A resume matching method and apparatus

Abstract

This application provides a resume matching method and apparatus, applicable to the field of data processing technology. The method includes: analyzing and mining unstructured resume data, semi-structured resume data, structured resume data, unstructured job requirement data, semi-structured job requirement data, and structured job requirement data to obtain implicit information from the resumes and job requirements; and generating resume matching information based on resume matching calculation parameters, resume matching parameter optimization step size, and the implicit information from the various resume data sets (unstructured, semi-structured, and structured), thereby improving recruitment efficiency and effectiveness. This application effectively improves the accuracy of resume-job requirement matching, enhances the reliability and rationality of resume matching, and thus improves recruitment efficiency and effectiveness.

Description

Technical Field

[0001] This application belongs to the field of data processing technology, and in particular relates to a resume matching method and apparatus. Background Technology

[0002] In today's digital age, the recruitment field is undergoing profound changes thanks to the rapid development of artificial intelligence and big data technologies. The growing demand from both businesses and job seekers for efficient and accurate resume matching is driving continuous iteration and updates in resume matching technology.

[0003] Existing technologies typically utilize Natural Language Processing (NLP) techniques to extract keywords from unstructured text in job application data and job requirement data, and then screen resumes based on the degree of keyword matching between the resume and the job description.

[0004] However, existing technologies rely solely on keyword matching and preset rules, making it difficult to delve into the semantic connections behind the text. The matching analysis has a single dimension, making it difficult to accurately interpret the experience descriptions of core competencies and implicit experiences in job resume data. It also cannot fully understand the deeper meaning of the job requirements in job demand data, leading to the omission or misjudgment of key information. Furthermore, it is difficult to adapt to the personalized needs of different industries and companies in the face of complex and ever-changing resume formats and diverse job requirements, resulting in poor accuracy and reliability of matching results. Summary of the Invention

[0005] In view of this, embodiments of this application provide a resume matching method and apparatus, which aim to solve the problems in the prior art, such as insufficient semantic understanding, inadequate integration of multiple types of data, and difficulty in adapting to complex and ever-changing resume formats and diverse job requirements, which make it difficult to improve the accuracy of resume matching.

[0006] The first aspect of this application provides a resume matching method, including:

[0007] Acquire job application resume data and job requirement data; the job application resume data includes unstructured job application resume data, semi-structured job application resume data, and structured job application resume data; the job requirement data includes unstructured job requirement data, semi-structured job requirement data, and structured job requirement data;

[0008] Based on multiple preset resume data semantic parsing vectors, semantic parsing processing is performed on the unstructured resume data, semi-structured resume data, and structured resume data to obtain the implicit information of the resume.

[0009] Semantic mining and analysis are performed on the unstructured, semi-structured, and structured job requirement data to obtain implicit job requirement information.

[0010] The unstructured resume data, semi-structured resume data, structured resume data, and implicit information of the resumes are aligned to obtain aligned resume data.

[0011] The unstructured job requirement data, semi-structured job requirement data, structured job requirement data, and implicit job requirement information are aligned to obtain aligned job requirement data.

[0012] Based on multiple randomly generated resume matching calculation parameters and multiple preset resume matching parameter optimization step sizes, the job resume alignment data and job requirement alignment data are matched and calculated to generate resume matching information.

[0013] A second aspect of this application provides a resume matching device, comprising:

[0014] The job application data and job requirement data acquisition module is used to acquire job application data and job requirement data; the job application data includes unstructured job application data, semi-structured job application data, and structured job application data; the job requirement data includes unstructured job requirement data, semi-structured job requirement data, and structured job requirement data.

[0015] The job resume implicit information generation module is used to perform semantic parsing processing on the unstructured resume data, semi-structured resume data and structured resume data based on multiple preset resume data semantic parsing vectors to obtain the job resume implicit information.

[0016] The job requirement implicit information generation module is used to perform semantic mining and analysis on the unstructured job requirement data, semi-structured job requirement data, and structured job requirement data to obtain the job requirement implicit information.

[0017] The job resume alignment data generation module is used to perform resume data alignment processing on the unstructured resume data, semi-structured resume data, structured resume data, and implicit information of the job resume to obtain job resume alignment data.

[0018] The job requirement alignment data generation module is used to perform job requirement alignment processing on the unstructured job requirement data, semi-structured job requirement data, structured job requirement data, and implicit job requirement information to obtain job requirement alignment data.

[0019] The resume matching information generation module is used to perform matching calculations on the job application resume alignment data and job requirement alignment data based on multiple randomly generated resume matching calculation parameters and multiple preset resume matching parameter optimization step sizes, and generate resume matching information.

[0020] A third aspect of this application provides a terminal device, the terminal device including a memory and a processor, the memory storing a computer program executable on the processor, the processor executing the computer program to implement the steps of the resume matching method as described in the first aspect above.

[0021] A fourth aspect of this application provides a computer-readable storage medium, comprising: storing a computer program, wherein when executed by a processor, the computer program implements the steps of the resume matching method described in the first aspect above.

[0022] The beneficial effects of this application's embodiments compared to existing technologies are as follows: This application mines implicit information from unstructured, semi-structured, and structured resume data that cannot be directly observed. Simultaneously, it extracts implicit information about job requirements from unstructured, semi-structured, and structured job requirement data. This breaks through the limitations of traditional matching methods that rely solely on explicit information. Alignment processing reduces comparison errors caused by differences in data formats. Furthermore, by optimizing matching parameters, it performs refined matching calculations on aligned resume data and aligned job requirement data. By fully combining explicit and implicit information, it adapts to the complex and ever-changing resume formats and diverse job requirements in the job market, effectively improving the matching accuracy between resumes and job requirements. This reduces misjudgments caused by non-standard information expression or lack of explicit key competencies, significantly improving the efficiency and effectiveness of recruitment. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram illustrating the implementation process of the resume matching method provided in Embodiment 1 of this application;

[0025] Figure 2 This is a schematic diagram illustrating the implementation process of the resume matching method provided in Embodiment 2 of this application;

[0026] Figure 3This is a schematic diagram illustrating the implementation process of the resume matching method provided in Embodiment 3 of this application;

[0027] Figure 4 This is a schematic diagram illustrating the implementation process of the resume matching method provided in Embodiment 4 of this application;

[0028] Figure 5 This is a schematic diagram illustrating the implementation process of the resume matching method provided in Embodiment 5 of this application;

[0029] Figure 6 This is a schematic diagram illustrating the implementation process of the resume matching method provided in Embodiment Six of this application;

[0030] Figure 7 This is a schematic diagram illustrating the implementation process of the resume matching method provided in Embodiment 7 of this application;

[0031] Figure 8 This is a schematic diagram of the resume matching device provided in the embodiments of this application;

[0032] Figure 9 This is a schematic diagram of the terminal device provided in the embodiments of this application. Detailed Implementation

[0033] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0034] To illustrate the technical solution described in this application, specific embodiments are provided below.

[0035] Figure 1 The flowchart illustrating the implementation of the resume matching method provided in Embodiment 1 of this application is shown below in detail:

[0036] Step S101: Obtain job application resume data and job requirement data; the job application resume data includes unstructured job application resume data, semi-structured job application resume data, and structured job application resume data; the job requirement data includes unstructured job requirement data, semi-structured job requirement data, and structured job requirement data.

[0037] In this embodiment, job application resume data can refer to various data sets that reflect personal abilities, experience, expectations, etc., submitted by job seekers through recruitment platforms, document uploads, or manual completion. This data can be manually entered by job seekers on recruitment platforms, such as by filling in work experience and skill tags; it can also be obtained by job seekers uploading PDF / Word format resume documents, with the platform's parsing tools extracting text and structured information; or it can be obtained through authorization via linked third-party accounts. Job requirement data can refer to data sets published by companies or recruiters through recruitment platforms, used to clarify recruitment standards, job responsibilities, and job requirements. This data can be obtained by companies filling out standardized forms on recruitment platforms, such as specifying educational requirements and salary ranges; or it can be obtained by entering job description text or batch importing job requirement documents. Unstructured resume data refers to data submitted by job seekers that lacks a fixed format and is primarily descriptive in natural language. This includes free text content such as personal introductions, job descriptions, project experience details, and self-evaluations. For example, job descriptions like "responsible for user growth projects on an e-commerce platform, optimizing conversion paths through data analysis" cannot be directly stored in database fields and require text parsing technology to extract information. This can be done by using OCR technology to recognize scanned text from text boxes filled in by job seekers or by using natural language processing tools to extract the free-form descriptions. Semi-structured resume data falls between structured and unstructured data, possessing a certain format but not strictly adhering to database specifications. It typically exists in the form of tags, lists, or semi-structured tables. It can include skill tag lists, such as "Python, Data Analysis, Project Management"; certificate names, such as "College English Test Band 6"; and work experience timelines, such as "2020.01-2023.05 Product Manager at a certain company." Structured data in job resumes refers to standardized data with a fixed format that can be directly stored in database fields, such as years of work experience, age, expected salary, and city of residence. It has clear field definitions and value ranges and can be directly read from standardized form fields filled out by job seekers, or quantified from vague expressions, such as mapping "more than three years of work experience" to "3 years". Unstructured data regarding job requirements refers to job-related content posted by companies without a fixed format, primarily using natural language descriptions. This can include job responsibilities, such as "responsible for building user growth models and optimizing core product metrics," or descriptions of soft skills required for the position, such as "possessing excellent cross-departmental communication and problem-solving abilities." This can be directly extracted from the job description text boxes and job detail description fields filled out by the company.Semi-structured job requirement data refers to job requirement data that has a certain format but is not strictly standardized. It is usually presented in the form of tags and conditional lists. It can include job requirement tags, such as "Requirement: 3+ years of backend development experience, proficient in Java"; it can include industry and field tags, such as "Industry: Internet / E-commerce, Field: ToB SaaS"; it can include skill proficiency descriptions, such as "Proficient in Python, familiar with SQL". It can be directly extracted from the job tags selected by the company and the skill requirement fields filled in, or it can be obtained by parsing the conditional statements in the job description. Structured job requirement data refers to standardized job requirement data with a fixed format that can be directly stored in database fields. It can include the number of positions to be filled, years of work experience requirements, education requirements, salary range, work location, etc. It has clear field definitions and value constraints. It can be directly read from the standardized recruitment form fields filled in by the company, or supplemented into a quantitative range by combining vague expressions with industry standards.

