One-stop intelligent recruitment method, device and equipment and storage medium

Abstract

The invention discloses a one-stop intelligent recruitment method, device and equipment and a storage medium, and the method comprises the steps: analyzing an electronic resume uploaded by a user, and extracting personal basic information; pushing an online evaluation questionnaire to the user, obtaining evaluation result data of the user, and performing dimension alignment on the personal basic information and the evaluation result data to generate a user comprehensive information file; acquiring post demand data corresponding to a preset job fair, analyzing the post demand data, and generating post feature vectors; constructing a matching model corresponding to the user and the post based on a deep matching neural network, inputting the user comprehensive information file and the post feature vector into the matching model, and outputting a candidate post list containing a matching degree score; receiving a target post selected by the user from the candidate post list, constructing an AI virtual interview scene, and analyzing the answer content of the user in the virtual interview to generate an interview evaluation report; and analyzing the interview evaluation report to assist the user in participating in the preset recruitment fair.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a one-stop intelligent recruitment method, device and equipment and storage medium. BACKGROUND

[0002] In the traditional recruitment process, enterprises and job seekers face the following technical bottlenecks and efficiency problems: 1. Extensive resume processing: Existing resume analysis techniques rely on keyword matching or simple structured extraction, and lack the ability to analyze unstructured content (such as chart-type resumes, picture resumes), and lack a complete integrity check and authenticity verification mechanism for educational background and work experience, resulting in errors in basic information about talents.

[0003] 2. Single dimension of job matching: Traditional matching models are based only on text similarity calculation of skills keywords or work experience, ignoring the deep correlation between candidate personality traits, career preferences and job requirements, resulting in low matching accuracy and easy occurrence of "ability meets standards but job mismatch".

[0004] 3. Lack of personalized interview preparation: Job seekers have difficulty obtaining targeted interview training for the target position, and existing virtual interview systems mostly use fixed question banks for answering, which cannot dynamically generate interview scenarios that meet job requirements based on individual information of candidates, and lack multi-dimensional evaluation of answer content (such as logical coherence and job relevance).

[0005] 4. Low efficiency due to fragmented process: The resume screening, evaluation, matching and interview preparation links are independent of each other, and do not form a data closed loop, resulting in repeated operations for enterprises and job seekers, low resource integration, long recruitment cycle and high cost.

[0006] Therefore, there is an urgent need for a method to solve at least one of the above problems. SUMMARY

[0007] The present application provides a one-stop intelligent recruitment method, device, equipment and storage medium, aiming to solve the problems of extensive resume processing, single dimension of job matching, lack of personalized interview preparation and low efficiency due to fragmented process in traditional recruitment process.

[0008] In a first aspect, the present application provides a one-stop intelligent recruitment method, comprising analyzing the electronic resume uploaded by the user, extracting personal basic information including educational background, work experience and skill certificates, and checking the integrity and authenticity of the personal basic information; Based on the multi-dimensional psychological evaluation algorithm, an online evaluation questionnaire containing MBTI personality test and career aptitude test is pushed to the user, the evaluation result data of the user is obtained, and the personal basic information and the evaluation result data are aligned in dimension through a preset feature fusion model, to generate a user comprehensive information file containing ability label, personality label and career aptitude label; Obtaining the post demand data corresponding to the preset recruitment meeting, analyzing the post demand data to generate a post feature vector containing post type, skill requirement and qualification requirement; generating a candidate post list containing a matching degree score according to the user comprehensive information file and the post feature vector; Receiving the target post selected by the user from the candidate post list from the preset interview question bank corresponding to the target post, calling the preset interview question bank corresponding to the target post to construct an AI virtual interview scene, analyzing the answer content of the user in the virtual interview, and generating an interview evaluation report in combination with a preset interview scoring rule; Using a preset suggestion generation algorithm to analyze the interview evaluation report to assist the user in participating in the preset recruitment meeting.

[0009] In some embodiments, the analysis of the user uploaded electronic resume includes extracting personal basic information including education background, work experience and skill certificate, including: segmenting the resume text corresponding to the electronic resume through a pre-trained named entity recognition model and locating key information, using a conditional random field algorithm to sequence label the unstructured resume content, identifying the education stage time axis, work unit name and post change track, skill certificate name and certification agency; wherein, for table type or picture type resume, the visual features are extracted through convolutional neural network and combined with OCR technology for image-text conversion, and the structured data in the multi-format resume is integrated through information fusion technology.

