Resume preliminary screening method and system fusing AI semantic analysis

Abstract

The invention discloses a resume preliminary screening method and system fusing AI semantic analysis, and relates to the technical field related to data processing, and the method comprises the steps: receiving a resume screening item of a target post, and dividing the resume screening item into a semantic clear item and a fuzzy skill item; constructing a deblurring adaptation template library; resume reconstruction is carried out according to a preset resume template; performing primary screening on the reconstructed resume set; and carrying out secondary screening on the first resume set, carrying out context-aware core coefficient analysis on the second resume set, carrying out third screening, analyzing the adaptation degree of each resume in the third resume set and a target post, sorting the resumes, and recommending the resumes to a target HR terminal. The technical problems of low resume screening precision and efficiency caused by insufficient evaluation of fuzzy skill items in resumes, single screening dimension and lack of context correlation analysis in the prior art are solved, and the technical effects of realizing automatic, multi-level and high-precision resume preliminary screening and remarkably improving the resume screening efficiency and post adaptation accuracy are achieved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a resume screening method and system that integrates AI semantic analysis. Background Technology

[0002] Resume screening is a crucial and time-consuming process in human resource management. Traditional resume screening mainly relies on keyword matching, rule filtering, or manual initial review. While simple and easy to use, it has significant limitations. For example, keyword matching cannot handle semantic complexity, especially when identifying vague descriptions of skills and experience; manual screening is inefficient, highly subjective, and struggles to handle large-scale resume processing. Artificial intelligence-based natural language processing and semantic analysis offer new solutions for resume screening. AI-based resume screening typically utilizes text mining, entity recognition, and semantic similarity calculation to parse and match resume content. However, job requirements often include both semantically explicit and vague skill requirements. Due to their abstract expression and strong contextual dependence, these are difficult to process directly using rules or simple semantic models. Furthermore, the diverse structures and inconsistent formats of resume documents make information extraction and standardization difficult, affecting screening accuracy. In addition, single-stage screening can easily overlook contextual information and the correlation between skills, potentially leading to incorrect or missed screenings.

[0003] Therefore, current technologies suffer from insufficient evaluation of vague skills in resumes, limited screening dimensions, and a lack of contextual analysis, resulting in low accuracy and efficiency in resume screening. Summary of the Invention

[0004] This application provides a resume screening method and system that integrates AI semantic analysis, which solves the technical problems of insufficient evaluation of ambiguous skill items in resumes, single screening dimensions, and lack of contextual analysis in existing technologies, resulting in low resume screening accuracy and efficiency. It achieves automated, multi-level, and high-precision resume screening, significantly improving resume screening efficiency and job matching accuracy.

[0005] This application provides a resume screening method integrating AI semantic analysis. The method includes: receiving resume screening items for a target position and dividing them into semantically explicit items and fuzzy skill items; performing defuzzification analysis based on skill association semantics on the fuzzy skill items to construct a defuzzification adaptation template library; receiving a resume dataset corresponding to the target position and reconstructing the resumes according to a preset resume template to generate a reconstructed resume set; performing a first screening on the reconstructed resume set based on the semantically explicit items to generate a first resume set; performing a second screening on the first resume set based on the fuzzy adaptation template library to obtain a second resume set; performing context-aware core coefficient analysis on the second resume set; performing a third screening based on the core coefficients; and combining the semantically explicit items to analyze the fit between each resume in the third resume set and the target position, sorting the resumes, and recommending them to the target HR terminal.

[0006] In a possible implementation, the resume screening method integrating AI semantic analysis further performs the following processing: reading the entity type of the fuzzy skill item, analyzing the associated entity attributes of the entity type and extracting entity relationships to construct a skill association graph; collecting multiple sets of actual interview samples for the target position, each set of actual interview samples including fuzzy skill item text and actual interview score samples; mapping the fuzzy skill item text to the skill association graph, extracting sub-graphs to construct defuzzification adaptation templates, and labeling the adaptation level with the corresponding actual interview score samples to construct the defuzzification adaptation template library.

[0007] In a possible implementation, the resume screening method that integrates AI semantic analysis also performs the following processing: collecting an associated knowledge base based on the entity type; performing skill association analysis on the associated knowledge base to construct a technology stack combination; and constructing the skill association graph using the technology stack combination.

[0008] In a possible implementation, the resume screening method that integrates AI semantic analysis also performs the following processing: skill association relationships include at least parent-child hierarchical relationships and co-occurrence relationships.