[0038] Step S102: Based on multiple preset resume data semantic parsing vectors, perform semantic parsing processing on the unstructured resume data, semi-structured resume data, and structured resume data to obtain the implicit information of the resume.

[0039] In this embodiment, the preset semantic parsing vector for resume data can be manually set. It can be combined with core competency dimensions and information features in the recruitment scenario. Through in-depth analysis of historical valid job application resume data and job requirement data, universally applicable semantic feature directions are extracted. These directions are then converted into vector forms that can be used for parsing to achieve the construction. Specifically, the construction method can first be based on common talent evaluation standards in the recruitment field to select dimensions strongly related to job matching, such as "professional skills mastery," "project experience depth," "soft skills suitability," "industry experience matching," and "education and qualification matching," with each dimension corresponding to a basic semantic parsing direction. Then, a large amount of verified valid job application resume data and corresponding job matching results are selected, and professional recruiters annotate the information in the samples related to each core dimension. For example, under the "professional skills mastery" dimension, statements such as "proficient in using Python for data analysis" and "proficient in Java backend development" in unstructured resume data are annotated, as are skill tags such as "Python, SQL, Java" in semi-structured resume data. Under the "project experience depth" dimension, statements such as "proficient in using Python for data analysis" and "proficient in Java backend development" are annotated, as are skill tags such as "Python, SQL, Java" in semi-structured resume data. Under the "Professional Skills Mastery" dimension, project descriptions such as "leading a project with tens of millions of users" and "independently completing the development of core modules" are marked, along with information such as project cycle and responsibilities. Then, for each marked core dimension, frequently occurring feature words, phrases, and sentence structures can be extracted from historical data. For example, the feature words for the "Professional Skills Mastery" dimension include skill proficiency descriptions such as "proficient," "master," and "master," as well as specific skill terms such as "Python," "Java," and "data analysis." The feature words for the "Soft Skills Suitability" dimension include "communication and coordination," "teamwork," and "problem solving," with expression patterns such as "responsible for cross-departmental communication" and "coordinating the team to achieve goals." Subsequently, the feature words, phrases, and expression patterns under each core dimension can be transformed into the basic components of a vector. Each dimension of the vector corresponds to a feature item, and the weight of the feature item is set according to its frequency and importance in historical successful matching cases. For example, in the initial vector of "Professional Skills Mastery," feature items such as "Python" and "proficient," which frequently appear and are important in technical job matching, have higher weights, while secondary expressions such as "basic understanding" have lower weights.Finally, the initial vector can be applied to the semantic parsing of some historical data. The parsing results are compared with the manually labeled results. Based on the differences, the weights and coverage of each feature item in the vector are repeatedly adjusted. For example, if the initial vector is insufficient in recognizing the related skill "big data processing" when parsing "using Python to process big data projects", the weights of feature items such as "big data" and "data processing" in the "professional skill mastery" vector are increased to form multiple stable and accurate preset resume data semantic parsing vectors. This results in multiple resume data semantic parsing vectors that can cover the key semantic parsing needs in the recruitment scenario and can be used for subsequent semantic parsing processing of unstructured resume data, semi-structured resume data, and structured resume data. Furthermore, for unstructured resume data, such as free text content like personal introductions, job descriptions, and project experience, pre-defined resume data semantic parsing vectors can be used for sentence-by-sentence or paragraph-by-paragraph semantic matching to identify key expressions related to the parsing vectors. For example, matching parsing vector features related to "teamwork" and "project management" from "led cross-departmental project implementation" can yield parsing vector features. For semi-structured resume data, such as skill tag lists, certificate names, and work experience timelines, pre-defined resume data semantic parsing vectors can be used to establish associations between tags and deeper capabilities. For example, matching the skill tags "Python, SQL" with the parsing vector of "data analysis ability," and associating "PMP certification" with the parsing vector of "project management ability," while also combining this with job promotion information in the timeline. The system includes a semantic parsing vector related to "career growth potential." For structured resume data, such as standardized data like years of work experience, education level, and expected salary, pre-defined semantic parsing vectors can transform numerical information into ability-related features. For example, it matches "5 years of work experience" with the "depth of industry experience" parsing vector, and associates "bachelor's degree or above" with the "basic professional knowledge" parsing vector. Through multiple rounds of semantic comparison and feature aggregation, the system deeply integrates scattered information from unstructured, semi-structured, and structured resume data with the pre-defined semantic parsing vectors, extracting implicit content that cannot be directly read from the raw data, such as "logical thinking ability," "problem-solving ability," and "industry adaptability," thus forming the implicit information in the resume.

[0040] Step S103: Perform semantic mining and analysis on the unstructured job requirement data, semi-structured job requirement data, and structured job requirement data to obtain implicit job requirement information.

[0041] In this embodiment, the text content in the unstructured job requirement data can be first broken down sentence by sentence and semantic association analysis can be performed to identify the implicit ability requirements. For example, potential requirements such as "model building ability" and "project execution ability" can be extracted from "responsible for building user growth models and promoting their implementation," and implicit requirements such as "adaptability" and "business sensitivity" can be extracted from "need to respond quickly to business changes and adjust strategies." Then, by establishing a mapping relationship between tags and core requirements, the semi-structured job requirement data can be transformed into deep requirement features. For example, the skill tag "proficient in Python and SQL" can be associated with the core requirement of "data analysis ability," and the tag "experience in the financial industry preferred" can be associated with the implicit requirement of "industry knowledge reserves." At the same time, combined with proficiency descriptions such as "proficient" and "familiar," the job requirements for different skills can be clarified. Based on the hierarchical requirements, we can then combine the semantic features of structured data with unstructured and semi-structured data to uncover implicit constraints. For example, from job descriptions such as "3-5 years of work experience" and "led large-scale projects," we can infer the implicit requirement for "depth of project experience"; from job tags such as "Master's degree or above" and "algorithm development," we can associate the potential requirement for "depth of professional knowledge." Through cross-validation and feature aggregation of multi-dimensional information, we can integrate semantic clues in unstructured job requirement data, tag associations in semi-structured job requirement data, and constraints in structured job requirement data to extract implicit content that cannot be directly obtained from a single data type, such as "preference for team collaboration mode," "stress resistance requirements," and "long-term development potential requirements," thus forming implicit information about job requirements.

[0042] Step S104: Perform resume data alignment processing on the unstructured resume data, semi-structured resume data, structured resume data, and implicit information of the resume to obtain aligned resume data.

[0043] In this embodiment, the data alignment dimension can be manually or automatically set based on the core evaluation indicators of job matching in the recruitment scenario, covering skill dimensions, experience dimensions, qualification dimensions, and ability dimensions. Each dimension contains several specific sub-items to ensure comprehensive coverage of key information in various types of resume data. This allows free text content such as personal introductions, job descriptions, and project experience to be split and mapped according to a unified alignment dimension. For example, the unstructured description "led a user growth project on an e-commerce platform, optimizing conversion paths through data analysis" can be broken down into "project experience" dimension, specifically "project type" and "core job title." The sub-items such as "Responsibilities" and "Skills Used" correspond to specific content such as "E-commerce Platform," "User Growth Strategy Development," and "Data Analysis," respectively, enabling unstructured data to be accurately matched with preset dimensions. Then, semi-structured information such as skill tag lists, certificate names, and work experience timelines can be standardized and unified according to the sub-item requirements of the aligned dimensions. For example, skill tags such as "Python, Data Analysis, Project Management" are categorized into the "Professional Skills" sub-item under the "Skills Dimension"; and the work experience timeline of "2020.01-2023.05 Product Manager at a certain company" is converted into the "Work Experience" sub-item under the "Experience Dimension." Standardized sub-items such as "Years of Experience" and "Job Title" eliminate discrepancies caused by different formats. This allows for precise filling of standardized data with fixed formats, such as years of work experience, education, age, and expected salary, according to pre-defined dimension sub-items. For example, "5 years" of work experience is mapped to the "Years of Experience" sub-item under the "Experience" dimension, and "Bachelor's Degree" is mapped to the "Highest Education Level" sub-item under the "Qualifications" dimension. This ensures seamless integration of structured data with the overall alignment framework. Furthermore, implicit content extracted from various data sources, such as "Logical Thinking Ability," "Problem-Solving Ability," and "Industry Suitability," can be mapped to the "Capability" dimension. Within the specific sub-items under "Dimension", for example, "Logical Thinking Ability" is added to the "Thinking Ability" sub-item under "Ability Dimension". This ensures that the aligned information not only includes explicit data but also covers deep implicit features. Finally, all transformed, integrated, and corresponding information is checked for consistency and merged to ensure that there are no conflicts or repetitions in the content of sub-items under the same dimension. The content scattered in the unstructured data, semi-structured data, structured data, and implicit information of job resumes is uniformly integrated into the preset alignment dimension framework to form aligned resume data with standardized format, unified dimensions, and complete information.

[0044] Step S105: Perform job requirement alignment processing on the unstructured job requirement data, semi-structured job requirement data, structured job requirement data, and implicit job requirement information to obtain job requirement aligned data.

[0045] In this embodiment, a unified job requirement alignment dimension system can be established first. This system can be built based on the core elements of job evaluation in recruitment scenarios, covering dimensions such as skill requirements, experience requirements, qualification requirements, and ability requirements. Specific sub-items are set under each dimension to ensure comprehensive inclusion of key information from various job requirement data. Then, free text content such as job descriptions, job highlights, and soft skills descriptions in job requirements can be broken down and mapped according to the unified alignment dimensions. For example, the unstructured description "Responsible for building and implementing user growth models, requiring rapid response to business changes" can be broken down into sub-items such as "core work content" and "ability requirements" under the "job requirements" dimension, corresponding to specific content such as "user growth model building" and "adaptability," enabling unstructured data to accurately match the preset dimensions. Furthermore, semi-structured information such as job requirement tags, skill proficiency descriptions, and industry and domain tags can be standardized and unified according to the sub-item requirements of the alignment dimensions. For example, "Proficient in Python and SQL, possessing financial industry experience" can be standardized and mapped. Tags and descriptions such as "experience priority" are categorized into sub-items such as "professional skills" and "industry experience" under the "skill requirements dimension". The proficiency description of "proficiency in data analysis tools" is converted into standardized content of the "skill mastery level" sub-item under the "skill requirements dimension", eliminating the differences caused by different expression formats. At the same time, implicit content such as "team collaboration mode preference", "stress resistance requirement", and "long-term development potential requirement" mined from various data can be mapped to specific sub-items under the "ability requirements dimension". For example, "team collaboration mode preference" is added to the "collaboration ability" sub-item under the "ability requirements dimension". This ensures that the aligned information not only includes explicit data but also covers deep implicit requirements. All converted, integrated, and mapped information is consistent and merged to ensure that there are no conflicts or repetitions in the content of sub-items under the same dimension. The content scattered in unstructured, semi-structured, and structured job requirement data, as well as implicit information in job requirements, is uniformly integrated into the preset alignment dimension framework to form job requirement aligned data.