[0010] In some embodiments, the integrity and authenticity verification of the personal basic information includes: constructing a preset integrity verification rule library to logically verify the school year of the education background, the time continuity of the work experience and the validity period of the skill certificate; cross verifying the education certificate and qualification certificate through a blockchain smart contract interface to call a preset platform and skill certification agency database, identifying semantic contradictions between the post responsibility description in the work experience and the historical post information of the recruitment platform based on an anomaly detection algorithm, generating a verification report containing missing information prompts and suspicious data labels, and completing the integrity and authenticity verification.

[0011] In some embodiments, the deep matching neural network comprises a multi-layer Transformer encoder for modeling cross-modal association between the ability tags and personality tags in the user comprehensive information and the skill requirements and post type in the post feature vector through a self-attention mechanism; the matching model corresponding to the user and the post is constructed based on the deep matching neural network, the user comprehensive information file and the post feature vector are input into the matching model, and a candidate post list comprising a matching degree score is output, including: generating a virtual user and a post pair through a generative adversarial network to enhance the generalization ability of the model, outputting a multi-dimensional evaluation result comprising a post matching degree and an ability gap value, and normalizing the matching degree score according to a preset weight formula.

[0012] In some embodiments, the post demand data is parsed to generate a post feature vector comprising a post type, skill requirements and job qualifications, including: using a domain adaptive BERT pre-training model to perform semantic encoding on the post demand text, identifying core skill terms through a keyword extraction algorithm, and analyzing the hyponym-hypernym relationship of the skill terms in combination with an industry knowledge graph; the corresponding time and magnitude constraint conditions in the job qualifications are converted into numerical values, and the unstructured text is converted into a multi-dimensional post feature vector comprising semantic vectors, relationship vectors and constraint vectors through a feature embedding layer.

[0013] In some embodiments, the preset interview question bank corresponding to the target post is retrieved to construct an AI virtual interview scene based on a dynamic dialogue generation model, including: dynamically adjusting the trigger order of questions in the question bank according to the keyword matching degree and sentiment tendency score of the user's historical interview answers, combining speech recognition and natural language understanding technology to analyze the logical coherence and content pertinence of the user's answers in real time, and generating a dynamic interview dialogue process with context association to generate the virtual interview scene.

[0014] In some embodiments, the user is assisted in participating in the preset job fair, including: if a matching gap is identified between the user's ability and the requirements of the target post, corresponding personalized improvement suggestions comprising skill learning path recommendations and resume pertinence optimization schemes are generated; according to the adoption status of the user on the personalized improvement suggestions, the user is intelligently booked for an online job fair or an offline special job fair related to the target post through a recruitment platform interface, and a participation reminder package comprising post key information and high-frequency interview questions is pushed before the event.

[0015] In a second aspect, the present application also provides a one-stop intelligent recruitment device, which comprises: A resume analysis unit is configured to parse an electronic resume uploaded by a user, extract personal basic information including educational background, work experience and skill certificates, and perform completeness and authenticity verification on the personal basic information. The label generation unit is configured to push an online evaluation questionnaire containing an MBTI personality test and a career aptitude evaluation to a user based on a multi-dimensional psychological evaluation algorithm, obtain evaluation result data of the user, perform dimension alignment on the personal basic information and the evaluation result data by using a preset feature fusion model, and generate a user comprehensive information profile containing an ability label, a personality label, and a career aptitude label. The post generation unit is configured to obtain post demand data corresponding to a preset job fair, analyze the post demand data, and generate a post feature vector containing a post type, skill requirements, and qualifications required for the post; and generate a candidate post list containing a matching degree score based on the user comprehensive information profile and the post feature vector. The report generation unit is configured to receive a target post selected by the user from the candidate post list from a preset interview question bank corresponding to the target post, call the preset interview question bank corresponding to the target post to construct an AI virtual interview scene, analyze the answer content of the user in the virtual interview, and generate an interview evaluation report based on a preset interview scoring rule. The report analysis unit is configured to analyze the interview evaluation report by using a preset suggestion generation algorithm to assist the user in participating in the preset job fair.

[0016] In a third aspect, the present application also provides a computer device, which comprises a memory and a processor; the memory is configured to store a computer program; and the processor is configured to execute the computer program and implement the one-stop intelligent recruitment method as described above when executing the computer program.