[0009] In a possible implementation, the resume screening method integrating AI semantic analysis further performs the following processing: determining whether semantically explicit items and vague skill items have opposing skill prerequisites; if so, collecting resume template field samples for skill prerequisites and constructing a skill prerequisite filter; inputting resumes that have been filtered out after the first and second screenings into the skill prerequisite filter for rescreening, generating marked resumes that meet the skill prerequisites, and supplementing and optimizing the second resume set.

[0010] In a possible implementation, the resume screening method integrating AI semantic analysis further performs the following processing: extracting any skill-related statements of fuzzy skill items from any resume in the second resume set and performing NLP preprocessing to generate any NLP processing information; identifying grammatical modification relationships in any NLP processing information to generate any syntax tree; performing semantic role labeling on the any syntax tree to generate an annotated syntax tree; and evaluating the annotated syntax tree according to a preset core scoring weight template to complete the core coefficient analysis.

[0011] In a possible implementation, the resume screening method that integrates AI semantic analysis also performs the following processing: the NLP preprocessing includes sentence segmentation, word segmentation and part-of-speech tagging, and entity recognition.

[0012] In a possible implementation, the resume screening method integrating AI semantic analysis further performs the following processing: reading the hierarchical scoring mechanism for the semantically explicit items; using the hierarchical scoring mechanism to perform hierarchical evaluation of the semantically explicit items for each resume in the third resume set, and obtaining hierarchical evaluation results; weighting the hierarchical evaluation results with core coefficients and then normalizing them to obtain the suitability of each resume for the target position.

[0013] In a possible implementation, the resume screening method that integrates AI semantic analysis also performs the following processing: extracting the professional experience statements from each resume in the third resume set; identifying resignation characteristics based on the professional experience statements to complete the career stability evaluation; and screening the third resume set four times according to the career stability evaluation results and a career stability threshold.

[0014] This application also provides a resume screening system integrating AI semantic analysis. The system includes: a resume screening item division module, used to receive resume screening items for the target position and divide them into semantically explicit items and fuzzy skill items; a defuzzification analysis module, used to perform defuzzification analysis based on skill association semantics for the fuzzy skill items and construct a defuzzification adaptation template library; a resume reconstruction module, used to receive the resume dataset corresponding to the target position and reconstruct the resumes according to a preset resume template to generate a reconstructed resume set; a reconstructed resume set screening module, used to perform a first screening on the reconstructed resume set based on the semantically explicit items to generate a first resume set; and a fit analysis module, used to perform a second screening on the first resume set based on the fuzzy adaptation template library to obtain a second resume set, perform context-aware core coefficient analysis on the second resume set, perform a third screening according to the core coefficients, and combine the semantically explicit items to analyze the fit between each resume in the third resume set and the target position, sort the resumes and recommend them to the target HR terminal.

[0015] This application proposes a resume screening method and system that integrates AI semantic analysis. The method receives resume screening items for a target position and categorizes them into semantically explicit items and fuzzy skill items. It then constructs a defuzzification and adaptation template library; reconstructs resumes according to preset templates; performs a first screening on the reconstructed resume set; performs a second screening on the first resume set; performs a third screening on the second resume set using context-aware core coefficient analysis; analyzes the fit of each resume in the third resume set with the target position; sorts the resumes; and recommends them to the target HR terminal. This addresses the technical problems in existing technologies, such as insufficient evaluation of fuzzy skill items in resumes, single screening dimensions, and lack of contextual analysis, leading to low resume screening accuracy and efficiency. The system achieves automated, multi-level, and high-precision resume screening, significantly improving resume screening efficiency and job matching accuracy. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a schematic diagram of the resume screening method integrating AI semantic analysis provided in the embodiments of this application.

[0018] Figure 2 This is a schematic diagram of the structure of a resume screening system that integrates AI semantic analysis, provided in an embodiment of this application.

[0019] Figure labeling: Module 10 for resume screening criteria, Module 20 for defuzzification analysis, Module 30 for resume reconstruction, Module 40 for reconstructed resume set screening, and Module 50 for suitability analysis. Detailed Implementation

[0020] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0023] This application provides a resume screening method that integrates AI semantic analysis, such as... Figure 1 As shown, the method includes: Step S100: Receive resume screening items for the target position and divide them into semantically explicit items and vague skill items.