[0046] Step S106: Based on multiple randomly generated resume matching calculation parameters and multiple preset resume matching parameter optimization step sizes, perform matching calculation processing on the job resume alignment data and job requirement alignment data to generate resume matching information.

[0047] In this embodiment, multiple randomly generated resume matching calculation parameters can include dimension weight parameters, sub-item matching threshold parameters, implicit feature association parameters, and comprehensive score correction parameters. The dimension weight parameters correspond to the importance weights of core dimensions such as skill requirements, experience requirements, qualification requirements, and ability requirements. For example, the weight of the skill requirements dimension can be randomly generated between 0.2 and 0.4, and the weight of the experience requirements dimension can be randomly generated between 0.15 and 0.3, ensuring that different dimensions influence the results in a reasonable proportion during the matching calculation. The sub-item matching threshold parameters can be the matching judgment thresholds for specific sub-items under each dimension. For example, the matching threshold for the "professional skills" sub-item in the skill requirements dimension, such as the semantic similarity between two skill tags needing to reach 0.6 or higher to be considered a match; and the matching threshold for the "work experience range" sub-item in the experience requirements dimension, such as the work experience in the resume matching the job requirements. The overlap ratio of the ranges must reach 0.7 or higher, and the sub-item matching threshold parameter can be randomly generated between 0.5 and 0.8; the implicit feature association parameter can be used to measure the correlation strength between implicit abilities in job resume alignment data and implicit requirements in job requirement alignment data, such as the association coefficient between "logical thinking ability" and "problem-solving requirements", which can be randomly generated between 0.3 and 0.5 to reflect the degree of influence of implicit features on the overall matching; the comprehensive score correction parameter can be used to adjust the correction coefficient when summarizing the scores of each dimension, such as the balance coefficient when the scores of the skill dimension and the experience dimension differ greatly, which can be randomly generated between 0.1 and 0.2 to avoid the extreme impact of a single dimension score being too high or too low on the overall result.The optimized dimensional weight parameters, sub-item matching threshold parameters, implicit feature association parameters, and comprehensive score correction parameters are used to perform matching calculations on job resume alignment data and job requirement alignment data. First, based on the optimized dimensional weight parameters, the proportions of skill requirements, experience requirements, qualification requirements, and ability requirements in the overall matching calculation can be determined. For example, the weight of the skill requirement dimension is 0.3, the weight of the experience requirement dimension is 0.25, the weight of the qualification requirement dimension is 0.2, and the weight of the ability requirement dimension is 0.25. Then, under each dimension, the matching status of the corresponding sub-item is judged by referring to the sub-item matching threshold parameters. For example, the "professional skills" sub-item in the skill requirement dimension needs to reach a matching threshold of 0.65. If the matching degree between "Python skills" in the job resume alignment data and "Python skill requirements" in the job requirement alignment data is 0.7, then it is judged... For a match to be considered, the "Years of Work Experience" sub-item in the experience requirement dimension must reach a matching threshold of 0.7. If the overlap ratio between the years of work experience in the job resume and the years of work experience required for the job is 0.8, it is considered a match. Then, for the ability requirement dimension, the correlation score between the implicit abilities in the job resume alignment data and the implicit requirements in the job requirement alignment data is calculated using implicit feature correlation parameters. For example, the correlation coefficient between "logical thinking ability" and "problem-solving ability" is 0.45. The score of this sub-item is obtained by combining the degree of matching between the two. Then, the scores of all sub-items under each dimension are weighted and summarized according to the dimension weight parameters to obtain a preliminary comprehensive score. Finally, the preliminary comprehensive score is adjusted by applying the comprehensive score correction parameters. If there is a large difference between the scores of the skill dimension and the experience dimension, the influence of the two is balanced by adjusting the parameters. Finally, resume matching information reflecting the overall fit between the job resume alignment data and the job requirement alignment data is generated.Multiple preset resume matching parameter optimization step sizes can be manually set. These can be determined by combining the characteristics of historical matching data and the sensitivity of different parameters, analyzing the impact of parameter adjustments on matching results. A large amount of historical resume alignment data, job requirement alignment data, and corresponding actual matching results (such as successful cases of hiring and interview passing) can be collected first to establish a basis for correlation analysis between parameter adjustments and matching accuracy. 1000 sets of historical successful matching cases and 1000 sets of failed cases can be selected as analysis samples. Then, parameters can be calculated for different types of resume matching to test the sensitivity of parameter adjustments to matching results. For example, using the weight parameter of the skill requirement dimension... Keeping other parameters constant, gradually increase the weight from 0.2 to 0.4, and observe the change in matching accuracy with each adjustment of 0.01: if the matching score of successful cases improves significantly after adjusting by 0.01, it indicates that the parameter is sensitive to the result. However, when the comprehensive score correction parameter is adjusted by 0.01, the matching score changes little, indicating that its sensitivity is low. Then, the basic optimization step size can be set according to the sensitivity. For highly sensitive parameters, a smaller optimization step size can be set, such as 0.01 each time, to avoid drastic parameter fluctuations due to excessively large step sizes, which would affect matching stability. For low-sensitivity parameters, a larger optimization step size can be set, such as 0.02 each time, to improve parameter optimization efficiency. Multiple randomly generated resume matching calculation parameters can cover matching weights, similarity thresholds, and feature correlation coefficients across different dimensions. These parameters quantify the degree of matching between job resume alignment data and job requirement alignment data across various dimensions, such as matching weights for skill requirements and similarity thresholds for experience requirements. The initial value of each parameter is generated randomly to cover a wider range of matching calculation possibilities. This allows for the introduction of multiple preset resume matching parameter optimization step sizes to progressively optimize the parameters during the calculation process. This ensures that parameter adjustments are neither too aggressive, leading to excessive fluctuations in matching results, nor too conservative, affecting optimization efficiency. For example, a smaller optimization step size can be set for the skill dimension to ensure matching accuracy, while a larger optimization step size can be set for the experience dimension to accommodate greater flexibility. The system matches job application data with available experience, and then performs multi-dimensional matching calculations on job application data and job requirement data based on several initially randomly generated resume matching parameters. For example, in the skill requirement dimension, it compares the professional skills sub-items in the job application data with those in the job requirement data, and calculates the skill matching score based on the matching weight of this dimension. In the experience requirement dimension, it refers to sub-items such as work experience range and project experience depth, and calculates the experience matching score based on the corresponding parameters. In the qualification requirement dimension, it generates a qualification matching score based on the matching of sub-items such as education and certificates, and generates a qualification matching score based on relevant parameters. In the ability requirement dimension, it compares the implicit ability features in the job application data with the implicit ability requirements in the job requirement data, and obtains the ability matching score using preset parameters.Then, the initial resume matching calculation parameters can be iteratively adjusted based on the optimization step size of multiple preset resume matching parameters. Specifically, the parameter values ​​can be gradually corrected according to the optimization step size by analyzing the differences between the matching scores of each dimension and historical successful matching cases. For example, if the matching score of the skill dimension deviates significantly from the actual successful cases, the skill matching weight can be adjusted according to the optimization step size of that dimension to make the calculation results closer to the actual matching situation. After multiple iterations, the comprehensive result of the matching scores of each dimension tends to be stable and the degree of consistency with historical data reaches the preset standard. Finally, the optimized matching scores of each dimension can be summarized to generate an overall matching score. At the same time, combined with the specific matching situation of each dimension, resume matching information containing detailed information such as skill matching degree, experience matching degree, qualification matching degree, and ability matching degree is formed, which fully presents the fit between the job resume and the job requirements.

[0048] The resume matching method provided in this application extracts implicit information from unstructured, semi-structured, and structured resume data that cannot be directly observed. Simultaneously, it extracts implicit information about job requirements from unstructured, semi-structured, and structured job requirement data. This method overcomes the limitations of traditional matching methods that rely solely on explicit information. By using alignment processing, it reduces comparison errors caused by differences in data format. Furthermore, by optimizing matching parameters, it performs refined matching calculations on aligned resume data and aligned job requirement data. By fully combining explicit and implicit information, it adapts to the complex and ever-changing resume formats and diverse job requirements in the job market, effectively improving the matching accuracy between resumes and job requirements. This reduces misjudgments caused by non-standard information expression or failure to reveal key abilities, significantly improving the efficiency and effectiveness of recruitment.

[0049] Figure 2 The flowchart illustrating the implementation of the resume matching method provided in Embodiment 2 of this application is shown. Its difference from Embodiment 1 described above lies in:

[0050] Multiple preset resume data semantic parsing vectors include preset resume data semantic focus vectors, preset resume data semantic association vectors, and preset resume data semantic integration vectors;

[0051] Step S102 specifically includes:

[0052] Step S201: Perform format conversion processing on the unstructured resume data, semi-structured resume data, and structured resume data to generate unstructured resume data vectors, semi-structured resume data vectors, and structured resume data vectors.

[0053] In this embodiment, the format conversion process can convert different types of job resume data into a unified vector form. For unstructured job resume data, such as job description text like "responsible for e-commerce platform user growth projects, optimizing conversion paths through data analysis," the text can be segmented using natural language processing tools and mapped into a fixed-dimensional numerical vector, with each dimension corresponding to the semantic features of a word. For semi-structured job resume data, such as a list of skill tags like "Python, data analysis, project management," each tag can be converted into binary features and concatenated into a vector, with a dimension value of 1 if the corresponding skill exists and 0 if it does not. For structured job resume data, such as "5 years of work experience, Bachelor's degree," the standardized numerical or categorical data can be directly converted into a vector, with a dimension value of 5 corresponding to the years of work experience and a dimension value mapped according to a preset rule, such as 2 for Bachelor's degree. Finally, unstructured job resume data vectors, semi-structured job resume data vectors, and structured job resume data vectors with unified dimensions are generated respectively.

