Method and system for screening and displaying resumes for human resource management

Abstract

The invention discloses a resume screening display method and system for human resource management. The method comprises the following steps: constructing a keyword-dimension mapping relation table based on a post description text through a natural language processing engine; analyzing the keyword frequency, syntactic position, modification intensity and semantic association degree to generate a post demand intensity vector; historical recruitment data are retrieved in combination with the post meta-information, and historical weight vectors are generated; self-adaptively calculating a smoothing coefficient according to the historical data volume, and fusing the double-source weight to generate a dynamic weight vector; standardized scoring is carried out on each dimension of the resume, a matching degree is calculated by adopting a normalization formula only based on effective dimensions, and score distortion caused by missing fields is avoided; and finally displaying results according to the matching degree sequence. According to the method, a preset weight mode is abandoned, one-post-one-strategy dynamic screening is realized, the matching accuracy, the operation efficiency and the man-machine collaborative decision-making quality are improved, and a scientific, efficient and transparent intelligent solution is provided for human resource management.

Description

Technical Field

[0001] This invention relates to the field of human resource management technology, and in particular to a resume screening and display method and system for human resource management. Background Technology

[0002] In the field of human resource management, resume screening is a core part of the recruitment process, and its scientific nature and efficiency directly affect the quality and cost control of talent acquisition for enterprises. Currently, mainstream recruitment systems generally adopt screening mechanisms based on preset rules, such as setting fixed weights for priority based on major or skills, or conducting initial screening through simple keyword matching. However, such methods have revealed multiple technical shortcomings in practical applications.

[0003] First, the weight configuration is highly rigid: the system only provides a limited number of preset modes for human resources specialists to choose from, and cannot dynamically generate differentiated weights based on the semantic details of the job description text. Furthermore, the system does not consider the impact of historical recruitment data on weight allocation, resulting in the use of preset weight templates when historical data is insufficient, causing a significant drop in matching accuracy. On the other hand, when data is abundant, it cannot make full use of the statistical patterns of historical data, making it impossible for the system to maintain accuracy in data-sparse scenarios or to leverage the value of historical experience in data-rich scenarios.

[0004] Secondly, the depth of historical data mining is insufficient: existing systems utilize historical recruitment data superficially, mostly for statistical display, lacking in-depth mining of the statistical correlation between historical recruitment results and resume dimensional features. Although existing technologies mention automatically adapting different weight templates based on job type identification, their weight allocation depends entirely on the preset job type, without establishing a dynamic correlation mechanism between historical data volume and weight allocation; although existing technologies involve resume matching, they do not consider the impact of historical data volume on weight, nor do they establish a scientific integration mechanism between historical data and current job requirements.

[0005] Third, the missing field handling mechanism has a fundamental flaw: when certain screening dimensions are missing in a resume, the existing system usually forces the score of that dimension to zero and includes it in the overall calculation, which leads to an unreasonable decrease in the matching degree. This makes it impossible for the system to accurately assess the matching degree of candidates when the resume data is incomplete, which seriously undermines the fairness and objectivity of the screening.

[0006] The aforementioned problems are intertwined and collectively constrain the overall performance of resume screening in terms of accuracy, fairness, efficiency, and interpretability. This makes it difficult for existing technologies to meet the urgent needs of enterprises for intelligent, data-driven recruitment tools. There is an urgent need for an innovative solution that can deeply integrate job semantic analysis with historical data mining, intelligently handle missing data, and achieve human-machine collaborative optimization. Summary of the Invention

[0007] The purpose of this invention is to provide a resume screening and display method and system for human resource management.

[0008] The problem this invention aims to solve is that existing resume screening technologies suffer from rigid weight configuration, reliance on preset modes that cannot dynamically adapt to differences in job descriptions, superficial mining of historical recruitment data, and low efficiency and strong subjectivity in manual weight configuration.

[0009] A method and system for screening and displaying resumes in human resource management, the technical solution of which is as follows: S1: Job Description Text Input and Semantic Analysis: The HR specialist inputs the plain text of the job description; the integrated natural language processing engine performs word segmentation, stop word filtering, part-of-speech tagging, and semantic role recognition on the input job description text; keywords and phrases related to job competency requirements, qualifications, and experience are extracted; based on a predefined dimension mapping rule library, the extracted keywords are associated with the corresponding resume screening dimensions to establish a structured keyword-dimensional mapping relationship table; S2: Job Demand Intensity Vector Generation: Based on the mapping table generated in S1, analyze the frequency of occurrence, syntactic position, modification intensity, and contextual semantic relevance of each keyword in the job description text; perform multi-dimensional semantic weighted evaluation on each resume screening dimension; normalize the evaluation results to generate a vector representing the demand intensity of the current job for each dimension; S3: Historical Weight Vector Extraction: Based on the industry attributes, job category, job level, and other metadata of the current position, retrieve the set of historical recruitment records with the highest matching degree from the historical recruitment database; call the pre-trained feature importance analysis model to deeply mine the statistical correlation between the features of each resume dimension in the historical records and the final recruitment results; generate a statistically significant historical weight vector based on the feature contribution index output by the model. S4: Adaptive generation of weight vector: Count the number of valid historical records retrieved in S3; calculate the smoothing coefficient based on the preset data volume threshold strategy; weight and merge the vector of demand intensity for each dimension of the current position generated in S2 with the historical weight vector generated in S3 according to the smoothing coefficient; perform normalization processing on the fusion result to generate a weight vector suitable for the current position screening; differentiated weight configurations can be generated for the same position type due to differences in job descriptions. S5: Intelligent Calculation and Missing Field Handling of Resumes Matching Degree: Batch acquisition of structured data of resumes to be screened; standardized scoring of each dimension of each resume based on a preset dimension scoring rule library; calculation of preliminary weighted score by combining the weight vector generated by S4 that is suitable for the current job selection; identification of missing screening dimension fields in each resume during the calculation process; retention of only the effective dimensions that actually exist in the resume and their corresponding weights, calculation of the sum of effective weights; normalization processing by dividing the preliminary weighted score by the sum of effective weights to generate the final matching degree score; and recording of the score details and calculation basis for each dimension. S6: Visualization of screening results: All resumes are automatically sorted from highest to lowest according to their final matching score, generating a structured list of screening results.

[0010] Furthermore, the S1 job description text input and semantic parsing also includes: The HR specialist inputs a plain text job description, and the system initiates a pre-defined natural language processing flow: It performs word segmentation, dividing the continuous text into meaningful word units; it performs stop word filtering, removing common words that do not contribute substantially to semantic analysis, and retaining substantive keywords related to job requirements, qualifications, and experience; it performs part-of-speech tagging on the filtered keywords, identifying the grammatical role of each keyword; and it performs semantic role recognition, analyzing the semantic role of each keyword in the sentence. Based on the above semantic analysis results, keywords and phrases related to job competency requirements, qualification conditions, and experience background are extracted; the extracted keywords are matched with a predefined dimension mapping rule library, which includes basic dimensions such as major, skills, internship, school, and award certificates, as well as a set of keywords or semantic patterns that can be associated under each dimension. The extracted keywords are compared with the keyword set or semantic patterns in the dimension mapping rule base to establish a structured keyword-dimension mapping relationship table. The matching process considers the semantic similarity, contextual relevance, and importance of keywords in job requirements. The above mapping relationship table is used as structured input for the subsequent generation of job requirement intensity vectors.