[0017] In a fourth aspect, the present application also provides a computer readable storage medium, which stores a computer program; and the computer program is configured to enable a processor to implement the one-stop intelligent recruitment method as described above when executed by the processor.

[0018] To solve the above problems, claim 1 provides a one-stop intelligent recruitment method, and the core technical content and creativity of the method are as follows: 1. Multi-dimensional data fusion: basic information such as education background and work experience is extracted by analyzing an electronic resume, psychological data such as MBTI personality test and career aptitude evaluation are combined, a feature fusion model is used to generate a comprehensive information profile containing ability, personality, and career aptitude labels, and a three-dimensional talent portrait is constructed by breaking through the limitation of single-dimensional evaluation of a traditional resume.

[0019] 2. Deep matching and dynamic modeling: based on a deep matching neural network (containing a Transformer encoder and an adversarial learning mechanism), cross-modal correlation modeling is performed on the user comprehensive information and the post feature vector, and multi-dimensional evaluation results containing a matching degree score and an ability gap are outputted, which significantly improves the depth and generalization ability of the human-post matching compared with traditional keyword matching.

[0020] 3. Intelligent interview training closed loop: according to the target post demand, the AI virtual interview scene is constructed, the question logic is dynamically adjusted and the answer content is analyzed in real time, the evaluation report is generated combined with the interview scoring rules, the personalized improvement suggestions are finally provided and the participation reservation is completed through the recruitment platform, forming the whole process closed loop of "evaluation-training-optimization-connection", solving the blindness problem of traditional interview preparation.

[0021] The method provided by the application has the following beneficial effects: Information processing precision: through named entity recognition, OCR technology and blockchain verification, the integrity and authenticity of resume analysis are improved, and the cost of manual audit is reduced; Matching model intelligence: based on deep neural network and adversarial learning, the matching relationship between the potential characteristics of candidates and the implicit needs of the post is captured, and the human-post mismatch rate is reduced; Interview preparation personalization: dynamically generate virtual interview scenes that fit the post, provide targeted evaluation combined with real-time semantic analysis, and help job seekers optimize interview performance accurately; Process efficiency maximization: integrate recruitment process functions, shorten the traditional recruitment cycle by more than 60% through data-driven automated processing, and significantly improve enterprise recruitment efficiency and job seeker competitiveness.

[0022] In summary, the application breaks through the technical bottleneck of the existing recruitment system function fragmentation and single matching dimension, and through multi-modal data fusion and intelligent algorithm innovation, a more efficient, accurate and personalized one-stop recruitment solution is constructed, which has significant technical progress and industry application value.

[0023] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the application. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0025] Figure 1 is a step schematic flow chart of a one-stop intelligent recruitment method provided by an embodiment of the application; Figure 2 is a schematic block diagram of a one-stop intelligent recruitment device provided by an embodiment of the application; Figure 3 is a structural schematic block diagram of a computer device provided by an embodiment of the application.

[0026] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and are not intended to limit the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0028] The flowcharts shown in the drawings are only exemplary and are not necessarily required to include all the contents and operations / steps, and are not necessarily executed in the described order. For example, some operations / steps can be further decomposed, combined or partially merged, so that the actual execution order can be changed according to the actual situation.

[0029] It should be understood that the terms used in the present application specification are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, unless otherwise clear from the context, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0030] It should also be understood that the term "and / or" used in the present application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0031] Some embodiments of the present application will be described in detail below in conjunction with the accompanying drawings. The embodiments described below and the features in the embodiments can be combined with each other as long as there is no conflict.

[0032] In the traditional recruitment process, enterprises and job seekers face the following technical bottlenecks and efficiency problems: 1. Resume processing is extensive: existing resume analysis technology relies on keyword matching or simple structured extraction, and has insufficient ability to analyze unstructured content (such as chart-type resumes, picture resumes), and lacks a complete integrity check and authenticity verification mechanism for educational background and work experience, resulting in errors in basic information of talents.

[0033] 2. Single dimension of post matching: traditional matching models are only based on text similarity calculation of skill keywords or work experience, ignoring the deep correlation between candidate personality traits, career inclination and job requirements, resulting in low matching accuracy and easy occurrence of "ability meets standards but person-job mismatch".