[0054] Step S202: Based on the unstructured data vector, semi-structured data vector, structured data vector, preset semantic focus vector, preset semantic association vector, and preset semantic integration vector of the resume, calculate the following semantic focus vectors for the resume: unstructured data, semi-structured data, structured data, semantic association vector, semantic association vector, semantic association vector, semantic integration vector, semantic integration vector, and semantic integration vector.

[0055] In this embodiment, the preset resume data semantic focus vector, the preset resume data semantic association vector, and the preset resume data semantic integration vector can be set manually. The pre-defined semantic focus vector for resume data is used to accurately capture the core semantic information in various types of resume data. The construction process involves first filtering core semantic dimensions, and based on historical successful matching cases, determining dimensions such as "core professional skills terms," ​​"keywords for project responsibilities," and "key qualification indicators." For example, the core semantic dimensions for technical positions include "programming languages," "development tools," and "project roles," while for operations positions they include "user growth," "event planning," and "data indicators." Then, high-frequency core expressions from historical resume data are collected. For instance, from unstructured data of technical resumes, text such as "proficient in using Python for data analysis" and "leading backend system development" are extracted; from semi-structured data, skill tags such as "Python, Java, SQL" are extracted; and from structured data, information such as "3+ years of development experience" and "Bachelor of Computer Science" is extracted. These core expressions are then feature-refined, using words such as "Python," "data analysis," and "backend development," and indicators such as "3 years of experience" and "Bachelor of Science" as the basic feature terms of the vector. The weight of each feature term is set according to its frequency of occurrence in successful cases. For example, "P..." "Python" appears frequently in data-related job matching and has a higher weight than "basic office software"; "project-led" has a higher weight than "participation in execution" in management positions. Finally, through historical data verification and optimization, the initial vector is applied to the semantic parsing of some resumes. If key expressions such as "big data processing" are found to be insufficiently focused, their weight in the vector is increased, ultimately forming a preset resume data semantic focus vector that can accurately lock the core semantics. The preset resume data semantic association vector is used to mine the implicit associations between different features in various types of resume data. The construction process can first determine the association analysis dimensions, covering "skill-responsibility association", "experience-ability association", "qualification-job type association", etc. For example, "Python skills" are often associated with "data analysis responsibilities", and "cross-departmental experience" is often associated with "communication skills". Then, the co-occurrence frequency of features in historical data is counted. Expressions such as "using Python to complete user behavior analysis" are extracted from a large amount of unstructured resume data, and the co-occurrence frequency of "Python" and "user behavior analysis" is recorded. The co-occurrence of "PMP certificate" and "project management" tags is counted from semi-structured resume data.This study analyzes the correlation frequency between "5 years of experience" and "team management" responsibilities in structured resume data. High-frequency correlated feature pairs are then used as core elements of a vector, such as "Python-Data Analysis," "PMP-Project Management," and "5 years of experience-Team Management." Feature pairs with strong correlations are assigned higher weights; for example, "Java-Backend Development" has a higher co-occurrence frequency than "Java-Frontend Development," so the former has a higher weight. Finally, adjustments are made based on job type, strengthening the correlation between "programming language-development tools" for technical positions and "event planning-user growth" for operations positions. The effect of the correlation vector on matching results is verified, and the weights of feature pairs are optimized to form a pre-defined semantic correlation vector for resume data. This pre-defined semantic integration vector integrates different types of features according to matching priority, forming a comprehensive semantic expression. The construction process can begin by setting the priority of integration dimensions. Based on the influence of each dimension on matching in the recruitment scenario, a priority order such as "skill requirements > experience depth > qualifications > soft skills" is determined. For example, in technical positions, "professional skills" has a higher priority than "experience depth." In management positions, "team experience" takes precedence over "basic skills." This involves collecting key features from various dimensions, extracting "core job descriptions" from unstructured resume data, "core skill tags" from semi-structured data, and "key qualification indicators" from structured data. Basic weights are then assigned according to priority, for example, 40% for skills, 30% for experience, 20% for qualifications, and 10% for soft skills. Feature integration rules are then established, merging features of the same priority based on relevance. For example, in the skills dimension, "programming languages" and "development tools" are merged into "technology stack features," and in the experience dimension, "project duration" and "scope of responsibilities" are merged into "experience depth features." Weights are adjusted based on the feature's suitability for the position; for example, in data-related positions, "SQL skills" have a higher weight than "Java skills." Finally, historical matching data is used to verify the integration effect. If it is found that "soft skills" have a greater actual impact on a certain type of position, their weight in the integration vector is increased to ensure the vector covers key features with reasonable priority, ultimately forming a pre-defined semantic integration vector for resume data.

[0056] In this embodiment, the preset resume data semantic focus vector is used to focus on the core semantics in the data. During construction, core keywords of each dimension are extracted from historical job resume data, and the semantic features of these keywords are combined into vectors. The weights are set according to the importance of the keywords in successful matching cases. For example, "Python" has a higher weight than "basic office software" in technical job matching. The preset resume data semantic association vector is used to capture the association relationship between data. During construction, the co-occurrence pattern of features in historical data is analyzed, and high-frequency association features are combined into vectors. Features with high association strength have higher weights. The preset resume data semantic integration vector is used to integrate multi-dimensional features. During construction, the core semantics and association relationships are combined, and key features of different dimensions are combined into vectors according to matching priority. The weights of skills and experience dimension features are higher than secondary features such as hobbies. The unstructured resume data vector can be semantically compared with the preset resume data semantic focus vector to extract features in the text that match the core semantics and generate the unstructured resume data semantic focus vector. Similarly, the semi-structured and structured data vectors are compared with the preset resume data semantic focus vector to generate the corresponding semantic focus vector. By comparing various data vectors from job applicant resumes with pre-defined semantic association vectors for resume data, the relationships between features such as skills and experience, education and qualifications are captured, generating semantic association vectors for various data types. Furthermore, by comparing these vectors with pre-defined semantic integration vectors for resume data, core semantics and related features are integrated to generate semantic integration vectors for various data types.

[0057] In this embodiment, the unstructured data vector, semi-structured data vector, and structured data vector of a job resume can be multiplied by a preset semantic focus vector, a preset semantic association vector, and a preset semantic integration vector of resume data, respectively. The multiplication results are used as the semantic focus vector, semantic focus vector, semantic focus vector, semantic focus vector, semantic association vector, semantic association vector, semantic association vector, semantic association vector, semantic integration vector, semantic integration vector, and semantic integration vector of resume data.

[0058] Step S203: Based on the semantic focus vector of the unstructured data of the job resume, the semantic focus vector of the semi-structured data of the job resume, the semantic focus vector of the structured data of the job resume, the semantic association vector of the unstructured data of the job resume, the semantic association vector of the semi-structured data of the job resume, and the semantic association vector of the structured data of the job resume, calculate the semantic association feature vector of the unstructured data of the job resume, the semantic association feature vector of the semi-structured data of the job resume, and the semantic association vector of the structured data of the job resume.

[0059] In this embodiment, the calculation process can fuse semantic focus vectors and semantic association vectors of the same data type to strengthen the combination of core semantics and association relationships. For unstructured resume data, the semantic focus vectors and semantic association vectors can be merged according to feature dimensions, the weighted average of overlapping features is taken, and the original values ​​of unique features are retained to generate a semantic association feature vector for unstructured resume data, highlighting the combination of core semantics and association logic in the text. For semi-structured resume data, the semantic focus vectors and semantic association vectors of skill tags can be fused to clarify the association strength between skills and generate a semantic association feature vector for semi-structured resume data. For structured resume data, the semantic focus vectors and semantic association vectors of work experience and education can be fused to reflect the association between structured data and abilities and generate a semantic association feature vector for structured resume data.

[0060] In this embodiment, the semantic focus vector of unstructured resume data can be multiplied with the semantic association vector of unstructured resume data, the semantic focus vector of semi-structured resume data can be multiplied with the semantic association vector of semi-structured resume data, and the semantic focus vector of structured resume data can be multiplied with the semantic association vector of structured resume data. The multiplication results are respectively used as the semantic association feature vector of unstructured resume data, the semantic association feature vector of semi-structured resume data, and the semantic association feature vector of structured resume data.

[0061] Step S204: Based on the semantic association feature vectors of the unstructured data, semi-structured data, and structured data of the job resume, calculate the semantic feature integration vectors of the unstructured data, semi-structured data, and structured data of the job resume.

[0062] In this embodiment, the calculation process can be used to deeply integrate semantic association feature vectors and semantic integration vectors of the same data type to form a comprehensive feature representation. For unstructured job resume data, its semantic association feature vectors and semantic integration vectors can be fused according to preset weights. For example, the weights of core responsibilities and related skills in the "project description" can be adjusted according to the priority set by the integration vector to strengthen features strongly related to job matching and generate a semantic feature integration vector for unstructured job resume data. For semi-structured job resume data, skill association feature vectors and semantic integration vectors can be fused, and the feature weights can be adjusted according to the importance ranking of skills to the job in the integration vector to generate a semantic feature integration vector for semi-structured job resume data. For structured job resume data, structured association feature vectors and semantic integration vectors can be fused, and the weights can be adjusted according to the priority of "years of work experience > education level" in the integration vector to generate a semantic feature integration vector for structured job resume data, ensuring that the key features of the structured data are highlighted.

[0063] In this embodiment, the semantic association feature vector of unstructured resume data can be multiplied with the semantic integration vector of unstructured resume data, the semantic association feature vector of semi-structured resume data can be multiplied with the semantic integration vector of semi-structured resume data, and the semantic association feature vector of structured resume data can be multiplied with the semantic integration vector of structured resume data. The multiplication results are respectively used as the semantic integration vector of unstructured resume data, the semantic integration vector of semi-structured resume data, and the semantic integration vector of structured resume data.

[0064] Step S205: Based on the preset semantic feature mapping vector of the job resume, the semantic feature integration vector of the unstructured data of the job resume, the semantic feature integration vector of the semi-structured data of the job resume, and the semantic feature integration vector of the structured data of the job resume are mapped to obtain the implicit information of the job resume.