[0011] Furthermore, the generation of the job demand intensity vector in S2 also includes: Based on the structured keyword-dimension mapping table generated by S1, a set of keywords related to each resume screening dimension is extracted. Multi-dimensional semantic feature analysis is then performed on the keywords in each set: Frequency analysis: The number of times each keyword appears in the job description text is counted; Syntactic position analysis: The syntactic position of the keywords in the job description text is analyzed. The job description text is parsed according to sentence structure, identifying the keywords in different positions such as titles, core requirement paragraphs, and priority conditions; Modification intensity analysis: The modifiers of the keywords are analyzed, identifying adjectives, adverbs, and other modifiers before and after the keywords; Contextual semantic relevance analysis: The semantic relevance of the keywords in the context is analyzed, examining the semantic relevance between the keywords and the core responsibilities and competency requirements of the job. A multi-dimensional semantic weighted evaluation is performed on each resume screening dimension, and the analysis results of the above four dimensions are integrated into a comprehensive evaluation value; the comprehensive semantic strength of each dimension is calculated by weighted summation. The comprehensive semantic strength of all resume screening dimensions is normalized, and the strength values ​​of each dimension are adjusted to the range of [0,1], while the sum of the strength values ​​of all dimensions is 1; the normalized vector is the vector representing the strength of the current job requirements for each dimension.

[0012] Furthermore, the historical weight vector extraction in S3 also includes: Extract metadata about the current job position, including industry attributes, job category, and rank, as key filtering criteria for historical data retrieval. Based on the extracted metadata, perform a search in the historical recruitment database to filter out a set of historical recruitment records that highly match the current job position. The retrieval process follows a multi-dimensional matching principle: first, a primary filtering is performed based on industry attributes; second, a secondary filtering is performed based on job category; and finally, a tertiary filtering is performed based on rank. The retrieved historical recruitment records are input into a pre-trained feature importance analysis model. This model learns based on the correlation between historical hiring results and resume features by analyzing the statistical correlation between the resume features of each dimension in the historical records and the final hiring results. Specifically, the model analyzes the co-occurrence patterns of each dimension feature and hiring results in a large number of historical records. The feature contribution index output by the feature importance analysis model is used to objectively evaluate each resume screening dimension; Normalize the feature contribution index; normalization adjusts the contribution values ​​of all dimensions to the [0,1] interval, while ensuring that the total weight is 1; the normalized weight vector is the historical weight vector with statistical significance.

[0013] Furthermore, the adaptive generation of the weight vector in S4 includes: The number of valid historical recruitment records retrieved in S3 is counted and compared with a preset data volume threshold, which is a benchmark value set based on industry practices. Based on the comparison between the data volume and the threshold, a smoothing coefficient is calculated. When the number of historical records is lower than the threshold, it indicates that the amount of historical data is insufficient, and the smoothing coefficient is set to a value close to 0. When the number of historical records is higher than the threshold, it indicates that the amount of historical data is sufficient, and the smoothing coefficient is set to a value close to 1. The calculation of the smoothing coefficient is continuous and adaptive. Smoothing coefficient Where m is the number of valid historical recruitment records retrieved in S3, m0 is the preset data volume threshold, and β is the parameter for controlling the smoothness of the transition. The current job demand intensity vector generated by S2 and the historical weight vector generated by S3 are weighted and fused according to the calculated smoothing coefficient. The fusion process follows the following logic: the higher the smoothing coefficient, the greater the contribution of the historical weight; the lower the smoothing coefficient, the greater the contribution of the demand intensity. Normalization is performed on the fused weight vector; the normalization process adjusts the weight values ​​of each dimension to the range of [0,1], ensuring that the sum of the weights of all dimensions is 1, thus forming a standardized weight vector; Weight vector of the current position ,in Let i be the fusion weight. Let the i-th dimension be the historical weight vector. Let i be the i-th dimension of the demand intensity vector.

[0014] Furthermore, the intelligent calculation of resume matching degree and handling of missing fields in S5 include: Batch acquisition of structured data of resumes to be screened. This data has been parsed and key fields have been extracted to form a structured resume feature representation; Based on a pre-defined dimensional scoring rule library, each resume is scored in a standardized manner across its various dimensions. The dimensional scoring rule library contains the scoring criteria for each dimension. The standardized scores for each dimension of each resume are combined with the weight vector generated by S4 to calculate the preliminary weighted score; the preliminary weighted score is the sum of the products of each dimension score and its corresponding weight. During the calculation process, the missing screening dimension fields in each resume are identified; only the valid dimensions that actually exist in the resume and their corresponding weights are retained, and the sum of valid weights is calculated; the sum of valid weights refers to the sum of the weights of all existing dimensions, not the sum of the weights of all dimensions. The initial weighted score is normalized by dividing by the sum of effective weights to generate the final matching score; matching degree = ,in, This represents the standardized score of the resume in dimension i. This represents the weight value of dimension i, and valid indicates the set of dimensions in the resume that contain valid data. Record the detailed scores and calculation basis for each dimension of each resume, including the scores, weights, weighted scores, and normalization processes for each dimension; these records are used for subsequent result display and interpretation.

[0015] Furthermore, a resume screening and display system for human resource management is provided to implement the resume screening and display method for human resource management described above. The resume screening and display system for human resource management includes: a job description input and semantic parsing module, a job requirement intensity analysis module, a historical weight vector extraction and configuration module, a resume input and preprocessing module, and a screening result visualization module. Job Description Input and Semantic Parsing Module: Receives plain text job descriptions input by HR specialists, integrates a natural language processing engine, performs word segmentation, stop word filtering, part-of-speech tagging, and semantic role recognition on the job description text; extracts keywords and phrases related to job competency requirements, qualifications, and experience background, and, based on a predefined keyword-dimension mapping rule base, associates keywords with resume screening dimensions to generate a structured mapping relationship table; Job Demand Intensity Analysis Module: Based on the keyword-dimensional mapping table, this module analyzes the frequency, syntactic position, modification intensity, and contextual semantic relevance of keywords in job descriptions; it quantifies the demand intensity of each resume screening dimension and generates a normalized job demand intensity vector; and it visualizes the demand weight of each dimension in the current job. The historical weight vector extraction and configuration module retrieves highly relevant historical records from the historical recruitment database based on job industry, category, and job level metadata. It then uses a pre-trained feature importance analysis model to mine the statistical correlation between each resume dimension and hiring results in the historical records, generating a historical weight vector. The module counts the number of valid historical records, calculates a smoothing coefficient based on a data volume threshold, and weights and merges the job demand intensity vector with the historical weight vector to generate a weight vector specific to the current job. It supports comparing and displaying the current weight with historical weights and can generate differentiated weight configurations for different job descriptions. Resume Input and Preprocessing Module: Provides a file upload interface, supporting batch import of resumes in formats such as .txt, .docx, and .pdf; performs text parsing on resumes, extracts structured fields such as name, school, major, work skills, and open-source projects, and generates resume feature vectors; The screening results visualization module: Based on a preset dimension scoring rule library, it standardizes the scoring of each dimension of each resume; combined with the weight vector, it calculates the preliminary weighted score of the resume; it identifies the missing screening dimension fields in the resume, retains only the valid dimensions and their weights, calculates the sum of the valid weights, and normalizes the preliminary score by dividing the sum of the valid weights to generate the final matching score; it sorts all resumes in descending order of matching score and generates a structured list containing ranking, name, and matching percentage.

[0016] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a resume screening and display method for human resource management as described above.

[0017] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the resume screening and display method for human resource management as described above.

[0018] The beneficial effects of this invention are: by organically integrating deep semantic analysis of job description text with intelligent mining of historical recruitment data, a dynamic screening system that runs through the entire process is constructed, realizing a substantial leap from experience-driven to data-driven resume screening.