[0024] Preferably, the resume screening criteria are the specific recruitment requirements and conditions set by the company's HR for the target job. The system receives resumes for the target position and categorizes them into semantically explicit criteria and fuzzy skill criteria. Semantically explicit criteria act as hard filters, handling hard conditions with clearly defined rules and easy-to-programmable implementation. Fuzzy skill criteria are the core evaluator, handling soft skills that require intelligent semantic analysis to quantify. Specifically, semantically explicit criteria refer to screening conditions judged using objective, quantitative, binary, or enumerable standards, without relying on deep semantic analysis of the context. These may include educational requirements, years of work experience, specific certifications, location requirements, and hard language requirements. And specific company experience, for example, educational requirements of bachelor's degree or above, the criteria for judgment is whether the candidate's education field is bachelor's, master's or doctoral; work experience of more than 3 years, the criteria for judgment is whether the total work experience calculated in the resume is greater than or equal to 3; PMP certification is required, the criteria for judgment is whether the certificate section of the resume contains the keyword PMP; location requirement is to be based in xx location, the criteria for judgment is whether the current location field in the resume is xx location; language requirement is a hard requirement of CET-6, the criteria for judgment is whether the certificate or skills section contains CET-6; specific company experience, the criteria for judgment is whether the company name field of the work experience matches.

[0025] Preferably, fuzzy skill items refer to those that cannot be directly judged by simple matching. Their meaning and true level are highly dependent on the context, modifiers, specific project experience, and other skill-related screening conditions. They include subjective modifiers indicating proficiency and ability level, such as "proficient," "familiar," "understand," "master," and "good." They must be defuzzified and interpreted using NLP techniques such as entity relationship recognition, semantic role labeling, and knowledge graphs. They also have strong context dependence and relevance and may include skills with degree modifiers, soft skills / abstract abilities, and broad technical fields. For example, proficient in Java programming, proficient in Python, possessing excellent communication skills, possessing strong teamwork spirit, having cloud computing experience, or being familiar with the big data technology ecosystem.

[0026] Step S200: Perform defuzzification analysis based on skill association semantics on the fuzzy skill items and construct a defuzzification adaptation template library.

[0027] Step S200 further includes step S210, reading the entity type of the fuzzy skill item, analyzing the associated entity attributes of the entity type and extracting entity relationships to construct a skill association graph; step S220, collecting multiple sets of actual interview samples for the target position, each set of actual interview samples including fuzzy skill item text and actual interview score samples; step S230, mapping the fuzzy skill item text to the skill association graph, extracting sub-graphs to construct defuzzification adaptation templates, and labeling the adaptation level with the corresponding actual interview score samples to construct the defuzzification adaptation template library.

[0028] Preferably, defuzzification analysis based on skill association semantics is performed on fuzzy skill items. Specifically, the entity type of the fuzzy skill item is read, that is, the core skill entity in the fuzzy skill item is identified and its category label is determined. Then, the entity is queried from related knowledge bases such as industry knowledge graphs, technical document libraries, and recruitment databases. The associated entity attributes of the entity type are analyzed, that is, the specific characteristics, parameters or sub-items of the skill entity. The logical relationship between the skill and other skills or concepts is extracted, such as parent-child hierarchical relationship or co-occurrence relationship. All entities are used as nodes and entity relationships are used as edges to construct a skill association graph. Then, multiple sets of actual interview sample data for the target position are found from historical data. Each set of actual interview samples includes fuzzy skill item text and actual interview score samples. The fuzzy skill item text refers to the skill point examined during the interview, and the actual interview score samples refer to the interviewer's final score or level evaluation of the candidate for this skill.

[0029] Preferably, the fuzzy skill item text input by HR is mapped to a skill association graph, and the closely related entities and relationships around the skill item are extracted to form a sub-graph, which is used to define the complete instruction structure of the skill entity and is used as a defuzzification adaptation template. Then, the adaptation level is labeled with the corresponding actual interview score sample, that is, the defuzzification adaptation template is associated with the actual score of the interview sample. Through learning from a large number of samples, multiple templates are built for a fuzzy skill item, and different adaptation levels such as "proficient", "familiar", and "understand" are labeled respectively. Finally, all the templates created for the "target position" with level labels are collected to form a defuzzification adaptation template library, which is used to scan and screen resumes with the corresponding templates, and can comprehensively calculate the score of the skill based on the number and importance of the related skills found.

[0030] Furthermore, step S210 also includes step S211, collecting associated knowledge bases according to the entity type; step S212, performing skill association analysis on the associated knowledge base and constructing a technology stack combination; and step S213, constructing the skill association graph using the technology stack combination.