[0065] In this embodiment, the preset semantic feature mapping vector for job resumes can be manually set. It can be used to accurately map explicit features from the integrated vectors of unstructured, semi-structured, and structured semantic features of job resumes to implicit ability dimensions. Its construction must be based on the correlation between explicit features and implicit abilities. The construction process can begin by determining the core implicit ability dimensions. Based on the core qualities that companies value in recruitment scenarios, key dimensions such as "problem-solving ability," "teamwork ability," "learning ability," "execution ability," and "industry adaptability" are selected. Each dimension corresponds to a category of... We directly observe implicit features that are crucial for job matching, and then collect and label historical explicit-implicit correlation samples. We select a large amount of historical job application data and corresponding job matching results, and professional recruiters label explicit features related to each implicit ability. For example, under the "problem-solving ability" dimension, we label expressions such as "independently solved project technical problems" and "optimized inefficient processes to improve efficiency" in unstructured job application data; skill tags such as "troubleshooting" and "solution optimization" in semi-structured job application data; and quantitative information such as "duration spent handling core issues" in structured job application data. Under the "teamwork ability" dimension, we label "leading cross-departmental collaboration"... The data includes textual descriptions such as "project collaboration" and "coordinating team resources to achieve goals," tags such as "team management" and "cross-departmental communication," and structured data such as "duration of leading a team of 5 or more people." Then, the mapping relationship between explicit features and implicit abilities is extracted. For each implicit ability dimension, high-frequency explicit features are extracted from the labeled samples as the basic elements of the mapping vector. For example, the mapping elements for the "problem-solving ability" dimension include textual features such as "independent solution," "process optimization," and "troubleshooting," as well as structured features such as "core problem handling time ≥ 6 months." For instance, in technical positions, the feature weight associated with "problem-solving ability" is higher than that of "process optimization." "Optimization" is achieved through various means. For example, in management positions, the weight of the "team resource coordination" feature associated with "team collaboration ability" is higher than that of "cross-departmental communication." Similarly, in marketing positions, the weight of the "activity implementation effect" feature associated with "execution ability" is higher than that of "proposal writing." Finally, historical data can be used for verification and optimization. The initial mapping vector can be applied to extract implicit information from some resumes, and the matching degree between the extraction results and the implicit abilities required for the actual job can be compared. If it is found that the "learning ability" dimension has an extraction deviation due to insufficient weight of the "quick mastery of skills" feature, then the weight of this feature can be increased to form a preset semantic feature mapping vector for job resumes that can accurately map explicit features to implicit abilities.The semantic feature integration vectors of unstructured, semi-structured, and structured resumes can be compared with preset semantic feature mapping vectors for resumes. Scorees for each implicit ability dimension can be calculated through feature matching. For example, the text feature of "leading cross-departmental projects" can be matched with the "teamwork ability" mapping vector, and a score for that dimension can be generated by combining weights. The feature of "quickly mastering skills" can be matched with the "learning ability" mapping vector to generate a corresponding score. The scores of each dimension can be summarized to form implicit information in the resume such as "problem-solving ability" and "teamwork ability." Alternatively, the semantic feature integration vectors of unstructured, semi-structured, and structured resumes can be multiplied with preset semantic feature mapping vectors for resumes, and the multiplication results can be weighted and summed to obtain the implicit information in the resume.

[0066] The resume matching method provided in this application performs hierarchical vector transformation and fusion on unstructured, semi-structured, and structured data in job resumes. It combines the semantic feature mapping vector of job resumes to accurately extract implicit capabilities, effectively improving the depth and accuracy of implicit information in job resumes, thereby achieving more accurate resume-job matching in complex recruitment scenarios.

[0067] Figure 3 The flowchart illustrating the implementation of the resume matching method provided in Embodiment 3 of this application is shown. The difference between this method and Embodiment 2 is that step S202 specifically includes:

[0068] Step S301: Based on the unstructured data vector, semi-structured data vector, and structured data vector of the job application resume, and the preset semantic focus vector of the resume data, calculate the semantic focus vector of the unstructured data, the semantic focus vector of the semi-structured data, and the semantic focus vector of the structured data of the job application resume.

[0069] In this embodiment, the unstructured data vector, semi-structured data vector, and structured data vector of a job resume can be multiplied by a preset semantic focus vector of resume data, respectively. The multiplication results are the unstructured data semantic focus vector, semi-structured data semantic focus vector, and structured data semantic focus vector of a job resume, respectively.

[0070] Step S302: Based on the unstructured data vector, semi-structured data vector, structured data vector of the job resume, and preset semantic association vector of resume data, calculate the semantic association vector of the unstructured data, the semantic association vector of the semi-structured data, and the semantic association vector of the structured data of the job resume.

[0071] In this embodiment, the unstructured data vector, semi-structured data vector, and structured data vector of a job resume can be multiplied by a preset semantic association vector of resume data, respectively. The multiplication results are the semantic association vector of unstructured data, the semantic association vector of semi-structured data, and the semantic association vector of structured data, respectively.

[0072] Step S303: Based on the unstructured data vector, semi-structured data vector, structured data vector, and preset semantic integration vector of the resume data, calculate the semantic integration vector of the unstructured data, the semantic integration vector of the semi-structured data, and the semantic integration vector of the structured data.

[0073] In this embodiment, the unstructured data vector, semi-structured data vector, and structured data vector of a job resume can be multiplied by a preset semantic integration vector of resume data, respectively. The results of the multiplication are the semantic integration vector of unstructured data, the semantic integration vector of semi-structured data, and the semantic integration vector of structured data, respectively.

[0074] The resume matching method provided in this application improves the accuracy of capturing semantic features in job application resume data, the robustness of feature association mining, and makes the integration of multi-dimensional resume data more orderly and reliable, thereby effectively improving the accuracy of implicit information extraction and the precision of matching calculation.

[0075] Figure 4 The flowchart illustrating the implementation of the resume matching method provided in Embodiment 4 of this application is shown. The difference between this method and Embodiment 1 is that step S103 specifically includes:

[0076] Step S401: Encode the unstructured job requirement data, semi-structured job requirement data, and structured job requirement data to generate unstructured job requirement data vectors, semi-structured job requirement data vectors, and structured job requirement data vectors.

[0077] In this embodiment, one-hot encoding can be used to convert different types of job requirement data into a unified vector form to support subsequent semantic analysis. For unstructured job requirement data, such as job description text like "responsible for building user growth models and optimizing core product metrics," natural language processing tools can be used to segment and semantically encode the text, mapping each word to a corresponding semantic feature value, and combining them into a fixed-dimensional unstructured job requirement data vector. For semi-structured job requirement data, such as skill tags and descriptions like "proficient in Python and SQL, with financial industry experience preferred," skill tags can be converted into binary features, and proficiency descriptions can be mapped to numerical values ​​according to preset rules, concatenating them to form a semi-structured job requirement data vector. For structured job requirement data, such as "3-5 years of work experience, bachelor's degree or above, salary range 20k-30k / month," the work experience range can be set to the median of 4 years, the education level can be mapped according to rules, and the salary range can be set to the median of 25k / month, converting them into a numerical vector, i.e., a structured job requirement data vector.

[0078] Step S402: The unstructured data vector, semi-structured data vector, and structured data vector of job requirements are concatenated to generate multiple concatenated data vectors of job requirements.

[0079] In this embodiment, methods such as sequential concatenation by data type or hierarchical concatenation by feature importance can be adopted: When concatenating sequentially, the unstructured data vector of job requirements, the semi-structured data vector of job requirements, and the structured data vector of job requirements are connected in turn to form a concatenated vector containing complete data features; when concatenating hierarchically by importance, the semi-structured data vector is placed at the front for technical positions and the unstructured data vector is placed at the front for management positions. At the same time, concatenated vectors of different lengths are generated to ensure that each concatenated vector can reflect a certain core feature combination of job requirements, and finally, multiple concatenated vectors of job requirement data covering different feature association dimensions are generated.

[0080] Step S403: Group the spliced ​​vectors of multiple job requirement data to generate multiple job requirement data vector groups.

[0081] In this embodiment, the spliced ​​vectors can be categorized based on the core dimensions of job requirements, so that each group of vectors focuses on a type of requirement feature. First, the grouping dimensions can be determined, such as "Skill Requirements Group," "Experience Requirements Group," "Ability Requirements Group," and "Qualification Requirements Group." Then, the spliced ​​vectors are divided according to the dominant dimension of the features: spliced ​​vectors primarily based on skill tags and skill proficiency features are grouped into the "Skill Requirements Group"; those primarily based on years of work experience and project experience are grouped into the "Experience Requirements Group"; those primarily based on soft skill descriptions and job-related features are grouped into the "Ability Requirements Group"; and those primarily based on education and certification features are grouped into the "Qualification Requirements Group." If a spliced ​​vector contains multiple types of features, it is categorized according to the dimension with the highest feature weight, ultimately generating multiple job requirement data vector groups with closely related features within each group.

[0082] Step S404: Perform semantic mining and analysis on the multiple job requirement data vector groups to obtain implicit job requirement information.

[0083] In this embodiment, implicit requirement features can be extracted in depth for each group of vectors and integrated across groups to achieve semantic mining and analysis. By analyzing the association patterns of skill tags in the vectors, such as "Python and SQL often co-occur," implicit features such as "data analysis ability is a core skill requirement" can be extracted. For the "experience requirement group," the implicit requirement of "experience in large-scale project implementation" can be inferred by combining years of work experience and project scale characteristics. For example, the experience depth requirement can be extracted from the feature "3-5 years of experience + leading projects with tens of millions of users." For the "ability requirement group," implicit ability requirements such as "preference for team collaboration modes" and "stress resistance requirements" can be extracted through correlation analysis of job descriptions and soft skill features. For example, the "business sensitivity" requirement can be extracted from the feature "rapid response to business changes." For the "qualification requirement group," implicit qualification requirements such as "depth of professional knowledge requirements" can be correlated by combining education level and job type characteristics. The implicit features mined from each group are cross-validated and integrated to eliminate conflicting information and form implicit information about job requirements.

[0084] The resume matching method provided in this application strengthens the correlation analysis of different types of requirement features by encoding and converting job requirement data, splicing multiple dimensions, grouping and focusing, and grouping semantic mining. This improves the pertinence and comprehensiveness of extracting implicit requirements, making the implicit information of job requirements more consistent with the core dimensions of actual recruitment needs. This lays a reliable feature foundation for the accurate matching of resume information with job seekers' resumes, effectively improving the accuracy and effectiveness of resume matching.

[0085] Figure 5 The flowchart illustrating the implementation of the resume matching method provided in Embodiment 5 of this application is shown. The difference between this method and Embodiment 4 above is that step S403 specifically includes:

[0086] Step S501: Based on the preset number of grouped job requirement data splicing vectors, randomly select from the multiple job requirement data splicing vectors to obtain multiple job requirement data main vectors.