[0019] The smoothing coefficient calculation formula used differs from the traditional simple weight template switching that adjusts based on job characteristics. It is calculated based on the amount of historical data m, realizing data-driven dynamic weight adjustment. When the number of historical records is lower than the threshold, it indicates that the amount of historical data is insufficient, and the smoothing coefficient is set to a value close to 0; when the number of historical records is higher than the threshold, it indicates that the amount of historical data is sufficient, and the smoothing coefficient is set to a value close to 1. The β parameter allows companies to adjust the smoothness according to their own business characteristics. For example, companies with high data sensitivity can set a larger β value to make the transition steeper, while companies with high data stability requirements can set a smaller β value to make the transition smoother.

[0020] The generated dynamic weight vector not only accurately responds to the immediate semantic needs of job descriptions, but also scientifically incorporates the statistical patterns of historical recruitment data, making the screening criteria highly consistent with the actual staffing needs of the job, and improving the consistency between screening results and actual recruitment decisions. When the number of historical recruitment records m is small (m < m0), the smoothing coefficient is approximately 0, and the system mainly relies on the job demand intensity vector to avoid weight distribution distortion caused by insufficient historical data; when m is large (m > m0), the smoothing coefficient is approximately 1, and the system makes full use of the guiding value of historical data. By adaptively adjusting the smoothing coefficient, the system can maintain high matching accuracy even when the amount of data is insufficient; when the amount of historical data is sufficient, the system can more accurately use historical data to guide the matching.

[0021] The normalization formula constructed in the matching degree calculation stage is designed to perform weighted summation and normalization based only on the effective dimensions actually existing in the resume. This achieves dynamic normalization of weights, making the matching degree between different resumes comparable. It fundamentally avoids the matching degree distortion problem caused by the forced setting of missing fields to zero in traditional methods, and avoids the underestimation or overestimation of the matching degree due to the lack of data in certain dimensions of the resume. This ensures that the matching degree score truly reflects the matching quality of the information provided by the candidate, and maintains the fairness and objectivity of the screening process.

[0022] Overall, this invention enables a synergistic improvement in resume screening in terms of accuracy, fairness, efficiency, and explainability. It not only optimizes the quality of human resource allocation but also enhances the scientific nature and credibility of recruitment decisions through a transparent and traceable screening process, providing the field of human resource management with an intelligent solution that combines technological innovation with practical industrial value. Attached Figure Description

[0023] Figure 1 A flowchart illustrating a resume screening and display method for human resource management. Figure 2 A chart ranking the top 10 resumes based on job descriptions of the same position and filtered with preset fixed weights; Figure 3 A chart showing the top 10 resumes sorted based on job description text for the same position and dynamic weighting; Figure 4 A chart ranking the top 10 resumes based on changed job description text and filtered with the same historical weight; Figure 5 A chart ranking the top 10 resumes based on job descriptions for the same position, filtered by different historical weights. Detailed Implementation

[0024] The present invention will be further described clearly and completely below, but the scope of protection of the present invention is not limited thereto.

[0025] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0026] Example 1 A method and system for screening and displaying resumes in human resource management, the technical solution of which is as follows: S1: Job Description Text Input and Semantic Analysis: The HR specialist inputs the plain text of the job description; the integrated natural language processing engine performs word segmentation, stop word filtering, part-of-speech tagging, and semantic role recognition on the input job description text; keywords and phrases related to job competency requirements, qualifications, and experience are extracted; based on a predefined dimension mapping rule library, the extracted keywords are associated with the corresponding resume screening dimensions to establish a structured keyword-dimensional mapping relationship table; S2: Job Demand Intensity Vector Generation: Based on the mapping table generated in S1, analyze the frequency of occurrence, syntactic position, modification intensity, and contextual semantic relevance of each keyword in the job description text; perform multi-dimensional semantic weighted evaluation on each resume screening dimension; normalize the evaluation results to generate a vector representing the demand intensity of the current job for each dimension; S3: Historical Weight Vector Extraction: Based on the industry attributes, job category, job level, and other metadata of the current position, retrieve the set of historical recruitment records with the highest matching degree from the historical recruitment database; call the pre-trained feature importance analysis model to deeply mine the statistical correlation between the features of each resume dimension in the historical records and the final recruitment results; generate a statistically significant historical weight vector based on the feature contribution index output by the model. S4: Adaptive generation of weight vector: Count the number of valid historical records retrieved in S3; calculate the smoothing coefficient based on the preset data volume threshold strategy; weight and merge the vector of demand intensity for each dimension of the current position generated in S2 with the historical weight vector generated in S3 according to the smoothing coefficient; perform normalization processing on the fusion result to generate a weight vector suitable for the current position screening; differentiated weight configurations can be generated for the same position type due to differences in job descriptions. S5: Intelligent Calculation and Missing Field Handling of Resumes Matching Degree: Batch acquisition of structured data of resumes to be screened; standardized scoring of each dimension of each resume based on a preset dimension scoring rule library; calculation of preliminary weighted score by combining the weight vector generated by S4 that is suitable for the current job selection; identification of missing screening dimension fields in each resume during the calculation process; retention of only the effective dimensions that actually exist in the resume and their corresponding weights, calculation of the sum of effective weights; normalization processing by dividing the preliminary weighted score by the sum of effective weights to generate the final matching degree score; and recording of the score details and calculation basis for each dimension. S6: Visualization of screening results: All resumes are automatically sorted from highest to lowest according to their final matching score, generating a structured list of screening results.

[0027] refer to Figure 1The diagram shows a resume screening and display method and system flowchart for human resource management.

[0028] Furthermore, the S1 job description text input and semantic parsing also includes: The HR specialist inputs a plain text job description for an AI engineer position: Requirements include proficiency in Python, machine learning project experience, a degree from a top-tier university (985), and preference given to candidates with internships at major companies. The system then initiates a pre-defined natural language processing flow: using a mature Chinese word segmentation tool and a pre-built HR-related dictionary, the text is divided into word units. Punctuation marks and general stop words are removed according to the Chinese stop word list standard, but restrictive words such as "requirements" and "preference" are retained—as they play a crucial role in semantic role recognition. The output is: ["requirements", "Python", "proficient", "machine learning", "project experience", "985", "degree", "major company", "internship", "preference"]. Standard part-of-speech tagging is performed on the cleaned words, focusing on content words such as nouns, proper nouns, and verbs; based on existing semantic role tagging techniques such as dependency parsing, the patient role components of core predicates such as "requirement" and "need" are identified; in "requirement to be proficient in Python", "requirement" is the predicate and "proficient in Python" is the patient role, "Python" is extracted as the core keyword and "proficient" is the modifier.

[0029] The extracted keywords are matched with a predefined dimension mapping rule base, which includes basic dimensions such as major, skills, internship, school, and award certificates, as well as a set of keywords or semantic patterns that can be associated under each dimension. The rule base is pre-built by domain technical personnel and supports synonym expansion, such as programming ≡ program design, which is achieved through a pre-set thesaurus.

[0030] The extracted keywords are compared with the keyword set or semantic patterns in the dimension mapping rule base to establish a structured keyword-dimension mapping relationship table; For each candidate keyword, calculate the overall matching confidence score; Confidence score = semantic similarity × contextual relevance weight × importance weight; Semantic similarity score rules: exact match = 1.0, thesaurus match = 0.9, calculated value is used when word vector cosine similarity ≥ 0.7; Contextual relevance weight score rules: appearing after requirement / must = 1.0, having / possessing = 0.8, priority = 0.6; Importance weight score rules: title / first paragraph = 1.2, core requirement paragraph (first 2 / 3 of the job description text) = 1.0, supplementary paragraph at the end = 0.8; The preset confidence threshold is 0.6. If the keyword matches multiple dimensions, the one with the highest confidence score is used; when the confidence scores are the same, the priority is determined by dimension (skills > major > school > internship > award certificate).