[0031] Preferably, the entity type is a predefined skill category label in natural language processing. Based on the identified entity type, relevant structured or unstructured data sources are automatically crawled or accessed to collect related knowledge bases, such as technical encyclopedia websites, technical documents, and professional social networking platforms. Data mining and natural language processing are used to analyze the skill association relationships in the related knowledge bases, including co-occurrence analysis and semantic similarity analysis. Co-occurrence analysis refers to counting the frequency of two skill entities appearing in the same document, the same text, or the same resume; the higher the frequency, the stronger the association. Semantic similarity analysis uses a word embedding model to convert skill names into vectors and calculates the semantic cosine similarity between vectors to find semantically similar skills. Then, a technology stack combination is constructed, which is a set of frequently used, grouped skill items. Finally, a skill association graph is constructed based on the technology stack combination, including treating each skill as a node and the association relationship between skills as an edge connecting the nodes, and converting the technology stack combination into edges in the graph. The final skill association graph is used for efficient resume querying and traversal.

[0032] Furthermore, step S212 also includes that the skill association relationship includes at least the skill parent-child hierarchy relationship and the co-occurrence relationship.

[0033] Preferably, the parent-child hierarchical relationship of skills is a vertical, hierarchical classification relationship used to describe the subordinate relationship between skills at the level of abstraction, inclusion, or technical category. This gives the skill association map a hierarchical structure, which is convenient for reasoning and expansion. Using the parent-child hierarchical relationship, not only can resumes with specific skill items be matched, but also resumes with their child skills can be matched. The co-occurrence relationship refers to a horizontal, statistical association relationship used to describe the probability of different skill items appearing at the same time in resumes, job descriptions, and project experience. It reveals the combined correlation between various skill items. Through the co-occurrence relationship, the authenticity of skills can be verified, and the depth and proficiency of skills can be evaluated.

[0034] Step S300: Receive the resume dataset corresponding to the target position, and reconstruct the resumes according to the preset resume template to generate a reconstructed resume set.

[0035] Preferably, a dataset of resumes corresponding to the target position is collected from recruitment websites, corporate emails, and internal talent pools. This includes all original resume files for the target position, in formats such as PDF, Word, and HTML web page text. The preset resume template refers to a predefined structured resume data pattern that specifies the classification and storage rules for resume information. Resume reconstruction according to the preset resume template involves using natural language processing and information extraction to automatically parse, identify, and fill the unstructured original resume text into the structured preset template. Specifically, OCR and text extraction are used to convert the file content into plain text and perform preprocessing, including format cleaning and encoding conversion. Then, an NLP model is used to identify and extract specific information fragments from the messy text, including identifying entities and entity relationships in the text. The extracted information is then categorized and filled according to the fields of the preset resume template to generate a reconstructed resume. This reconstructed resume set improves the accuracy and efficiency of the analysis.

[0036] Step S400: Based on the semantically explicit items, the reconstructed resume set is filtered once to generate a first resume set.

[0037] Preferably, the reconstructed resume set is screened using semantically explicit items. This involves comparing the specific field values ​​of each reconstructed resume with the semantically explicit items, and generating multiple judgment rules based on the resume screening items for the target position. These rules include checking whether the education field is compliant, whether the years of work experience field is compliant, whether the certificate field contains certificate keywords, and whether the current location field is compliant. If all of these conditions are met, the resume is selected as the first resume, and multiple resumes are then selected to form the first resume set.

[0038] Step S500: Based on the fuzzy matching template library, the first resume set is screened a second time to obtain the second resume set. The second resume set is then subjected to context-aware core coefficient analysis. After three screenings based on the core coefficients, the suitability of each resume in the third resume set with the target position is analyzed in conjunction with semantically explicit items. The resumes are then sorted and recommended to the target HR terminal.