[0087] In this embodiment, the preset number of grouped job requirement data splicing vectors can be manually set. The construction process can begin by analyzing the characteristic dimension distribution of historical job requirement data. For example, technical positions typically need to cover three core dimensions: "skill requirements," "experience requirements," and "ability requirements," while management positions need to cover four core dimensions: "team management," "strategic planning," and "communication and coordination." Based on this, the initial range of grouping numbers can be set. Then, the total number of actual job requirement data splicing vectors can be adjusted. If the total number of vectors is 20, the number of groups can be set to 4-5 to ensure that each group contains a sufficient number of vectors to support subsequent analysis. For example, for technical positions, the preset number of grouped job requirement data splicing vectors can be set to 3, corresponding to the "skill requirements group," "experience requirements group," and "ability requirements group." Vectors with the same number of grouping numbers as the job requirement data splicing vectors can be randomly selected from multiple job requirement data splicing vectors as initial master vectors, i.e., multiple job requirement data master vectors.

[0088] Step S502: Based on the splicing vector of multiple job requirement data and the multiple job requirement data master vector, multiple job requirement data servant vectors are obtained.

[0089] In this embodiment, the remaining vectors in the concatenated vector of multiple job requirement data that were not extracted as the main vector of multiple job requirement data can be determined as the servant vectors of multiple job requirement data.

[0090] Step S503: Based on the master vector and servant vector of the multiple job requirement data, calculate the distance information between the master and servant vectors of the multiple job requirement data.

[0091] In this embodiment, the degree of correlation between the two can be determined by calculating the cosine similarity between the servant vector of each job requirement data and the principal vector of each job requirement data, which is the distance information between the principal and servant vectors of the job requirement data.

[0092] Step S504: Based on the master-servant vector distance information of the multiple job requirement data, the master vector and servant vector of the multiple job requirement data are divided to generate multiple master-servant vector groups of job requirement data.

[0093] In this embodiment, each job requirement data servant vector can be grouped into the group containing the nearest job requirement data master vector, forming multiple job requirement data master-servant vector groups with the master vector as the core and the servant vectors as subordinates.

[0094] Step S505: Calculate the median and average of the master-servant vector groups of the multiple job demand data to obtain the median vector and the average vector of the master-servant groups of the multiple job demand data.

[0095] In this embodiment, for each job requirement data master-servant vector group, the distribution of feature values ​​of all vectors in the group can be statistically analyzed, and the vector with the middle position of the feature values ​​can be taken as the median vector to reflect the typical level of the feature within the group; the average level of the feature values ​​of all vectors in each job requirement data master-servant vector group can be taken to form the mean vector to reflect the overall trend of the feature within the group.

[0096] Step S506: Determine the updated master vector of job requirement data based on the median vector of the master-servant group of job requirement data, the mean vector of the master-servant group of job requirement data, and the preset threshold for the difference between the master-servant group vectors of job requirement data.

[0097] In this embodiment, the preset threshold for the difference between the master and servant vectors of the job requirement data can be set manually. It can be constructed by analyzing the reasonable range of differences between the master vector and the median and mean vectors within the group in historical grouped data. If the difference between the median or mean vector of the master and servant groups of the job requirement data and the original master vector of the job requirement data exceeds the threshold, then the median or mean vector is used as the updated master vector of the job requirement data. If the difference between the median or mean vector of the master and servant groups of the job requirement data and the original master vector of the job requirement data does not exceed the threshold, then the original master vector of the job requirement data remains unchanged and can be used as the updated master vector of the job requirement data for subsequent calculations.

[0098] Step S507: Determine whether the multiple updated job requirement data main vectors are the same as the multiple job requirement data main vectors; if yes, proceed to step S508; if no, proceed to step S509.

[0099] In this embodiment, the feature values ​​of the updated job requirement data principal vector and the original job requirement data principal vector can be compared. If the core features of all job requirement data principal vectors remain unchanged, they are determined to be the same; if the core features of any job requirement data principal vector are adjusted, they are determined to be different.

[0100] Step S508: Generate multiple job requirement data vector groups based on the master-servant vector groups of the multiple job requirement data.

[0101] In this embodiment, when the updated job requirement data master vector is the same as the original job requirement data master vector, it indicates that the grouping has stabilized. At this time, the multiple job requirement data master-servant vector groups are directly determined as the final multiple job requirement data vector groups.

[0102] Step S509: The multiple updated job requirement data master vectors are used as multiple job requirement data master vectors, and the process returns to step S502.

[0103] In this embodiment, when the updated principal vector of job requirement data is different from the original principal vector of job requirement data, it indicates that the grouping is not yet stable. It is necessary to redetermine the servant vector, calculate the principal-servant distance, and divide the new principal-servant vector group until the principal vector of job requirement data is no longer updated, so as to ensure that the grouping result can accurately reflect the core dimension features of job requirements.

[0104] The resume matching method provided in this application divides job requirement data into master-servant vectors to achieve dynamic grouping of job requirement data. The stability of the grouping is verified by using median and mean vectors, making the division of job requirement data vector groups more in line with the core dimensions of actual needs. This significantly improves the accuracy and reliability of grouping, provides a stable feature foundation for subsequent semantic mining, and enhances the depth and accuracy of extracting implicit information from job requirements.

[0105] Figure 6 The flowchart illustrating the implementation of the resume matching method provided in Embodiment Six of this application is shown. The difference between this method and Embodiment Five is that step S506 specifically includes:

[0106] Step S601: Determine whether the difference between the median vector of the master-servant group of the job demand data and the mean vector of the master-servant group of the job demand data is greater than a preset threshold for the difference between the master-servant group vectors of the job demand data; if yes, proceed to step S602; if no, proceed to step S603.

[0107] In this embodiment, the preset threshold for the difference between the master and servant vectors of job requirement data can be set manually. It can be set based on the reasonable fluctuation range of the median and mean vectors in historical grouped data. For example, analysis of historical data for technical positions might reveal that when the difference between the median and mean vectors of the master and servant groups of job requirement data exceeds 15%, it indicates abnormal fluctuations in the data within the group. In this case, the master vector update rule needs to be adjusted through threshold control. The preset threshold for the difference between the master and servant vectors of job requirement data can be set to 15%.

[0108] Step S602: Use the mean vector of the master and servant groups of the job demand data as the updated master vector of the job demand data.

[0109] In this embodiment, when the difference between the median vector of the master and servant groups of job demand data and the mean vector of the master and servant groups of job demand data is greater than the threshold of the difference between the master and servant groups of job demand data, it indicates that there is a large fluctuation in the vector characteristics within the group. If there are individual extreme values ​​that raise or lower the median, the mean vector of the master and servant groups of job demand data will be used as the updated master vector of job demand data to avoid the interference of extreme values ​​on the master vector.

[0110] Step S603: Use the median vector of the master-servant group of the job requirement data as the updated master vector of the job requirement data.

[0111] In this embodiment, when the difference between the median vector of the master-servant group of job requirement data and the mean vector of the master-servant group of job requirement data does not exceed the threshold of the difference between the master-servant group vectors of job requirement data, it indicates that the vector features within the group are evenly distributed and the median vector of the master-servant group of job requirement data can stably represent the typical features within the group. Therefore, the median vector of the master-servant group of job requirement data can be used as the updated master vector of job requirement data.

[0112] The resume matching method provided in this application distinguishes the data fluctuation within each master-servant vector group of job requirement data by using a preset threshold for the difference between the master and servant vectors of job requirement data. It dynamically selects the median vector or the mean vector as the updated master vector of job requirement data, which avoids the interference of extreme values ​​on the group master vector and ensures that the master vector can accurately reflect the typical features within the group. This improves the stability and rationality of the division of job requirement data vector groups and enhances the effectiveness and robustness of subsequent semantic mining and resume matching.

[0113] Figure 7 The flowchart illustrating the implementation of the resume matching method provided in Embodiment Seven of this application is shown. Its difference from Embodiment One described above lies in:

[0114] The multiple randomly generated resume matching calculation parameters include randomly generated resume matching calculation weight parameters, randomly generated resume matching similarity threshold parameters, and randomly generated resume matching deviation correction parameters;

[0115] The preset resume matching parameter optimization step size includes a preset resume matching calculation weight parameter optimization step size, a preset resume matching similarity threshold parameter optimization step size, and a preset resume matching deviation correction parameter optimization step size.

[0116] Step S106 specifically includes:

[0117] Step S701: Calculate the similarity between the job resume alignment data and the job requirement alignment data to obtain resume matching similarity information.

[0118] In this embodiment, the calculation process needs to cover the matching degree between job resume alignment data and job requirement alignment data across each core dimension. For example, in the skill requirement dimension, the semantic similarity of the "Python skill proficiency" sub-item is compared; in the experience requirement dimension, the overlap ratio of work experience ranges is calculated; and in the ability requirement dimension, the correlation strength of implicit features is analyzed. By weighted summarization of the similarity across dimensions such as skills, experience, qualifications, and abilities, comprehensive resume matching similarity information is generated, for example, the comprehensive similarity between a resume and a job posting is 75%.

[0119] Step S702: Determine whether the resume matching similarity information is less than or equal to the randomly generated resume matching similarity threshold parameter; if yes, proceed to step S703; if no, skip the job resume alignment data and job requirement alignment data corresponding to the resume matching similarity information.

[0120] In this embodiment, the randomly generated resume matching similarity threshold parameter can be randomly generated between 0.5 and 0.8, for example, the initially generated resume matching similarity threshold parameter is 0.6. If the resume matching similarity information is greater than 0.6, it means that the initial matching has met the standard, and no further correction of the resume matching calculation parameters is required; if the resume matching similarity information is less than or equal to 0.6, it is necessary to proceed to the subsequent optimization process of the resume matching calculation parameters to avoid missing potential matching resumes.

[0121] Step S703: Based on the randomly generated resume matching calculation weight parameters and the randomly generated resume matching deviation correction parameters, the job resume alignment data and job requirement alignment data are matched to generate resume matching variable information.

[0122] In this embodiment, the randomly generated resume matching calculation weight parameters can be set according to dimensions, such as a weight of 0.3 for the skill requirement dimension, 0.25 for the experience requirement dimension, 0.2 for the qualification requirement dimension, and 0.25 for the ability requirement dimension. The randomly generated resume matching deviation correction parameters can be generated between 0.1 and 0.2 to balance extreme scores. The matching scores of each dimension can be calculated first according to the weight parameters, and then the outliers can be adjusted by the deviation correction parameters to generate resume matching variable information containing the corrected scores of each dimension.