[0031] The above mapping table is used as structured input, with fields including: keyword | mapping dimension | semantic similarity | contextual relevance weight | importance weight | confidence | starting character index | ending character index, for the subsequent generation of job requirement intensity vector.

[0032] Furthermore, the generation of the job demand intensity vector in S2 also includes: Based on the structured keyword-dimension mapping table generated by S1, a set of keywords related to each resume screening dimension is extracted. Multi-dimensional semantic feature analysis is then performed on the keywords in each set: For example, frequency analysis: the frequency score is min(keyword occurrences / 3, 1), where 3 is the industry experience value; syntactic position analysis: the syntactic position score is based on character index location, with the first 10% of characters in the title area scoring 1.5, core requirement paragraphs containing sentences like "requirement" and "must" scoring 1.2, priority condition sentences containing "priority" or "bonus" scoring 0.8, and others scoring 1.0; modification strength analysis: the modification strength score is based on a pre-defined modification strength table, with "must / requirement" scoring 1.5; "proficient / expert" scoring 1.2; "understand / familiar" scoring 0.8; "priority / bonus" scoring 0.6; and no modification scoring 1.0; contextual semantic relevance analysis: the job description text is segmented by keywords such as "job requirements" and "job responsibilities" to score contextual semantic relevance, with occurrences in the "job requirements" section scoring 1.0, occurrences only in the "job responsibilities" section scoring 0.7, and other positions scoring 0.5.

[0033] A multi-dimensional semantic weighted evaluation was conducted for each resume screening dimension, and the analysis results of the above four dimensions were integrated into a comprehensive evaluation value. The evaluation process considered the relative importance of each dimension and calculated the comprehensive semantic strength of each dimension by weighted summation. The weights of the four dimensions—frequency of occurrence, syntactic position, modification strength, and contextual semantic relevance—were 0.3, 0.3, 0.2, and 0.2, respectively. For example, when recruiting an AI engineer, the requirements are: proficiency in Python, machine learning project experience, a degree from a 985 university, and preference will be given to candidates with internships at major companies. The Python score would be S=0.3×0.333+0.3×1.2+0.2×1.2+0.2×1.0=0.90, and the score for internships at major companies would be S=0.3×0.333+0.3×0.8+0.2×1.0+0.2×1.0=0.74.

[0034] The comprehensive semantic strength of all resume screening dimensions is normalized, and the strength values ​​of each dimension are adjusted to the range of [0,1] to ensure that the strength values ​​of each dimension are comparable, while making the sum of the strength values ​​of all dimensions equal to 1. The normalized vector is the vector representing the demand strength of each dimension for the current position. This vector objectively reflects the relative importance of each dimension in the job description.

[0035] Furthermore, the historical weight vector extraction in S3 also includes: Extract metadata about the current job position, including industry attributes, job category, and rank, as key filtering conditions for historical data retrieval. Based on the extracted metadata, perform a search in the historical recruitment database to filter out historical recruitment records that highly match the current job position. The retrieval process follows a multi-dimensional matching principle: first, a primary filtering is performed based on industry attributes, with the matching rule being an exact string match; second, a secondary filtering is performed based on job category, with the matching rule being the first two digits of the job category code, using a predefined coding system RD=R&D, MK=Marketing; finally, a tertiary filtering is performed based on rank, with the matching rule being that the rank is within the range of [current rank - 1, current rank + 1]; records are only included in the record set if all three conditions are met; if the search results are empty, all dimensions of the historical weight vector are set to 0.

[0036] The retrieved historical recruitment records are input into a pre-trained feature importance analysis model. This model learns based on the correlation between historical hiring results and resume features. It is a pre-trained logistic regression model (binary classification: hired = 1, not hired = 0). By analyzing the statistical correlation between the features of each resume dimension in the historical records and the final hiring results, the model evaluates the actual contribution of each dimension to the hiring decision. Specifically, the model quantifies the influence of each dimension on the hiring decision in historical recruitment practices by analyzing the co-occurrence patterns of each dimension feature and hiring results in a large number of historical records. The input data consists of each resume in the retrieved historical records that has been rated by human resource managers, and the label is whether the resume was ultimately hired (0 / 1).

[0037] The feature contribution index output by the feature importance analysis model is used to objectively evaluate each resume screening dimension. The contribution index reflects the actual weight of the dimension in the hiring decision in historical recruitment. The higher the value, the more important the dimension is. The SHAP library is used to calculate the average absolute SHAP value of each dimension. To ensure the scientific validity and reliability of the historical weight vector, the feature contribution index is normalized. The normalization process adjusts the contribution values ​​of all dimensions to the [0,1] interval, making the weights of each dimension comparable, while ensuring that the sum of the weights is 1. The normalized weight vector is the historical weight vector with statistical significance, which objectively reflects the actual contribution of each dimension to the hiring decision in the historical recruitment data.

[0038] Furthermore, the adaptive generation of the weight vector in S4 includes: The number of valid historical job postings retrieved in S3 is counted. This number serves as an objective indicator of the richness of historical data. This number is compared with a preset data volume threshold, which is a benchmark value set based on industry practices to determine whether the historical data is sufficient to support the stability of the weight calculation. Based on the comparison results between data volume and threshold, a smoothing coefficient is calculated. m represents the number of valid historical recruitment records retrieved in S3, m0 is the preset data volume threshold, and β is 0.03 to control the smoothness of the transition. When the number of historical records is lower than the threshold, it indicates that the amount of historical data is insufficient, and the smoothing coefficient is set to a value close to 0. At this time, more emphasis is placed on the immediate needs of the job description text. When the number of historical records is higher than the threshold, it indicates that the amount of historical data is sufficient, and the smoothing coefficient is set to a value close to 1. At this time, more emphasis is placed on the stability of historical experience. The calculation of the smoothing coefficient is continuous and adaptive, ensuring a smooth transition in the weight generation process when the amount of data changes, and avoiding abrupt changes.

[0039] The demand intensity vector for each dimension of the current job generated in S2 is combined with the historical weight vector generated in S3, and then weighted and merged according to the calculated smoothing coefficient. ,in The fusion weights for dimension i are unnormalized. Let be the i-th dimension of the historical weight vector (already normalized). Let i be the i-th dimension of the demand intensity vector (already normalized). The fusion process follows the logic as follows: the higher the smoothing coefficient, the greater the contribution of historical weights; the lower the smoothing coefficient, the greater the contribution of demand intensity. The fused weight vector comprehensively reflects the dual influence of the job description's immediate needs and historical recruitment experience, avoiding the one-sidedness of relying solely on the job description or historical data.

[0040] To ensure the comparability and usability of the weight vectors, normalization is performed on the fused weight vectors. The normalization process adjusts the weight values ​​of each dimension to the range of [0,1], ensuring that the sum of the weights of all dimensions is 1, thus forming a standardized weight vector. The normalized weight vectors are generated entirely by data and have no preset fixed values.