[0039] Preferably, a secondary screening of the first resume set is performed using a fuzzy matching template library to evaluate the skill depth and authenticity of each resume in the first resume set. Specifically, the matching template corresponding to any fuzzy skill item is used as a standard to compare the detailed content of each resume, determine whether the associated skills defined in the template also appear in the resume, and count their frequency and context, thereby assigning a preliminary score to each fuzzy skill item in each resume. The resumes selected through this screening form a second resume set, which includes resumes that have passed the initial skill verification. Then, a context-aware core coefficient analysis is performed on the second resume set, that is, to understand the specific role and true meaning of skill words in sentences and judge the value of their skill level. Specifically, syntactic analysis and semantic role labeling are used to decompose the sentences in the resumes to identify key signals, such as action intensity, fuzzy achievement quantification, and technical depth, and then a core coefficient is generated to represent the true level of core skills and experience depth of the resume. The resumes are then sorted according to the core coefficient and a coefficient threshold is set to screen resumes that match the skills but have significantly insufficient experience depth. The resumes that pass the screening are combined to form a third resume set. Then, the suitability of each resume in the third resume set to the target position is analyzed by combining semantically explicit items. That is, the suitability is determined by weighted summation of the fuzzy skill item score and the semantically explicit item score for each resume in the third resume set. The weight may be 70% for skill quality and 30% for hard requirements. Finally, all resumes are sorted from high to low suitability and pushed to the target HR terminal, such as the backend of the company's recruitment system, so that they receive a precisely sorted candidate list, thereby significantly improving the efficiency of resume screening and the accuracy of job matching, and improving the efficiency and accuracy of recruitment decisions.

[0040] Furthermore, step S500 also includes step S510, determining whether semantically explicit items and fuzzy skill items have opposing skill prerequisites; step S520, if so, collecting resume template field samples for skill prerequisites and constructing a skill prerequisite filter; step S530, inputting the resumes that were filtered out after the first and second screenings into the skill prerequisite filter for rescreening, generating marked resumes that meet the skill prerequisites, and supplementing and optimizing the second resume set.

[0041] Preferably, the process involves determining whether semantically explicit and vaguely defined skill items have opposing skill prerequisites. This includes situations where there might be a trade-off or compromise between semantically explicit and vaguely defined skill items, such as a resume failing to meet hard requirements but demonstrating exceptional skill level that leads the recruiter to make an exception. For example, regardless of other conditions, a candidate resume with skill X is considered for the next round, or if a candidate resume demonstrates master-level skill Y, the requirement for years of work experience can be relaxed. If opposing skill prerequisites exist, a sample of resume template fields that meet these conditions is collected. This sample is then analyzed and trained to construct a skill prerequisite filter, which accurately identifies resumes containing extremely important markers from the rejected resumes. The resumes that were rejected after the first and second rounds of filtering are then input into the skill prerequisite filter for re-filtering to determine the marked resumes that meet the skill prerequisites. Finally, the selected resumes are added back to the second resume set, thus supplementing and optimizing the candidate resumes.

[0042] Furthermore, step S500 also includes step S540, performing NLP preprocessing on any skill-related statement of any fuzzy skill item extracted from any resume in the second resume set to generate any NLP processing information; step S550, performing grammatical modification relation recognition on any NLP processing information to generate any syntax tree; step S560, performing semantic role labeling on the any syntax tree to generate an annotated syntax tree; and step S570, evaluating the annotated syntax tree according to a preset core scoring weight template to complete the core coefficient analysis.

[0043] Step S540 further includes the NLP preprocessing including sentence segmentation, word segmentation and part-of-speech tagging, and entity recognition.

[0044] Preferably, extract any skill-related statement from any resume in the second resume set that contains fuzzy skill items. For example, if I use Python and Pandas libraries to lead a user behavior data analysis project, build a predictive model, and improve click-through rate accuracy by 30%, perform NLP preprocessing on it, including sentence segmentation and word segmentation, part-of-speech tagging, and text entity recognition. That is, cut the sentence into independent lexical units, label the part of speech of each word, and identify named entities in the text to form any NLP processing information containing multiple linguistic tags. Then, perform grammatical modification relationship recognition on any NLP processing information, that is, use grammatical parsing to determine the grammatical modification relationship of all words, including who modifies whom, etc. By identifying which part of speech is the object of which part of speech, and which is the core verb of the sentence, the surface structure of the sentence is revealed. Then, by combining part-of-speech tags, any syntax tree can be constructed, clearly demonstrating the grammatical relationships between words. Semantic role labeling is performed on any syntax tree, including determining the semantic role of each component in the event described by the sentence, generating an annotated syntax tree. Finally, the annotated syntax tree is evaluated according to a preset core scoring weight template. This template defines the weight of each semantic role in demonstrating the skill level. Then, each semantic role in the annotated syntax tree is traversed, and a weighted sum is performed according to the core scoring weight template to generate a core coefficient, representing the value and depth of the skill experience described in the sentence.

