview question database for each node. The question types include: open-ended questions, closed-ended questions, and situational questions; verify whether the job applicant actually possesses the skills and experience listed on the resume; S43, classifying resume tags into skill tags, experience tags, and personality tags; the robot extracts questions corresponding to resume tags from the interview question database, and determines question templates based on the characteristics of the tags, and uses the template filling algorithm to dynamically generate questions based on the specific tags on the resume; S44. When the job seeker can answer the question accurately and in detail, the label is considered to have been verified; the node status in the knowledge graph is updated to mark the HR requirement label corresponding to the resume label as verified.

6. According to claim 5, an intelligent recruitment method based on human-computer interaction and resume mining, Features: In step S4, the step of mining temporary tags of job seekers includes: S45, the robot checks the job-related tags in the knowledge graph, confirms that the HR requirement tags are not explicitly mentioned or verified in the job seeker's resume, and marks them as missing tags; S46. Evaluate the importance of missing labels based on job requirements and the relationship network in the knowledge graph, and determine their suitability and impact on the job; S47. Based on the type and importance of the missing label, the robot selects questions corresponding to the missing label from the interview question database and guides the job seeker to discuss the experience, project or ability related to the missing label; S48. Mining the information of job seekers obtained from question S47, the robot constructs temporary labels, associates the temporary labels with corresponding nodes in the knowledge graph, and corrects the knowledge graph.

7. The intelligent recruitment method based on human-computer interaction and resume mining according to claim 6, Features: In step S46, The importance of missing labels, combined with the node importance calculation process in the job knowledge graph, assuming that the missing label is m, the corresponding importance is S 缺失标签重要性评分 =ω f ×f+ω m ×m, where ω f ,ω m are the weights for the frequency of occurrence and the importance of missing labels; The fitness of missing labels is evaluated by calculating the average semantic similarity between missing labels and verified labels. The set of verified HR demand labels g is G. sim(m,g) represents the average semantic similarity between missing labels and verified HR requirement labels; The impact of missing tags is calculated by combining the importance of missing tags and the relationship strength of other verified tags with them, and the relationship strength of S4 is calculated, where d is the co-occurrence frequency of missing tags to verified HR demand tags, ω d ,ω sim(m,g) is the weight corresponding to the co-occurrence frequency and the average semantic similarity, S 影响评分 =S 缺失标签重要性评分 ·(1+S 关系强度评分 )=(ω f ×f+ω m ×m)·(1+(ω d ×d+ω sim(m,g) ×sim(m,g))); Combining the above scores, we can get a comprehensive evaluation score S of the missing label's suitability and impact on the position. 缺失标签 =α·S 缺失标签重要性评分 +β·S 缺失标签适合度评分 +γ·S 影响评分 , where α, β, γ are the corresponding weight coefficients.

8. According to claim 6, an intelligent recruitment method based on human-computer interaction and resume mining, Features: The steps to modify the knowledge graph include: S481. When the temporary label is highly correlated with an existing node in the knowledge graph, the temporary label is added as a child node of the node; S482, when the temporary tag represents a low relevance to an existing node in the knowledge graph, creating a supplementary node in the knowledge graph, and connecting the supplementary node to the relevant node; S483, updating the relationship strength between the temporary label and other nodes in the knowledge graph; Let the set of temporary labels u be U, and the semantic similarity between them and nodes v∈V Among them, u k ,v k is the vector representation of temporary labels and knowledge graph nodes; update the relationship strength, S 更新关系强度 =μ·sim(u,v)+(1-μ)·S 关系强度评分 , μ is the weight of the semantic similarity of the temporary label; update importance, S 更新重要性 =μ·sim(u,v)+(1-μ)·S 节点重要性评分 .

9. The intelligent recruitment method based on human-computer interaction and resume mining according to claim 8, Features: In step S5, the updated relationship strength, updated importance and attribute quantitative score of the job knowledge graph are used to recalculate the comprehensive score of the resume tag and the score of the temporary tag, and the overall evaluation score of the job seeker is generated in combination; The score of the resume tag is expressed as The score of the temporary label is expressed as The overall evaluation score of the job seeker is S = ω 1 S 1 +ω 2 S 2 ,ω 1 +ω 2 =1;ω 1 ,ω 2 are the weights of resume tag and temporary tag for HR requirement tag, where ω 0 is the baseline weight of the temporary label, |U| is the number of sets of corresponding temporary labels u.

10. An intelligent recruitment system based on human-computer interaction and resume mining, based on the intelligent recruitment method based on human-computer interaction and resume mining described in claims 1-9, include: The job requirement parsing module is used to receive job requirement information provided by HR, parse and extract key information through natural language processing technology, and convert it into HR requirement tags; Resume collection module, used to collect resume data from the Internet, parse and clean it; Resume label generation module, used to extract key information from resumes and generate resume labels; Resume screening module, used to screen and sort resumes based on the matching of HR requirement tags and resume tags; Position knowledge graph construction module, used to construct position-related knowledge graphs; Interview question database module, which stores and manages interview questions for different job requirements and tags; Robot chat module, the robot extracts questions from the interview question database and chats with job seekers online; Label verification module, the robot verifies the authenticity of resume labels; Tag mining module: the robot mines temporary tags of job seekers; The knowledge graph correction module is used to correct the job knowledge graph based on the information obtained in the online chat; The job applicant evaluation model building module is used to build a job applicant evaluation model by combining the job knowledge graph, resume tags and temporary tags, and generate an overall evaluation score; The job applicant recommendation module recommends job applicants who meet HR needs based on the job applicant evaluation model and generates an evaluation report.

Images