[0041] Furthermore, the intelligent calculation of resume matching degree and handling of missing fields in S5 include: The system can acquire structured data of resumes to be screened in batches. This data has been parsed and key fields have been extracted to form a structured resume feature representation. No additional data processing is required from human resources specialists, ensuring the efficiency and accuracy of data acquisition. Based on a pre-defined dimensional scoring rule library, which is initialized and configured by the company's human resources administrator, the rule library standardizes the scoring of each dimension of each resume. The dimensional scoring rule library contains the scoring criteria for each dimension. For example, the scoring rules for the skills dimension are: ≥3=100, 2=80, 1=60, 0=0; the scoring rules for the school dimension are: 985=100, 211=80, Double First-Class=70, ordinary undergraduate=60; the scoring rules for the major dimension are: exact match=100, relevant=80, irrelevant=0; the scoring rules for the internship dimension are: relevant internship=100, irrelevant internship=60, none=0 (but missing internships are not included in the calculation); the scoring rules for the award certificate dimension are: national level=100, provincial level=80, school level=60, none=0.

[0042] The standardized scores for each dimension of each resume are combined with the weight vector generated by S4 to calculate the preliminary weighted score. The preliminary weighted score is the sum of the products of each dimension score and its corresponding weight, representing the overall performance of the resume in each dimension.

[0043] During the calculation process, the missing screening dimension fields in each resume are identified. Values ​​of empty strings / null / NaN are considered missing, values ​​of preset missing keywords such as none / not yet available are considered missing, and valid non-empty values ​​are considered present. Only the valid dimensions that actually exist in the resume and their corresponding weights are retained, and the sum of valid weights is calculated. The sum of valid weights refers to the sum of the weights of all existing dimensions, not the sum of the weights of all dimensions. The initial weighted score is normalized by dividing by the sum of effective weights to generate the final matching score. Normalization ensures the fairness of the matching score calculation, preventing missing fields from negatively impacting the contributions of other dimensions. Matching score = ,in, This represents the standardized score of the resume in dimension i. This represents the weight value of dimension i, and valid indicates the set of dimensions in the resume that contain valid data. This formula ensures that the matching degree calculation is based only on the information actually provided in the resume, and the result is strictly limited to a reasonable numerical range.

[0044] Record the detailed scores and calculation basis for each resume across all dimensions, including the scores, weights, weighted scores, and normalization processes for each dimension; these records are used for subsequent result presentation and interpretation, making the screening process transparent and traceable.

[0045] Furthermore, a resume screening and display system for human resource management is provided to implement the resume screening and display method for human resource management described above. The resume screening and display system for human resource management includes: a job description input and semantic parsing module, a job requirement intensity analysis module, a historical weight vector extraction and configuration module, a resume input and preprocessing module, and a screening result visualization module. Job Description Input and Semantic Parsing Module: Receives plain text job descriptions input by HR specialists, integrates a natural language processing engine, performs word segmentation, stop word filtering, part-of-speech tagging, and semantic role recognition on the job description text; extracts keywords and phrases related to job competency requirements, qualifications, and experience background, and, based on a predefined keyword-dimension mapping rule base, associates keywords with resume screening dimensions to generate a structured mapping relationship table; Job Demand Intensity Analysis Module: Based on the keyword-dimensional mapping table, this module analyzes the frequency, syntactic position, modification intensity, and contextual semantic relevance of keywords in job descriptions; it quantifies the demand intensity of each resume screening dimension and generates a normalized job demand intensity vector; and it visualizes the demand weight of each dimension in the current job. The historical weight vector extraction and configuration module retrieves highly relevant historical records from the historical recruitment database based on job industry, category, and job level metadata. It then uses a pre-trained feature importance analysis model to mine the statistical correlation between each resume dimension and hiring results in the historical records, generating a historical weight vector. The module counts the number of valid historical records, calculates a smoothing coefficient based on a data volume threshold, and weights and merges the job demand intensity vector with the historical weight vector to generate a weight vector specific to the current job. It supports comparing and displaying the current weight with historical weights and can generate differentiated weight configurations for different job descriptions. Resume Input and Preprocessing Module: Provides a file upload interface, supporting batch import of resumes in formats such as .txt, .docx, and .pdf; performs text parsing on resumes, extracts structured fields such as name, school, major, work skills, and open-source projects, and generates resume feature vectors; The screening results visualization module: Based on a preset dimension scoring rule library, it standardizes the scoring of each dimension of each resume; combined with the weight vector, it calculates the preliminary weighted score of the resume; it identifies the missing screening dimension fields in the resume, retains only the valid dimensions and their weights, calculates the sum of the valid weights, and normalizes the preliminary score by dividing the sum of the valid weights to generate the final matching score; it sorts all resumes in descending order of matching score and generates a structured list containing ranking, name, and matching percentage.

[0046] Example 2 This embodiment sorts resumes according to the initially preset weights. For the same job title and the same JD input text, it outputs a sorting table of resumes based on matching degree.

[0047] The HR manager has received 48 resumes. These 48 resumes have been sorted according to the job requirements and the HR manager's subjective preferences. The sorted resume list is shown in the table below.

[0048] Ranking Name Ranking Name Ranking Name Ranking Name 1 Lin Wei 13 Lin Yuxin 25 Zheng Haoran 37 Wu Zihao 2 Wang Hao 14 Zhang Muyang 26 Ma Xiaoqi 38 Zheng Yutong 3 Li Xue 15 Zhao Yichen 27 Wang Zixuan 39 Ma Mingyu 4 Chen Siyuan 16 Su Haoran 28 Chen Siqi 40 Wang Haoyu 5 Zhao Siyu 17 Xia Ruotong 29 Li Yuchen 41 Li Yuxin 6 Ma Yuchen 18 Lu Zeyu 30 Zhang Sihan 42 Zhang Zixuan 7 Gao Yutong 19 Lin Zihao 31 Zhao Mingyu 43 Zhao Haoyu 8 Xia Zeyu 20 Zhao Sihan 32 Sun Yuxin 44 Sun Ziqi 9 Su Wanqing 21 Wang Yuxuan 33 Lin Haoyu 45 Lin Siyu 10 Lu Junhao 22 Liu Siyu 34 Gao Ziqi 46 Gao Mingxuan 11 Ye Zixuan 23 Zhou Mingyu 35 Huang Mingxuan 47 Chen Zhe 12 Chen Junyu 24 Wu Yutong 36 Zhou Siyu 48 Liu Yang .

[0049] The HR manager entered the job description text as follows: Recruiting AI algorithm engineers, requiring proficiency in Python and TensorFlow, experience in machine learning projects, graduates from 985 universities preferred, and those with internship experience in major companies preferred.

[0050] The job requirement intensity analysis is as follows: internship experience accounts for 34.8%, work skills account for 30.4%, major accounts for 17.4%, school accounts for 8.7%, open source projects account for 8.7%, and awards / certificates account for 0%. The key requirements are: major: algorithm, machine learning, learning; school: 985; internship experience: project, experience, big company, internship; work skills: algorithm, proficient in Python, TensorFlow; open source projects: project.

[0051] Based on fixed weights: major weight 25%, school weight 15%, internship experience weight 15%, work skills weight 35%, open source projects weight 5%, and awards and certificates weight 5%.

[0052] Based on the scores of each of these 48 resumes across various dimensions, and combined with the matching degree calculation formula, the matching degree is calculated as follows: Matching Degree = The resulting resume ranking list is shown in the table below.