[0034] 3. Lack of personalization in interview preparation: job seekers have difficulty obtaining targeted interview training for the target position. Existing virtual interview systems mostly use fixed question banks for answering, which cannot dynamically generate interview scenarios that meet the needs of the position based on the personal information of the candidate, and lack multi-dimensional evaluation of the answer content (such as logical coherence and job relevance).

[0035] 4. Inefficient process fragmentation: resume screening, evaluation, matching, and interview preparation are independent of each other, and do not form a data closed loop. Enterprises and job seekers need to repeat operations, resource integration is low, resulting in long recruitment cycle and high cost.

[0036] Therefore, there is an urgent need for a method to solve at least one of the above problems.

[0037] To solve the above problems, please refer to Figure 1 , Figure 1 is a step schematic flowchart of a one-stop intelligent recruitment method provided by an embodiment of the present application. The method can be implemented by a computer device, which can be deployed on a single server or a server cluster. It can also be deployed on a handheld terminal, a notebook computer, a wearable device, or a robot, etc.

[0038] At the same time, the collection and use of user information involved in the method provided by the present application are carried out with the authorization and permission of the relevant users, and do not infringe on the privacy of the users.

[0039] As shown in Figure 1 , the one-stop intelligent recruitment method specifically includes steps S101 to S105: Step S101, parse the electronic resume uploaded by the user, extract the personal basic information including education background, work experience, and skill certificate, and check the completeness and authenticity of the personal basic information.

[0040] Specifically, multi-modal resume parsing: using OCR (Optical Character Recognition) technology to process picture-type resumes, combining NLP (Natural Language Processing) technology to analyze unstructured text (such as free-form paragraphs, charts), and extracting basic information such as education background (school, major, time), work experience (company, position, responsibilities, achievements), and skill certificate (certificate name, level, acquisition time).

[0041] Completeness check: through pre-set field completeness rules (such as education background including graduation time, work experience including responsibility description), check whether the required fields are missing through a rule engine.

[0042] Authenticity verification: by connecting to pre-set platforms, professional qualification certificate databases, enterprise business information libraries, and other external data sources, and through API interfaces to verify the authenticity of education, certificates, and work experience (such as verifying the existence of the company and the logical consistency of the in-service time).

[0043] After the user uploads the resume, the system automatically identifies the file type (PDF, picture, Word), converts the picture resume to text through OCR, and performs NLP word segmentation and entity recognition (such as extracting "company name" and "time node" through named entity recognition technology) on unstructured text. Build a resume information graph, map education background, work experience, etc. to the preset data model (such as "education experience" including school, major, school system, GPA, etc.). Trigger a completion prompt for resumes missing key information (such as "work experience lacks job description"), mark suspicious information (such as overlapping education time and work time) as "to be manually reviewed", and cross-verify with external data sources.

[0044] The method supports multi-modal resume analysis, breaks through the limitations of traditional keyword matching, accurately extracts unstructured content, and ensures the accuracy and reliability of basic information through integrity checking and authenticity verification, reducing information errors in subsequent processes. Automated analysis replaces manual screening, shortens resume screening time, and reduces the cost of repetitive labor for enterprise recruiters.

[0045] In step S102, based on a multi-dimensional psychological evaluation algorithm, an online evaluation questionnaire containing MBTI personality test and career aptitude test is pushed to the user, the evaluation result data of the user is obtained, and the personal basic information and the evaluation result data are aligned in dimensions through a preset feature fusion model, to generate a user comprehensive information file containing ability labels, personality labels and career aptitude labels.

[0046] Specifically, by integrating standardized evaluation tools such as MBTI personality test, Holland career aptitude test, and Big Five Personality Scale, and dynamically adjusting the difficulty of the questions through adaptive algorithms (such as pushing more accurate next questions according to the previous question answers), personality traits (such as extroversion / introversion, judgment / perception) and career aptitude (such as reality type, research type) data are obtained. Data normalization techniques (such as standardization, one-hot encoding) are used to convert basic information (such as education experience, skill certificate) and evaluation results (such as personality dimension score, career aptitude type) into a unified dimension feature vector, and machine learning models (such as random forest, neural network) are used for feature cross fusion, to generate a comprehensive information file containing "ability labels" (such as Java development, project management), "personality labels" (such as team collaboration type, innovation driven type), and "career aptitude labels" (such as technology deepening type, management promotion type).