[0045] Furthermore, step S500 also includes step S580, reading the hierarchical scoring mechanism for the semantically explicit items; step S590, using the hierarchical scoring mechanism to perform hierarchical evaluation of the semantically explicit items for each resume in the third resume set, and obtaining the hierarchical evaluation result; step S5100, weighting the hierarchical evaluation result with the core coefficient and then normalizing it to obtain the suitability of each resume with the target position.

[0046] Preferably, the semantically explicit item grading and scoring mechanism is a pre-defined rule used to score the degree of the hard conditions for judgment. For example, more than 10 years of work experience is 100 points, 7-10 years is 85 points, and 5-7 years is 70 points; a doctoral degree is 100 points, a master's degree is 85 points, and a bachelor's degree is 70 points. Then, the semantically explicit item grading and evaluation of each resume in the third resume set is carried out using the grading and scoring mechanism. That is, each resume in the third resume set is traversed, and the hard condition fields in its structured data are scored according to the grading and scoring mechanism to determine the grading and evaluation results, including the score of each resume for each hard condition. Then, the grading and evaluation results including the comprehensive score of hard conditions are weighted and summed with the core coefficient representing skill quality. The weight is determined by the recruitment strategy and usually has a greater bias towards skill quality. Then, normalization processing is performed, that is, all scores are converted to the same standard range through mathematical methods. Finally, the suitability of each resume to the target position is generated, which greatly improves the efficiency of resume screening and the recruitment success rate.

[0047] Furthermore, step S500 also includes extracting the career experience statements from each resume in the third resume set; identifying resignation characteristics based on the career experience statements to complete the career stability evaluation; and performing four rounds of screening on the third resume set according to the career stability evaluation results and a career stability threshold.

[0048] Preferably, the professional experience statements of each resume are extracted from the third set of resumes. This includes accurately locating and extracting all text fields related to work experience, such as company name, job title, employment period, and job description. Then, based on the professional experience statements, turnover characteristics are identified, which may include: identification of short tenure, calculating the duration of each job and identifying whether there are multiple consecutive job experiences that are shorter than the industry or job standard; identification of frequent job-hopping patterns, calculating the number of job changes within the total years of employment and identifying whether there is an excessively high job change frequency; identification of career gaps, analyzing the time interval between two adjacent jobs and identifying excessively long gaps that cannot be reasonably explained; identification of downward career trajectory, judging whether career development has declined or stagnated by analyzing the sequence of job titles, such as from senior engineer to engineer, or long-term lack of promotion at the same job level; and identification of industry / functional jumps, analyzing whether there is a lack of continuity in the industry and function of different companies, as frequent jumps between industries and functions may indicate unclear career planning. Finally, a career stability score or risk level label is generated for each candidate to complete the career stability evaluation and obtain the career stability evaluation result. Finally, based on the career stability evaluation results, the third set of resumes was screened four times according to the career stability threshold. The career stability threshold is a pre-set minimum stability score. Multiple career stability evaluation results in the third set of resumes were compared with the career stability threshold. If the career stability evaluation result is better than or equal to the career stability threshold, it is retained; if the career stability evaluation result is worse than the career stability threshold, it is eliminated from the third set of resumes.

[0049] In the above text, refer to Figure 1 This paper describes in detail a resume screening method incorporating AI semantic analysis according to embodiments of the present invention. Next, we will refer to... Figure 2 This invention describes a resume screening system that integrates AI semantic analysis according to an embodiment of the present invention.

[0050] The resume screening system integrating AI semantic analysis according to embodiments of the present invention addresses the technical problems in existing technologies, such as insufficient evaluation of ambiguous skill items in resumes, single screening dimensions, and lack of contextual analysis, leading to low resume screening accuracy and efficiency. It achieves automated, multi-level, and high-precision resume screening, significantly improving resume screening efficiency and job suitability accuracy. Figure 2As shown, the resume screening system integrating AI semantic analysis includes: a resume screening item division module 10, a defuzzification analysis module 20, a resume reconstruction module 30, a reconstructed resume set screening module 40, and a fit analysis module 50.

[0051] The resume screening item segmentation module 10 receives resume screening items for the target position and divides them into semantically explicit items and fuzzy skill items; the defuzzification analysis module 20 performs defuzzification analysis based on skill association semantics for the fuzzy skill items and constructs a defuzzification adaptation template library; the resume reconstruction module 30 receives the resume dataset corresponding to the target position and reconstructs the resumes according to preset resume templates to generate a reconstructed resume set; the reconstructed resume set screening module 40 performs a first screening on the reconstructed resume set based on the semantically explicit items to generate a first resume set; the fit analysis module 50 performs a second screening on the first resume set based on the fuzzy adaptation template library to obtain a second resume set, performs context-aware core coefficient analysis on the second resume set, performs a third screening based on the core coefficients, and combines the semantically explicit items to analyze the fit between each resume in the third resume set and the target position, sorts the resumes, and recommends them to the target HR terminal.