[0053] Ranking Name Match Ranking Name Match Ranking Name Match 1 Wang Yuxuan 85.25% 17 Ma Yuchen 70.75% 33 Chen Siqi 60.75% 2 Chen Junyu 81.75% 18 Li Yuchen 70.75% 34 Xia Zeyu 60% 3 Lu Junhao 79.75% 19 Zhao Yichen 70.25% 35 Zhang Muyang 60% 4 Su Wanqing 78.25% 20 Su Haoran 69.25% 36 Xia Ruotong 60% 5 Liu Yang 77.9% 21 Zhao Haoyu 67.75% 37 Zhao Sihan 57.06% 6 Zhao Siyu 77.75% 22 Ye Zixuan 67.65% 38 Lin Zihao 56.25% 7 Lin Wei 75.75% 23 Zhou Siyu 67.25% 39 Wang Zixuan 54.71% 8 Liu Siyu 75.29% 24 Zheng Yutong 67.25% 40 Zhang Sihan 54.71% 9 Wang Hao 74.75% 25 Wu Yutong 67.06% 41 Li Yuxin 54.71% 10 Zhou Mingyu 74.75% 26 Ma Xiaoqi 67.06% 42 Gao Ziqi 54.12% 11 Zhao Mingyu 74.25% 27 Gao Yutong 66.75% 43 Zheng Haoran 51.75% 12 Lin Siyu 74.25% 28 Zhang Zixuan 63.75% 44 Huang Mingxuan 51.75% 13 Li Xue 72.35% 29 Lin Yuxin 63.53% 45 Lin Haoyu 50.59% 14 Chen Zhe 72% 30 Sun Yuxin 62.94% 46 Wu Zihao 50.59% 15 Lu Zeyu 71.76% 31 Ma Mingyu 62.94% 47 Wang Haoyu 50.59% 16 Chen Siyuan 71.25% 32 Gao Mingxuan 62.94% 48 Sun Ziqi 50% .

[0054] refer to Figure 2 The chart shown is a top 10 ranking of resumes based on job descriptions for the same position and a preset fixed weight filter.

[0055] Example 2 illustrates a typical shortcoming of traditional resume screening systems: they employ preset fixed weights (major 25%, school 15%, internship experience 15%, work skills 35%, open-source projects 5%, awards and certificates 5%). With the same job description input, the system outputs a fixed order, but fails to adapt to the semantic details of the job description. In Example 2, fixed weights ensure that for the same job description text, ranking fluctuations caused by HR fatigue or emotions are avoided, such as the order in subjective rankings changing due to daily status changes; all resumes are scored according to uniform rules to avoid implicit discrimination caused by subjective preferences such as prestige school complex or gender implications; limitations: the weights are out of sync with the actual needs of the job description, and it is impossible to dynamically generate differentiated weights based on the semantic details of the job description text (e.g., using the same weight for technical positions and campus recruitment positions), resulting in a deviation between the screening results and the actual hiring needs of enterprises.

[0056] Example 3 This embodiment performs a sorting based on the dynamic weights set by the present invention. For the same job posting and the same JD input text, it outputs a sorting table of resumes based on matching degree.

[0057] The HR manager entered the job description text as follows: Recruiting AI algorithm engineers, requiring proficiency in Python and TensorFlow, experience in machine learning projects, graduates from 985 universities preferred, and those with internship experience in major companies preferred.

[0058] The analysis of job demand intensity is as follows: internship experience accounts for 34.8%, work skills account for 30.4%, major accounts for 17.4%, school accounts for 8.7%, open source projects account for 8.7%, and awards and certificates account for 0%.

[0059] based on Where m is 20, m0 is 30, and α is calculated to be 0.27; based on The historical weights for each subject are as follows: major (0.148), school (0.294), internship experience (0.0147), work skills (0.263), open-source projects (0.0203), and awards / certificates (0.256).

[0060] The calculated weights are as follows: major weight 16.7%, school weight 14.3%, internship experience weight 25.8%, work skills weight 29.3%, open source projects weight 6.9%, and awards and certificates weight 6.9%.

[0061] Based on the scores of each of these 48 resumes across various dimensions, and combined with the matching degree calculation formula, the matching degree is calculated as follows: Matching Degree = The resulting resume ranking list is shown in the table below.

[0062] Ranking Name Match Ranking Name Match Ranking Name Match 1 Wang Yuxuan 84.46% 17 Su Haoran 71.64% 33 Sun Yuxin 61.01% 2 Chen Junyu 82.12% 18 Li Xue 71.6% 34 Ma Mingyu 61.01% 3 Lu Junhao 81.27% 19 Zhao Yichen 71.34% 35 Gao Mingxuan 61.01% 4 Su Wanqing 79.19% 20 Lu Zeyu 69.86% 36 Zhang Muyang 60.71% 5 Zhao Siyu 78.5% 21 Zhao Haoyu 69.8% 37 Xia Ruotong 58.93% 6 Lin Wei 77.65% 22 Chen Zhe 69.4% 38 Zhao Sihan 58.64% 7 Wang Hao 75.66% 23 Zhou Siyu 69.1% 39 Zheng Haoran 58.54% 8 Zhou Mingyu 75.66% 24 Zheng Yutong 69.1% 40 Huang Mingxuan 58.54% 9 Chen Siyuan 75.01% 25 Zhang Zixuan 66.17% 41 Wang Zixuan 53.1% 10 Zhao Mingyu 74.97% 26 Ye Zixuan 65.91% 42 Zhang Sihan 53.1% 11 Lin Siyu 74.97% 27 Ma Xiaoqi 65.76% 43 Li Yuxin 53.1% 12 Liu Yang 73.66% 28 Wu Yutong 64.97% 44 Gao Ziqi 52.16% 13 Liu Siyu 72.88% 29 Chen Siqi 63.93% 45 Lin Haoyu 49.14% 14 Gao Yutong 72.25% 30 Lin Zihao 62.86% 46 Wu Zihao 49.14% 15 Ma Yuchen 72.04% 31 Lin Yuxin 61.95% 47 Wang Haoyu 49.14% 16 Li Yuchen 72.04% 32 Xia Zeyu 61.51% 48 Sun Ziqi 48.21% .

[0063] refer to Figure 3 The chart shown is a top 10 ranking of resumes based on job description text for the same position and dynamic weight filtering.

[0064] In Example 3, the weights are generated by semantic parsing of the job description text, extracting keywords through a natural language processing engine, analyzing frequency, syntactic position, modification intensity, etc., and calculating the fusion weights based on the smoothing coefficient, so that the weights are highly consistent with the core requirements of the job. However, the impact of historical data on weight allocation was not considered, and the historical data mining was still superficial. When the amount of historical data was insufficient, the weight mainly relied on the job demand intensity vector, which did not make full use of the statistical regularity of historical data. While maintaining objectivity, the system achieved a preliminary integration of machine efficiency and artificial intelligence.

[0065] Example 4 This embodiment performs a sorting based on the dynamic weight calculation formula set by the present invention. For the same job, different JD input texts are changed, and a sorting table of resumes based on matching degree is output.

[0066] The HR manager entered the job description text as follows: Recruiting AI algorithm engineers, requiring proficiency in Python programming and the PyTorch deep learning framework, solid practical experience in machine learning projects; Master's degree or above, with preference given to candidates from Double First-Class universities; those with internship or relevant project experience at leading internet companies will be given priority.

[0067] The analysis of job demand intensity is as follows: internship experience accounts for 35.48%, work skills account for 30.96%, major accounts for 14.12%, school accounts for 10.63%, open source projects account for 8.85%, and awards and certificates account for 0.12%.

[0068] based on Where m is 20, m0 is 30, and α is calculated to be 0.27; based on The historical weights for each category are as follows: major (0.148), school (0.294), internship experience (0.0147), work skills (0.263), open-source projects (0.02), and awards / certificates (0.256).