[0123] Step S704: Calculate the resume matching accuracy information based on the job application alignment data, job requirement alignment data, resume matching variable information, and the preset job application matching database.

[0124] In this embodiment, the pre-built job resume matching database can be manually constructed and may contain multiple historical successful matching cases, such as alignment data and matching results of job applications and corresponding positions. The construction process requires collecting at least 1000 sets of historical data, labeling them with "high match," "medium match," and "low match" tags and corresponding features. The current resume matching variable information can be compared with historical data of similar positions in the database, and the degree of consistency between the corrected score and the actual matching result can be statistically analyzed. For example, if a certain variable information matches the features of 80% of the historical high-match cases, then the resume matching accuracy information is 80%.

[0125] Step S705: Determine whether the resume matching accuracy information is greater than the preset resume matching accuracy threshold information; if yes, proceed to step S706; if no, proceed to step S707.

[0126] In this embodiment, the preset resume matching accuracy threshold can be set manually, or it can be set based on the average accuracy of historical data. For example, if analysis shows that the matching result is more reliable when the accuracy exceeds 70%, then the resume matching accuracy threshold is set to 70%. When the resume matching accuracy is greater than the preset resume matching accuracy threshold, the match is considered valid; when the resume matching accuracy is less than or equal to the preset resume matching accuracy threshold, the match is considered invalid, and further optimization of the resume matching calculation parameters is required.

[0127] Step S706: Generate resume matching information based on the resume matching variable information.

[0128] In this embodiment, the scores of each dimension in the resume matching variable information can be integrated after correction to generate resume matching information that includes the overall matching score and the matching degree of the subdivided dimensions. For example, the resume matching information includes an overall matching score of 78 points and subdivided dimension matching degrees of 85% for skills, 70% for experience, 80% for qualifications, and 75% for ability, thus fully presenting the suitability of the job application resume and the job position.

[0129] Step S707: Based on the preset optimization step size of the resume matching calculation weight parameter, the preset optimization step size of the resume matching similarity threshold parameter, and the preset optimization step size of the resume matching deviation correction parameter, optimize the randomly generated resume matching calculation weight parameter, the randomly generated resume matching similarity threshold parameter, and the randomly generated resume matching deviation correction parameter to obtain the optimized resume matching calculation weight parameter, the optimized resume matching similarity threshold parameter, and the optimized resume matching deviation correction parameter.

[0130] In this embodiment, the preset optimization step size for the resume matching calculation weight parameter, the preset optimization step size for the resume matching similarity threshold parameter, and the preset optimization step size for the resume matching deviation correction parameter can all be manually set. The preset optimization step size for the resume matching calculation weight parameter can be 0.01, the preset optimization step size for the resume matching similarity threshold parameter can be 0.02, and the preset optimization step size for the resume matching deviation correction parameter can be 0.01. The resume matching calculation weight parameter of the key dimensions can be increased by the step size, the resume matching similarity threshold parameter can be decreased, or the resume matching deviation correction parameter can be adjusted. The resume matching calculation weight parameter, the resume matching similarity threshold parameter, and the resume matching deviation correction parameter adjusted according to the step size are used as the optimized resume matching calculation weight parameter, the optimized resume matching similarity threshold parameter, and the optimized resume matching deviation correction parameter.

[0131] Step S708: Use the optimized resume matching calculation weight parameter as the randomly generated resume matching calculation weight parameter, use the optimized resume matching similarity threshold parameter as the randomly generated resume matching similarity threshold parameter, use the optimized resume matching deviation correction parameter as the randomly generated resume matching deviation correction parameter, and return to step S702.

[0132] In this embodiment, the optimized parameters can replace the initial parameters, and the matching judgment process can be re-entered. The matching accuracy can be gradually improved through iterative adjustments until the matching accuracy meets the standard.

[0133] The resume matching method provided in this application is used to dynamically iteratively optimize multiple resume matching calculation parameters. It reduces invalid calculations by using a resume matching similarity threshold parameter and corrects parameter balance dimension differences by using resume matching deviation, thereby improving the accuracy and effectiveness of resume matching information and enhancing the efficiency and reliability of matching resumes with job requirements.

[0134] Corresponding to the method in the above embodiments, Figure 8 A structural block diagram of the resume matching device provided in the embodiments of this application is shown. For ease of explanation, only the parts related to the embodiments of this application are shown. Figure 8 The example resume matching device can be the execution subject of the resume matching method provided in the aforementioned embodiment one.

[0135] Reference Figure 8 The resume matching device includes:

[0136] The job application data and job requirement data acquisition module 810 is used to acquire job application data and job requirement data; the job application data includes unstructured job application data, semi-structured job application data, and structured job application data; the job requirement data includes unstructured job requirement data, semi-structured job requirement data, and structured job requirement data.

[0137] The job resume implicit information generation module 820 is used to perform semantic parsing processing on the unstructured data, semi-structured data and structured data of the job resume based on multiple preset resume data semantic parsing vectors to obtain the implicit information of the job resume.

[0138] The job requirement implicit information generation module 830 is used to perform semantic mining and analysis on the unstructured job requirement data, semi-structured job requirement data and structured job requirement data to obtain the job requirement implicit information.

[0139] The job resume alignment data generation module 840 is used to perform resume data alignment processing on the unstructured resume data, semi-structured resume data, structured resume data, and implicit information of the job resume to obtain job resume alignment data.

[0140] The job requirement alignment data generation module 850 is used to perform job requirement alignment processing on the unstructured job requirement data, semi-structured job requirement data, structured job requirement data, and implicit job requirement information to obtain job requirement alignment data.

[0141] The resume matching information generation module 860 is used to perform matching calculations on the job application resume alignment data and job requirement alignment data based on multiple randomly generated resume matching calculation parameters and multiple preset resume matching parameter optimization step sizes, and generate resume matching information.

[0142] For details on how each module in the resume matching device provided in this application implements its respective function, please refer to the foregoing. Figure 1 The description of Embodiment 1 shown will not be repeated here.

[0143] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0144] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0145] The resume matching method provided in this application can be applied to terminal devices such as mobile phones and tablets. This application does not impose any restrictions on the specific type of terminal device.

[0146] Figure 9 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. For example... Figure 9 As shown, the terminal device 9 of this embodiment includes: at least one processor 90 ( Figure 9 (Only one is shown in the image), memory 91, which stores a computer program 92 that can run on the processor 90. When the processor 90 executes the computer program 92, it implements the steps in the above-described resume matching method embodiments, for example... Figure 1 Steps S101 to S106 are shown. Alternatively, when the processor 90 executes the computer program 92, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 8 The functions of modules 810 to 860 are shown.

[0147] The terminal device 9 can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor 90 and a memory 91. Those skilled in the art will understand that... Figure 9 This is merely an example of terminal device 9 and does not constitute a limitation on terminal device 9. It may include more or fewer components than shown, or combine certain components, or use different components.

[0148] The processor 90 can be a central processing unit, or it can be other general-purpose processors, digital signal processors, application-specific integrated circuits (ASICs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0149] In some embodiments, the memory 91 may be an internal storage unit of the terminal device 9, such as a hard disk or memory of the terminal device 9. The memory 91 may also be an external storage device of the terminal device 9, such as a plug-in hard disk or smart memory card equipped on the terminal device 9. Furthermore, the memory 91 may include both internal and external storage units of the terminal device 9. The memory 91 is used to store operating systems, applications, bootloaders, data, and other programs, such as the program code of computer programs. The memory 91 can also be used to temporarily store data that has been sent or will be sent.

[0150] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0151] This application also provides a terminal device, which includes at least one memory, at least one processor, and a computer program stored in the at least one memory and executable on the at least one processor. When the processor executes the computer program, it causes the terminal device to implement the steps in any of the above method embodiments.

[0152] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0153] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0154] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0155] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A resume matching method, characterized in that, include: Acquire job application data and job requirement data; the job application data includes unstructured job application data, semi-structured job application data, and structured job application data; The job requirement data includes unstructured job requirement data, semi-structured job requirement data, and structured job requirement data. Based on multiple preset resume data semantic parsing vectors, semantic parsing processing is performed on the unstructured resume data, semi-structured resume data, and structured resume data to obtain the implicit information of the resume. Semantic mining and analysis are performed on the unstructured, semi-structured, and structured job requirement data to obtain implicit job requirement information. The unstructured resume data, semi-structured resume data, structured resume data, and implicit information of the resumes are aligned to obtain aligned resume data. The unstructured job requirement data, semi-structured job requirement data, structured job requirement data, and implicit job requirement information are aligned to obtain aligned job requirement data. Based on multiple randomly generated resume matching calculation parameters and multiple preset resume matching parameter optimization step sizes, the job resume alignment data and job requirement alignment data are matched and calculated to generate resume matching information. The multiple randomly generated resume matching calculation parameters include randomly generated resume matching calculation weight parameters, randomly generated resume matching similarity threshold parameters, and randomly generated resume matching deviation correction parameters; The preset resume matching parameter optimization step size includes a preset resume matching calculation weight parameter optimization step size, a preset resume matching similarity threshold parameter optimization step size, and a preset resume matching deviation correction parameter optimization step size. The step of performing matching calculations on the job resume alignment data and job requirement alignment data based on multiple randomly generated resume matching calculation parameters and multiple preset resume matching parameter optimization step sizes to generate resume matching information specifically includes: Calculate the similarity between the job resume alignment data and the job requirement alignment data to obtain resume matching similarity information; When the resume matching similarity information is less than or equal to the randomly generated resume matching similarity threshold parameter, the resume matching calculation weight parameter and the randomly generated resume matching deviation correction parameter are used to match the job application resume alignment data and the job requirement alignment data to generate resume matching variable information. Based on the job resume alignment data, job requirement alignment data, resume matching variable information, and the preset job resume matching database, the resume matching accuracy information is calculated. Determine whether the resume matching accuracy information is greater than a preset resume matching accuracy threshold information; If so, then generate resume matching information based on the resume matching variable information; If not, then based on the preset resume matching calculation weight parameter optimization step size, the preset resume matching similarity threshold parameter optimization step size, and the preset resume matching deviation correction parameter optimization step size, the randomly generated resume matching calculation weight parameter, the randomly generated resume matching similarity threshold parameter, and the randomly generated resume matching deviation correction parameter are optimized to obtain the optimized resume matching calculation weight parameter, the optimized resume matching similarity threshold parameter, and the optimized resume matching deviation correction parameter. The optimized resume matching calculation weight parameter is used as the randomly generated resume matching calculation weight parameter, the optimized resume matching similarity threshold parameter is used as the randomly generated resume matching similarity threshold parameter, and the optimized resume matching deviation correction parameter is used as the randomly generated resume matching deviation correction parameter. Then, the process returns to the step where, when the resume matching similarity information is less than or equal to the randomly generated resume matching similarity threshold parameter, the job resume alignment data and job requirement alignment data are matched according to the randomly generated resume matching calculation weight parameter and the randomly generated resume matching deviation correction parameter to generate resume matching variable information.