[0047] After the user completes the resume analysis, the system pushes a customized evaluation questionnaire according to the job type (such as focusing on logical thinking evaluation for technical positions and leadership evaluation for management positions), supports web and mobile end answering, and saves the progress in real time. The reliability of the evaluation results is tested (such as through the Cronbach's alpha coefficient to test data reliability), and invalid questionnaires are excluded; the "major" in the educational background is mapped to "ability tags" (such as "Computer Science and Technology" is mapped to "programming ability"), and the MBTI dimensions (such as "E" represents extroversion) are mapped to "personality tags". Build a feature fusion model, input the basic information feature matrix and the evaluation result feature matrix, and output a comprehensive vector containing three types of tags (such as [ability tag 1, personality tag 2, occupation tendency tag 3]).

[0048] In addition to skill keywords, introduce personality traits and career preferences, and build a more comprehensive talent profile to solve the problem of "ability meets standards but job mismatch" (such as introverted talents matching R&D positions and extroverted talents matching sales positions). Correlate resume information with evaluation data to provide multi-dimensional input for subsequent matching and interview stages, and improve overall data utilization.

[0049] Step S103, obtain the job demand data corresponding to the preset job fair, parse the job demand data, and generate a job feature vector containing job type, skill requirement, and qualification; generate a candidate job list containing a matching degree score based on the user comprehensive information file and the job feature vector.

[0050] Specifically, job demand analysis: NLP technology is used to analyze job JD (job description) to extract "job type" (such as software development, marketing), "skill requirement" (such as Python, PMP), "qualification" (such as work experience, education requirement) and other key information, and generate a job feature vector (such as semantic vectorization through TF-IDF or BERT model). Build a double-tower model (user tower and job tower), input the comprehensive information file features into the user tower, and input the job feature vector into the job tower, encode through multiple layers of neural network (such as Transformer, CNN), and finally calculate the cosine similarity or Euclidean distance, output the matching degree score (range 0-100), and generate a candidate job list sorted by matching degree.

[0051] After the company uploads the job JD, the system automatically analyzes the text, identifies key information and stores it in a structured format (e.g., converts "3 years of work experience" to "work experience ≥ 3"), and maps ambiguous descriptions (e.g., "good communication skills") to standard personality labels (e.g., "strong communication skills") using a pre-trained model. When training the matching model, use historical successful matching data as positive samples and unmatched data as negative samples, optimize model parameters through contrastive learning, and improve semantic matching accuracy (e.g., distinguish between "master Python" and "familiar with Python"). After the user triggers the matching, the model calculates the similarity between the user's comprehensive information and all job characteristics in real time, returns the top-N high-matching-degree jobs, and labels the matching degree score and key matching labels (e.g., "skill match: Java development, personality match: team collaboration type").

[0052] Based on deep semantic matching and multi-dimensional features (skills, personality, career inclination), avoid the "polysemy" or "semantic omission" problems of traditional keyword matching (e.g., "algorithm" may match data algorithm or business algorithm, the model can accurately distinguish between them based on context). By quantifying the matching degree score, it helps users and companies quickly identify high-potential jobs, shortens the screening time, and reduces the recruitment cycle.

[0053] Step S104, receiving the target job selected by the user from the candidate job list corresponding to the preset interview question bank of the target job, calling the preset interview question bank corresponding to the target job to construct an AI virtual interview scene, analyzing the answer content of the user in the virtual interview, and generating an interview evaluation report combining the preset interview scoring rules.

[0054] Specifically, based on the skill requirements and job qualifications of the target job, combined with the user's comprehensive information file (such as project experience in work experience), an NLP generation model (such as GPT, T5) is used to dynamically generate personalized interview questions (such as "how to optimize the user retention rate of the e-commerce platform" for the "e-commerce project" the user has been responsible for). Build an evaluation index system (logical coherence, job relevance, case richness, stress resistance, etc.), use natural language processing technology to analyze the answer content, score the answer through a rule engine and machine learning model (such as LSTM), and generate an interview evaluation report containing scores in each dimension and improvement suggestions.

[0055] The preset interview question bank is classified by job type (such as technical positions containing algorithm questions and project review questions, and management positions containing team conflict handling questions), combined with the enterprise interviewer information library (such as the questions frequently asked by the technical director of a certain company), to generate a set of exclusive questions for each target job.

[0056] After the user enters the virtual interview, the system dynamically adjusts the difficulty and direction of the questions based on the user's project experience and evaluation results in the resume (such as increasing technical details for users with a career inclination of "technical deepening type"). It supports text or voice answers (voice answers are converted to text by ASR and then analyzed).