[0069] The calculated weights are as follows: major weight 14.3%, school weight 15.7%, internship experience weight 26.3%, work skills weight 29.7%, open source projects weight 7%, and awards and certificates weight 6.9%.

[0070] Based on the scores of each of these 48 resumes across various dimensions, and combined with the matching degree calculation formula, the matching degree is calculated as follows: Matching Degree = The resulting resume ranking list is shown in the table below.

[0071] Ranking Name Match Ranking Name Match Ranking Name Match 1 Wang Yuxuan 83.66% 17 Li Yuchen 71.08% 33 Lin Yuxin 60.38% 2 Chen Junyu 81.91% 18 Li Xue 71.04% 34 Sun Yuxin 59.45% 3 Lu Junhao 81.02% 19 Zhao Yichen 70.38% 35 Ma Mingyu 59.45% 4 Su Wanqing 78.93% 20 Chen Zhe 68.85% 36 Gao Mingxuan 59.45% 5 Zhao Siyu 78.23% 21 Zhao Haoyu 68.8% 37 Zhao Sihan 58.76% 6 Lin Wei 77.34% 22 Lu Zeyu 68.45% 38 Zheng Haoran 58.68% 7 Chen Siyuan 75.19% 23 Zhou Siyu 68.11% 39 Huang Mingxuan 58.68% 8 Wang Hao 74.75% 24 Zheng Yutong 68.11% 40 Xia Ruotong 57.3% 9 Zhou Mingyu 74.75% 25 Ma Xiaoqi 65.13% 41 Wang Zixuan 51.39% 10 Zhao Mingyu 74.05% 26 Zhang Zixuan 65.13% 42 Zhang Sihan 51.39% 11 Lin Siyu 74.05% 27 Ye Zixuan 64.42% 43 Li Yuxin 51.39% 12 Gao Yutong 73.17% 28 Wu Yutong 63.49% 44 Gao Ziqi 50.44% 13 Liu Yang 72.68% 29 Lin Zihao 63.05% 45 Lin Haoyu 47.35% 14 Liu Siyu 71.55% 30 Chen Siqi 62.86% 46 Wu Zihao 47.35% 15 Su Haoran 71.19% 31 Xia Zeyu 62.55% 47 Wang Haoyu 47.35% 16 Ma Yuchen 71.08% 32 Zhang Muyang 60.92% 48 Sun Ziqi 46.41% .

[0072] refer to Figure 4 The chart shown is a top 10 ranking of resumes based on changed job description text and the same historical weighting.

[0073] In Example 4, when a new description is added to the job description text, a targeted ranking is generated. However, subjective ranking only applies to the original job description text. When new requirements arise, HR needs to spend time re-sorting, and inconsistencies in standards can easily occur due to memory bias. When a new description is added to the job description text, the system outputs a new ranking immediately, avoiding distortion of screening standards due to HR misunderstandings. This scenario demonstrates dynamic screening capabilities, eliminating the need for HR specialists to readjust weights. The system automatically adapts to changes in the job description, directly addressing the business pain point of frequent adjustments to job description texts in recruitment. However, it still does not solve the problem of the impact of historical data volume on weight allocation, and cannot automatically adjust weights when the historical data volume changes.

[0074] Example 5 This embodiment performs a sorting based on the dynamic weights set in this invention. These dynamic weights are updated according to historical data. For the same JD input text, a sorting table of resumes based on matching degree is output.

[0075] The HR manager entered the job description text as follows: Recruiting AI algorithm engineers, requiring proficiency in Python and TensorFlow, experience in machine learning projects, graduates from 985 universities preferred, and those with internship experience in major companies preferred.

[0076] The analysis of job demand intensity is as follows: internship experience accounts for 34.8%, work skills account for 30.4%, major accounts for 17.4%, school accounts for 8.7%, open source projects account for 8.7%, and awards and certificates account for 0%.

[0077] based on Where m is 70, m0 is 30, and α is calculated to be 0.98; based on The historical weights for each subject are as follows: major (0.146), school (0.296), internship experience (0.0153), work skills (0.263), open-source projects (0.0206), and awards / certificates (0.259).

[0078] The calculated weights are as follows: major weight 14.7%, school weight 29.2%, internship experience weight 2.2%, work skills weight 26.4%, open source project weight 2.2%, and awards and certificates weight 25.4%.

[0079] Based on the scores of each of these 48 resumes across various dimensions, and combined with the matching degree calculation formula, the matching degree is calculated as follows: Matching Degree = The resulting resume ranking list is shown in the table below.

[0080] Ranking Name Match Ranking Name Match Ranking Name Match 1 Lu Junhao 83.49% 17 Wang Hao 66.47% 33 Lin Zihao 58.92% 2 Lin Wei 80.63% 18 Zhou Mingyu 66.47% 34 Zhou Siyu 58.65% 3 Li Xue 80.31% 19 Lin Yuxin 65.4% 35 Zheng Yutong 58.65% 4 Chen Junyu 78.52% 20 Zhao Mingyu 63.93% 36 Zhang Muyang 58.44% 5 Chen Siyuan 77.91% 21 Lin Siyu 63.93% 37 Zhang Zixuan 56.01% 6 Gao Yutong 75.93% 22 Zhao Sihan 63.29% 38 Chen Siqi 55.91% 7 Su Wanqing 75.88% 23 Wu Yutong 63.14% 39 Wang Zixuan 55.04% 8 Zhao Siyu 75.66% 24 Xia Ruotong 62.93% 40 Zhang Sihan 55.04% 9 Chen Zhe 74.74% 25 Ma Yuchen 61.29% 41 Li Yuxin 55.04% 10 Wang Yuxuan 74.4% 26 Li Yuchen 61.29% 42 Gao Ziqi 52.45% 11 Liu Yang 72.05% 27 Zhao Haoyu 61.19% 43 Lin Haoyu 52.34% 12 Lu Zeyu 70.8% 28 Zhao Yichen 61.07% 44 Wu Zihao 52.34% 13 Ma Xiaoqi 70.06% 29 Su Haoran 61% 45 Wang Haoyu 52.34% 14 Liu Siyu 68.54% 30 Sun Yuxin 60.44% 46 Zheng Haoran 51.2% 15 Ye Zixuan 68.1% 31 Ma Mingyu 60.44% 47 Huang Mingxuan 51.2% 16 Xia Zeyu 67.73% 32 Gao Mingxuan 60.44% 48 Sun Ziqi 49.75% .

[0081] refer to Figure 5 The chart shown is a top 10 ranking of resumes based on job description text for the same position, filtered by different historical weights.

[0082] In Example 5, with the same job description input, the amount of historical data increased from m=20 to m=70. The system changed from mainly relying on the job demand intensity vector to mainly relying on the historical weight vector, tending to reflect the actual hiring patterns of enterprises in the historical data. The weighting system integrates job requirements with historical hiring patterns that implicitly reflect HR's true preferences, avoiding discrepancies between personal preferences and corporate standards in subjective rankings. For example, if an HR professional overemphasizes award certificates, but historical data shows that their contribution is only 2%, the system may be flawed. In subjective ranking, HR may subjectively deduct points due to missing fields in resumes, leading to the misscreening of high-quality candidates. This embodiment ensures that the comparison is based only on valid information, and the fairness is quantitatively verified. When historical data is insufficient, its weight is automatically reduced to avoid the excessive influence of recent case experience on the judgment in subjective ranking. Example 5 demonstrates that for the same job type, the weights should differ when the amount of historical data varies; the amount of historical data is a key factor affecting weight allocation, and the system should adjust the weights according to the amount of data.