2. The resume matching method as described in claim 1, characterized in that, Multiple preset resume data semantic parsing vectors include preset resume data semantic focus vectors, preset resume data semantic association vectors, and preset resume data semantic integration vectors; The step of performing semantic parsing processing on unstructured resume data, semi-structured resume data, and structured resume data based on multiple preset resume data semantic parsing vectors to obtain implicit information from the resumes specifically includes: The unstructured, semi-structured, and structured resume data are converted to generate unstructured, semi-structured, and structured resume data vectors. Based on the unstructured data vector, semi-structured data vector, structured data vector, preset semantic focus vector, preset semantic association vector, and preset semantic integration vector of the resume, the following vectors are calculated: unstructured data semantic focus vector, semi-structured data semantic focus vector, structured data semantic focus vector, unstructured data semantic association vector, semi-structured data semantic association vector, structured data semantic association vector, unstructured data semantic integration vector, semi-structured data semantic integration vector, and structured data semantic integration vector of the resume. Based on the semantic focus vector of unstructured resume data, semantic focus vector of semi-structured resume data, semantic focus vector of structured resume data, semantic association vector of unstructured resume data, semantic association vector of semi-structured resume data, and semantic association vector of structured resume data, the semantic association feature vector of unstructured resume data, semantic association feature vector of semi-structured resume data, and semantic association feature vector of structured resume data are calculated. Based on the semantic association feature vectors of unstructured data, semistructured data, and structured data of job resumes, the semantic integration vectors of unstructured data, semistructured data, and structured data of job resumes are calculated to obtain the semantic feature integration vectors of unstructured data, semistructured data, and structured data of job resumes. Based on the preset semantic feature mapping vector of job resumes, the semantic feature integration vector of unstructured data, the semantic feature integration vector of semi-structured data, and the semantic feature integration vector of structured data of job resumes are mapped to obtain the implicit information of job resumes.

3. The resume matching method as described in claim 2, characterized in that, The step of calculating the semantic focus vector, semantic focus vector, semantic focus vector, semantic focus vector, structural data focus vector, semantic association vector, semantic association vector, semantic integration vector, and semantic integration vector of the resume based on the unstructured data vector, semi-structured data vector, structured data vector, preset semantic focus vector, preset semantic association vector, and preset semantic integration vector of the resume specifically includes: Based on the unstructured data vector of the job resume, the semi-structured data vector of the job resume, the structured data vector of the job resume, and the preset semantic focus vector of the resume data, the semantic focus vector of the unstructured data of the job resume, the semantic focus vector of the semi-structured data of the job resume, and the semantic focus vector of the structured data of the job resume are calculated. Based on the unstructured data vector of the job resume, the semi-structured data vector of the job resume, the structured data vector of the job resume, and the preset semantic association vector of the resume data, the semantic association vector of the unstructured data of the job resume, the semantic association vector of the semi-structured data of the job resume, and the semantic association vector of the structured data of the job resume are calculated. Based on the unstructured data vector, semi-structured data vector, and structured data vector of the job resume, and the preset semantic integration vector of resume data, the semantic integration vectors of the unstructured, semi-structured, and structured data of the job resume are calculated.

4. The resume matching method as described in claim 1, characterized in that, The step of performing semantic mining and analysis on the unstructured, semi-structured, and structured job requirement data to obtain implicit information about job requirements specifically includes: The unstructured, semi-structured, and structured job requirement data are encoded to generate unstructured, semi-structured, and structured job requirement data vectors. The unstructured data vector, semi-structured data vector, and structured data vector of job requirements are concatenated to generate multiple concatenated data vectors of job requirements. The concatenated vectors of multiple job requirement data are grouped to generate multiple job requirement data vector groups; Semantic mining and analysis are performed on the multiple job requirement data vector groups to obtain implicit information about job requirements.

5. The resume matching method as described in claim 4, characterized in that, The step of grouping the concatenated vectors of multiple job requirement data to generate multiple job requirement data vector groups specifically includes: Based on the preset number of grouped job requirement data splicing vectors, the multiple job requirement data splicing vectors are randomly selected to obtain multiple job requirement data main vectors. Based on the concatenation vector of multiple job requirement data and the multiple job requirement data master vector, multiple job requirement data servant vectors are obtained; Based on the master vector and servant vector of the multiple job requirement data, the distance information between the master and servant vectors of the multiple job requirement data is calculated. Based on the master-servant vector distance information of the multiple job requirement data, the master vector and servant vector of the multiple job requirement data are divided to generate multiple master-servant vector groups of job requirement data. Calculate the median and average of the master-servant vector groups of the multiple job demand data to obtain the median vector and the average vector of the master-servant groups of the multiple job demand data. Based on the median vector of the master and servant groups of the job requirement data, the mean vector of the master and servant groups of the job requirement data, and the preset threshold for the difference between the master and servant groups of the job requirement data, the updated master vector of the job requirement data is determined. Determine whether the multiple updated job requirement data principal vectors are the same as the multiple job requirement data principal vectors; If so, then generate multiple job requirement data vector groups based on the master-servant vector groups of the multiple job requirement data; If not, then the multiple updated job requirement data master vectors are used as multiple job requirement data master vectors, and the process returns to the step of concatenating the multiple job requirement data vectors and the multiple job requirement data master vectors to obtain multiple job requirement data servant vectors.

6. The resume matching method as described in claim 5, characterized in that, The step of determining the updated master vector of job requirement data based on the median vector of the master-servant group of job requirement data, the mean vector of the master-servant group of job requirement data, and a preset threshold for the difference between the master-servant group vectors of job requirement data specifically includes: Determine whether the difference between the median vector of the master-servant group of the job requirement data and the mean vector of the master-servant group of the job requirement data is greater than a preset threshold for the difference between the master-servant group vectors of the job requirement data. If so, the mean vector of the master and servant groups of the job requirement data will be used as the updated master vector of the job requirement data. If not, then the midpoint vector of the master-servant group of the job requirement data will be used as the updated master vector of the job requirement data.

7. A resume matching device, characterized in that, include: The job application data and job requirement data acquisition module is used to acquire job application data and job requirement data; the job application data includes unstructured job application data, semi-structured job application data, and structured job application data; the job requirement data includes unstructured job requirement data, semi-structured job requirement data, and structured job requirement data. The job resume implicit information generation module is used to perform semantic parsing processing on the unstructured resume data, semi-structured resume data and structured resume data based on multiple preset resume data semantic parsing vectors to obtain the job resume implicit information. The job requirement implicit information generation module is used to perform semantic mining and analysis on the unstructured job requirement data, semi-structured job requirement data, and structured job requirement data to obtain the job requirement implicit information. The job resume alignment data generation module is used to perform resume data alignment processing on the unstructured resume data, semi-structured resume data, structured resume data, and implicit information of the job resume to obtain job resume alignment data. The job requirement alignment data generation module is used to perform job requirement alignment processing on the unstructured job requirement data, semi-structured job requirement data, structured job requirement data, and implicit job requirement information to obtain job requirement alignment data. The resume matching information generation module is used to perform matching calculations on the job resume alignment data and job requirement alignment data based on multiple randomly generated resume matching calculation parameters and multiple preset resume matching parameter optimization step sizes, and generate resume matching information. The multiple randomly generated resume matching calculation parameters include randomly generated resume matching calculation weight parameters, randomly generated resume matching similarity threshold parameters, and randomly generated resume matching deviation correction parameters; The preset resume matching parameter optimization step size includes a preset resume matching calculation weight parameter optimization step size, a preset resume matching similarity threshold parameter optimization step size, and a preset resume matching deviation correction parameter optimization step size. The step of performing matching calculations on the job resume alignment data and job requirement alignment data based on multiple randomly generated resume matching calculation parameters and multiple preset resume matching parameter optimization step sizes to generate resume matching information specifically includes: Calculate the similarity between the job resume alignment data and the job requirement alignment data to obtain resume matching similarity information; When the resume matching similarity information is less than or equal to the randomly generated resume matching similarity threshold parameter, the resume matching calculation weight parameter and the randomly generated resume matching deviation correction parameter are used to match the job application resume alignment data and the job requirement alignment data to generate resume matching variable information. Based on the job resume alignment data, job requirement alignment data, resume matching variable information, and the preset job resume matching database, the resume matching accuracy information is calculated. Determine whether the resume matching accuracy information is greater than a preset resume matching accuracy threshold information; If so, then generate resume matching information based on the resume matching variable information; If not, then based on the preset resume matching calculation weight parameter optimization step size, the preset resume matching similarity threshold parameter optimization step size, and the preset resume matching deviation correction parameter optimization step size, the randomly generated resume matching calculation weight parameter, the randomly generated resume matching similarity threshold parameter, and the randomly generated resume matching deviation correction parameter are optimized to obtain the optimized resume matching calculation weight parameter, the optimized resume matching similarity threshold parameter, and the optimized resume matching deviation correction parameter. The optimized resume matching calculation weight parameter is used as the randomly generated resume matching calculation weight parameter, the optimized resume matching similarity threshold parameter is used as the randomly generated resume matching similarity threshold parameter, and the optimized resume matching deviation correction parameter is used as the randomly generated resume matching deviation correction parameter. Then, the process returns to the step where, when the resume matching similarity information is less than or equal to the randomly generated resume matching similarity threshold parameter, the job resume alignment data and job requirement alignment data are matched according to the randomly generated resume matching calculation weight parameter and the randomly generated resume matching deviation correction parameter to generate resume matching variable information.

8. A terminal device, characterized in that, The terminal device includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 6.

Images