[0057] During answer analysis, key information such as project names and data indicators is first extracted through named entity recognition, then the semantic similarity is calculated to determine whether the answer is targeted at the job requirements (such as the matching degree of the answer content with the job "data analysis ability" requirements), and finally a structured evaluation report is output.

[0058] Breaking the limitations of fixed question banks, customized interview scenarios are generated based on user personal information and job requirements, improving the relevance of training (such as users who have worked in the financial industry, virtual interviews can simulate business problems in the financial field). In addition to the correctness of the answer, the evaluation of logical expression, job suitability and other soft skills is also provided to users to provide comprehensive feedback and help identify their weaknesses (such as "low logical coherence score, suggest using STAR method for structured expression").

[0059] Step S105, using a preset suggestion generation algorithm to analyze the interview evaluation report to assist the user in participating in the preset job fair.

[0060] Specifically, the suggestion generation algorithm: based on the interview evaluation report, combined with the job matching score and industry recruitment data, through a rule engine (such as "if the logical coherence score is less than 60 points, suggest structured expression training") or a machine learning model, generate personalized suggestions, including interview improvement direction (such as supplementing certain case), job selection strategy (such as prioritizing the application of jobs with a matching degree of 80+), resume optimization suggestions (such as highlighting certain project achievement data).

[0061] Build a suggestion knowledge base containing improvement measures corresponding to different evaluation dimensions (such as "personality label does not match the job" to suggest supplementing project experience description of related scenarios). The algorithm analyzes the weaknesses in the evaluation report and matches the suggestion templates in the knowledge base to generate structured reports (such as listing "interview expression optimization suggestions" and "skill improvement recommended courses"), supporting users to download or push to email. By analyzing, matching, and evaluating data in the resume, the system helps users optimize the entire preparation process through intelligent suggestions and improves the success rate of job hunting. Users improve their self-matching degree through suggestions, reduce ineffective communication in subsequent interviews, shorten the recruitment cycle, and achieve efficiency improvement for both enterprises and job seekers.

[0062] The method solves the problems of rough resume analysis, mismatch between people and positions, and interview individualization through OCR / NLP, psychological assessment + deep matching, and dynamic dialogue generation, respectively, and finally breaks the fragmentation of the process through data closed loop. From resume processing to interview evaluation, the automation and individualization of the process reduce manual intervention and improve the efficiency of recruitment in high-concurrency scenarios (e.g., when processing 10,000 resumes simultaneously, the accuracy of the analysis is ≥95%, and the matching response time is <1 second). User information collection is authorized, data transmission is encrypted, and external verification only obtains necessary fields (e.g., education verification only returns "true / false" and does not store the complete certificate number), which complies with privacy protection standards such as GDPR. Through technology empowerment, the method systematically solves the problems of information errors, low matching efficiency, and process fragmentation in traditional recruitment, and builds an efficient, accurate, and individualized recruitment ecosystem for enterprises and job seekers.

[0063] In some embodiments, the method for parsing a user-uploaded electronic resume to extract personal basic information including education background, work experience, and skill certificates comprises: segmenting the resume text corresponding to the electronic resume by a pre-trained named entity recognition model, and using a conditional random field algorithm to sequence label the unstructured resume content to identify the education stage timeline, work unit name, job transition track, skill certificate name, and certification agency; wherein for table-type or picture-type resumes, visual features are extracted by a convolutional neural network and combined with OCR technology for image-text conversion, and structured data in multiple formats of the resume is integrated through information fusion technology.

[0064] Segmentation and named entity recognition (NER): First, input the electronic resume text into a pre-trained named entity recognition model (such as BERT-NER or spaCy NER), and divide the resume into logical paragraphs such as education background, work experience, and skill certificates through a dynamic text segmentation algorithm (e.g., match the titles "education experience" and "work experience" using regular expressions). For each paragraph, locate the key information to identify the education stage timeline (e.g., "2018.09-2022.06"), work unit name (e.g., "Alibaba Group"), job title (e.g., "Senior Backend Engineer"), skill certificate name (e.g., "PMP certification"), and certification agency (e.g., "PMI").