[0083] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0084] All formulas in this invention are dimensionless and calculated by taking their numerical values. Dimensionlessness can be achieved through various methods such as standardization, which will not be elaborated here. The formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas can be set by those skilled in the art according to the actual situation.

[0085] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A resume screening display method for human resource management, characterized by, include: S1: Job Description Text Input and Semantic Parsing: Obtain the job description text, extract keywords related to the job requirements through natural language processing, associate the keywords with the resume screening dimensions according to a predefined dimension mapping rule library, and generate a keyword-dimensional mapping relationship table; S2: Job Demand Intensity Vector Generation: Based on the keyword-dimensional mapping table, perform multi-dimensional semantic weighted evaluation on each resume screening dimension to generate a job demand intensity vector; S3: Historical weight vector extraction: Based on the meta-information of the current position, retrieve matching historical recruitment records from the historical recruitment database, call the pre-trained feature importance analysis model to analyze the correlation between resume dimension features and hiring results in the historical recruitment records, and generate historical weight vectors; S4: Adaptive generation of weight vector: Count the number of valid historical records retrieved in S3, calculate the difference between the number of valid historical records and the preset threshold, multiply the difference by the negative value of the smoothness parameter, perform an exponential function operation with base e on the negative value, add 1 to the operation value, and take the reciprocal of the result as the final smoothing coefficient. The value of the smoothing coefficient depends on the number of valid historical records and the preset threshold. The job demand intensity vector generated in S2 and the historical weight vector generated in S3 are weighted and fused according to the smoothing coefficient. The weighting coefficient of the historical weight vector is the smoothing coefficient, and the job demand intensity vector is 1 minus the value of the smoothing coefficient. The fusion result is then normalized to generate the weight vector of the current job. When the number of valid historical records is less than a preset threshold, the weight vector of the current position depends on the demand intensity vector; when the number of valid historical records is greater than the preset threshold, the weight vector of the current position depends on the historical weight vector. S5: Obtain the structured data of the resumes to be screened, score each dimension according to the preset dimension scoring rules, and calculate the weighted score based on the scores and the weight vector of the current position as described in S4. During the calculation process, missing screening dimensions in the resume are identified, and the sum of effective weights is calculated based only on the existing effective dimensions and their corresponding weights. The weighted score is then divided by the sum of effective weights for normalization to generate a matching score. S6: Visualization of screening results: Sort all resumes in descending order of the matching score and generate a list of screening results.

2. The resume screening display method for human resource management according to claim 1, wherein The S1 job description text input and semantic parsing also includes: The job description text is processed sequentially by word segmentation, stop word filtering, part-of-speech tagging, and semantic role recognition to extract keywords and phrases related to job competency requirements, qualification requirements, and experience background. The keywords and phrases are matched with a predefined dimension mapping rule library, which contains multiple resume screening dimensions and keyword sets or semantic patterns corresponding to each dimension. The matching process is based on the semantic similarity and contextual relevance of the keywords. Generate the keyword-dimension mapping table based on the matching results.

3. The resume screening display method for human resource management according to Claim 1, wherein The generation of the job demand intensity vector in S2 also includes: Based on the keyword-dimension mapping table, for each resume screening dimension, multi-dimensional semantic feature analysis is performed on the associated keywords. The multi-dimensional semantic feature analysis includes: counting the frequency of keywords in the job description text, analyzing the syntactic position of keywords in the text, analyzing the modifier strength of keywords, and analyzing the contextual semantic relevance of keywords. Based on the results of the multi-dimensional semantic feature analysis, a weighted calculation is performed on each resume screening dimension to generate a comprehensive semantic strength value for each dimension. The comprehensive semantic strength values ​​of each dimension are normalized so that the sum of the normalized strength values ​​of each dimension is 1, thereby generating the job requirement strength vector.

4. The resume screening and display method for human resource management as described in claim 1, characterized in that, The historical weight vector extraction in S3 also includes: Extract the metadata of the current position, which includes industry attributes, position category, and job level; Based on the metadata, historical recruitment records are retrieved from the historical recruitment database according to a multi-level matching strategy, which includes filtering based on industry attributes, job categories, and job levels in sequence. The retrieved historical recruitment records are input into a pre-trained feature importance analysis model. The model analyzes the correlation between the features of each resume dimension and the hiring results, and outputs the feature contribution index of each dimension. The feature contribution index is normalized to generate the historical weight vector.

5. The resume screening and display method for human resource management as described in claim 1, characterized in that, The adaptive generation of the weight vector in S4 includes: The number of valid historical recruitment records retrieved in S3 is counted; the number of valid historical records is compared with a preset threshold, and a smoothing coefficient is calculated based on the comparison result; The smoothness coefficient Where m is the number of valid historical recruitment records retrieved in S3, m0 is the preset data volume threshold, and β is the parameter for controlling the smoothness of the transition. The job demand intensity vector generated in S2 and the historical weight vector generated in S3 are weighted and fused according to the smoothing coefficient; the weighted and fused vector is normalized to generate the weight vector of the current job. The weight vector of the current position ,in Let i be the fusion weight. Let the i-th dimension be the historical weight vector. Let i be the i-th dimension of the demand intensity vector.

6. The resume screening and display method for human resource management as described in claim 1, characterized in that, The intelligent calculation of resume matching degree and handling of missing fields in S5 include: Batch acquire structured data of resumes to be screened; standardize the scoring of each dimension of each resume according to a preset dimension scoring rule library; calculate the sum of the products of each dimension score and its corresponding weight as a preliminary weighted score based on the standardized score and the weight vector of the current position. Identify the missing screening dimensions in each resume, determine the effective dimension set, and calculate the sum of the weights of each dimension in the effective dimension set as the effective weight sum; divide the preliminary weighted score by the effective weight sum to generate a matching score; matching score = ,in, This represents the standardized score of the resume in dimension i. This represents the weight value of dimension i, and valid indicates the set of dimensions in the resume that contain valid data; it records the score, weight, weighted score and normalization calculation basis for each dimension of each resume.

7. A resume screening and display system for human resource management, characterized in that, To implement the resume screening and display method for human resource management as described in any one of claims 1-6, the resume screening and display system for human resource management includes: a job description input and semantic parsing module, a job demand intensity analysis module, a historical weight vector extraction and configuration module, a resume input and preprocessing module, and a screening result visualization module. Job Description Input and Semantic Parsing Module: Performs natural language processing on the input job description text, extracts keywords and phrases related to the job requirements, and generates a keyword-dimension mapping relationship table based on a predefined dimension mapping rule library; Job Demand Intensity Analysis Module: Based on the keyword-dimensional mapping relationship table, multi-dimensional semantic feature analysis and weighted evaluation are performed on each resume screening dimension to generate a normalized job demand intensity vector; The historical weight vector extraction and configuration module retrieves matching historical recruitment records from the historical recruitment database based on the meta-information of the current position, calls a pre-trained feature importance analysis model to generate a historical weight vector, calculates a smoothing coefficient based on the number of valid historical records and a preset threshold, and normalizes the weight vector of the current position after weighted fusion with the historical weight vector to generate the weight vector of the current position. Resume Input and Preprocessing Module: Acquires structured data of resumes to be screened; The screening results visualization module: It standardizes the scoring of each dimension of the resume according to the preset dimension scoring rule library, calculates the matching degree score by combining the weight vector of the current position, sorts the resumes according to the matching degree score, and generates a screening results list.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps in the resume screening and display method for human resource management as described in any one of claims 1-6.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps in the resume screening and display method for human resource management as described in any one of claims 1-6.

Images