[0052] The specific configuration of the defuzzification analysis module 20 will be described in detail below. The defuzzification analysis module 20 further includes: reading the entity type of the fuzzy skill item, analyzing the associated entity attributes of the entity type and extracting entity relationships to construct a skill association graph; collecting multiple sets of actual interview samples for the target position, each set of actual interview samples including fuzzy skill item text and actual interview score samples; mapping the fuzzy skill item text to the skill association graph, extracting sub-graphs to construct defuzzification adaptation templates, and labeling the adaptation level with the corresponding actual interview score samples to construct the defuzzification adaptation template library.

[0053] The specific configuration of the deblurring analysis module 20 will be described in detail below. The deblurring analysis module 20 further includes: collecting an associated knowledge base based on the entity type; performing skill association analysis on the associated knowledge base to construct a technology stack combination; and constructing the skill association graph using the technology stack combination.

[0054] The specific configuration of the deblurring analysis module 20 will be described in detail below. The deblurring analysis module 20 further includes: skill association relationships, which at least include parent-child hierarchical relationships and co-occurrence relationships.

[0055] The specific configuration of the fit analysis module 50 will be described in detail below. The fit analysis module 50 further includes: determining whether semantically explicit items and ambiguous skill items have opposing skill prerequisites; if so, collecting resume template field samples for the skill prerequisites and constructing a skill prerequisite filter; inputting resumes that were filtered out after the first and second rounds of filtering into the skill prerequisite filter for re-filtering, generating marked resumes that meet the skill prerequisites, and supplementing and optimizing the second resume set.

[0056] The specific configuration of the fit analysis module 50 will be described in detail below. The fit analysis module 50 further includes: performing NLP preprocessing on any skill-related statement extracted from any resume in the second resume set for any fuzzy skill item, generating any NLP processing information; identifying grammatical modification relationships on any NLP processing information, generating any syntax tree; performing semantic role annotation on the syntax tree, generating an annotated syntax tree; and evaluating the annotated syntax tree according to a preset core scoring weight template to complete the core coefficient analysis.

[0057] The specific configuration of the fit analysis module 50 will be described in detail below. The fit analysis module 50 further includes: the NLP preprocessing includes sentence segmentation, word segmentation and part-of-speech tagging, and entity recognition.

[0058] The specific configuration of the fit analysis module 50 will be described in detail below. The fit analysis module 50 further includes: reading the hierarchical scoring mechanism for the semantically explicit items; performing hierarchical evaluation of the semantically explicit items for each resume in the third resume set using the hierarchical scoring mechanism to obtain hierarchical evaluation results; and weighting and normalizing the hierarchical evaluation results with the core coefficients to obtain the fit degree between each resume and the target position.

[0059] The specific configuration of the fit analysis module 50 will be described in detail below. The fit analysis module 50 further includes: extracting the professional experience statements from each resume in the third resume set; identifying turnover characteristics based on the professional experience statements to complete the career stability evaluation; and performing four rounds of screening on the third resume set according to the career stability evaluation results and a career stability threshold.

[0060] The resume screening system integrating AI semantic analysis provided in this embodiment of the invention can execute the resume screening method integrating AI semantic analysis provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0061] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0062] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A resume screening method integrating AI semantic analysis, characterized in that, include: The resume screening criteria for the target position are divided into semantically explicit criteria and vague skill criteria; For the aforementioned fuzzy skill items, a defuzzification analysis based on skill association semantics is performed to construct a defuzzification adaptation template library; Receive the resume dataset corresponding to the target position, and reconstruct the resumes according to the preset resume template to generate a reconstructed resume set; Based on the semantically explicit terms, the reconstructed resume set is filtered once to generate a first resume set; The first resume set is filtered a second time based on the fuzzy matching template library to obtain the second resume set. The second resume set is then subjected to context-aware core coefficient analysis. After filtering a third time based on the core coefficients, the suitability of each resume in the third resume set with the target position is analyzed in conjunction with semantically explicit items. The resumes are then sorted and recommended to the target HR terminal.

2. The resume screening method integrating AI semantic analysis as described in claim 1, characterized in that, For the aforementioned fuzzy skill items, a defuzzification analysis based on skill association semantics is performed, and a defuzzification adaptation template library is constructed, including: Read the entity type of the fuzzy skill item, analyze the associated entity attributes of the entity type and extract the entity relationships to construct a skill association graph; Multiple sets of actual interview samples for the target position were collected. Each set of actual interview samples included fuzzy skill item text and actual interview score samples. The text of fuzzy skill items is mapped to the skill association graph, sub-graphs are extracted to construct defuzzification adaptation templates, and adaptation levels are labeled with corresponding actual interview scoring samples to construct the defuzzification adaptation template library.

3. The resume screening method integrating AI semantic analysis as described in claim 2, characterized in that, Read the entity type of the fuzzy skill item, analyze the associated entity attributes of the entity type and extract entity relationships, and construct a skill association graph, including: Collect associated knowledge bases based on the entity type; Skill association analysis is performed on the associated knowledge base to construct a technology stack combination; The skill association graph is constructed using the aforementioned technology stack.

4. The resume screening method integrating AI semantic analysis as described in claim 3, characterized in that, Skill association relationships include at least parent-child hierarchical relationships and co-occurrence relationships.

5. The resume screening method integrating AI semantic analysis as described in claim 1, characterized in that, After receiving the second set of resumes, it also includes: Determine whether semantically explicit items and vague skill items have opposing skill prerequisites; If so, collect resume template field samples based on skill prerequisites and build a skill prerequisite filter; The resumes that were rejected after the first and second screenings are input into the skills prerequisite filter for rescreening, generating marked resumes that meet the skills prerequisite priority conditions, and supplementing and optimizing the second resume set.

6. The resume screening method integrating AI semantic analysis as described in claim 5, characterized in that, Context-aware core coefficient analysis was performed on the second set of resumes, including: For any resume in the second resume set, extract any skill-related statement from any fuzzy skill item and perform NLP preprocessing to generate any NLP processing information; Recognize syntactic modification relations for any NLP processed information and generate any syntax tree; Semantic role labeling is performed on any of the syntax trees to generate an annotated syntax tree; The labeled syntax tree is evaluated according to the preset core scoring weight template to complete the core coefficient analysis.

7. The resume screening method integrating AI semantic analysis as described in claim 6, characterized in that, The NLP preprocessing includes sentence segmentation, word segmentation and part-of-speech tagging, and entity recognition.

8. The resume screening method integrating AI semantic analysis as described in claim 6, characterized in that, The suitability of each resume in the third resume set for the target position is analyzed by combining semantically explicit terms, including: Read the hierarchical scoring mechanism for the semantically explicit items; The aforementioned hierarchical scoring mechanism is used to perform hierarchical evaluation of semantically explicit items in each resume in the third resume set, and the hierarchical evaluation results are obtained. The grading evaluation results are weighted with the core coefficient and then normalized to obtain the suitability of each resume to the target position.

9. The resume screening method integrating AI semantic analysis as described in claim 1, characterized in that, After three rounds of screening based on the core coefficient, the following are also included: Extract the professional experience statements from each resume in the third resume set; Based on the stated professional experience, exit characteristics are identified to complete the career stability assessment. Based on the job stability evaluation results, the third set of resumes was screened four times according to the job stability threshold.

10. A resume screening system integrating AI semantic analysis, characterized in that, The system is used to implement the resume screening method integrating AI semantic analysis as described in any one of claims 1 to 9, and the system includes: The resume screening criteria module is used to receive resume screening criteria for the target position and divide them into semantically explicit criteria and vague skill criteria. The defuzzification analysis module is used to perform defuzzification analysis based on skill association semantics for the fuzzy skill items and to build a defuzzification adaptation template library. The resume reconstruction module is used to receive the resume dataset corresponding to the target position, and reconstruct the resumes according to the preset resume template to generate a reconstructed resume set. The reconstructed resume set filtering module is used to filter the reconstructed resume set once based on the semantically explicit items, and generate a first resume set; The fit analysis module is used to perform a second screening of the first resume set based on the fuzzy fit template library to obtain a second resume set. The second resume set is then subjected to context-aware core coefficient analysis. After a third screening based on the core coefficients, the fit of each resume in the third resume set with the target position is analyzed in conjunction with semantically explicit items. The resumes are then sorted and recommended to the target HR terminal.

Images