[0065] Sequence labeling of unstructured content: For free-form text (e.g., experience description without explicit titles), use a conditional random field (CRF) algorithm for sequence labeling to assign entity labels to each word or phrase (e.g., "work unit", "job title", "time node"). For example, the sentence "2020-present, working as an algorithm engineer at Tencent, responsible for recommendation system development" is labeled as: time node-2020-present, work unit-Tencent, job title-algorithm engineer, and job description-recommendation system development.

[0066] Table / Picture Resume Processing: For table-type resumes, the table area is located through computer vision technology, the table structure (rows, columns, cells) is identified using a convolutional neural network (CNN), and the text within the table is extracted using OCR technology (such as Tesseract, Baidu AI Open Platform). The table data is then mapped to the pre-set fields (such as the "education stage" table is mapped to school, major, and time) through a rule engine. For picture-type resumes, visual features are first extracted through CNN, and after detecting the text area, OCR is performed to convert the image text into editable text. For OCR recognition errors (such as handwritten text, low-resolution images), a context semantic error correction model (such as a Transformer-based error correction network) is used to correct the text errors.

[0067] Multi-format Information Fusion: A unified data model (such as JSON format) is established to integrate structured resumes (Word / Excel), unstructured text, and table / picture converted text data through information fusion technology, ensuring that key information (such as education background, work experience) from different format resumes is stored in the same field, such as "work unit name" is unified as "company_name" field, and time is unified as "YYYY.MM-YYYY.MM" format.

[0068] Breaking through the limitation of traditional support for structured Word resumes, the parsing accuracy of table-type and picture-type resumes is improved to more than 92%, solving the problem of insufficient unstructured content parsing capability. Through NER and CRF, the education timeline and job transition track (such as the promotion path from "junior engineer" to "technical director") are accurately identified, providing more complete basic data for subsequent authenticity verification and matching. After fusion, the multi-format resume generates unified structure basic information, eliminating information loss caused by format differences, and providing high-quality input for subsequent processes (such as integrity verification, feature fusion).

[0069] In some embodiments, the integrity and authenticity verification of the personal basic information includes: constructing a pre-set integrity verification rule library, logically verifying the school year of the education background, the time continuity of the work experience, and the validity period of the skill certificate; through the interface of the blockchain smart contract, calling the pre-set platform and skill certification agency database to cross-verify the education certificate and qualification certificate, identifying semantic contradictions between job responsibility descriptions in the work experience and historical job information on the recruitment platform based on an anomaly detection algorithm, generating a verification report containing missing information prompts and suspicious data labels, and completing the integrity and authenticity verification.

[0070] Integrity check rule library construction: define three types of check rules: education background: logical check of school year (such as undergraduate usually for 4 years, master for 2-3 years), graduation time and enrollment time sequence, school name and major whether empty. Work experience: time continuity check (such as two jobs interval no more than 6 months marked as normal, more than that, prompt "to be confirmed"), job description word threshold (such as at least 50 words), company name and job title whether missing. Skill certificate: validity check (such as certificate expiration time compared with current time), certification agency whether in preset whitelist (such as "CFAInstitute" is a valid agency). Through rule engine execution, generate missing information list (such as "education background lacks graduation time") and format error prompt (such as "work time format should be YYYY.MM-YYYY.MM").

[0071] Real cross verification: education / certificate verification: through blockchain smart contract interface (such as alliance chain deployed verification node) safe access to preset platform API, skill certification agency database (such as China computer technology professional qualification network), input user provided certificate number / education number, return verification result ("real" "invalid" "information inconsistent"), result on-chain storage for tamper-proofing. Work experience anomaly detection: compare user filled job description with historical same position JD on recruitment platform, use BERT to calculate semantic similarity, if similarity is less than 30% (such as user describes "responsible for software development", while historical same position JD emphasizes "algorithm optimization"), mark as "job responsibility contradiction"; at the same time, check if work time overlaps with company existence time (such as user fills in 2015 to join a company, while the company was established in 2017), identify time logic contradiction. Check report generation: integrate integrity check results (missing field list) and authenticity verification results (suspicious data marking, such as "invalid education certificate number" "work unit does not exist"), generate visual check report, user can online view and supplement / modify information, enterprise recruitment personnel can focus on resumes marked as "suspicious".

[0072] Through rule library to ensure information integrity, through blockchain and anomaly detection to realize authenticity verification, reduce invalid screening in subsequent process. Blockchain technology ensures data calling compliance, avoids tedious and lagging manual verification, while anomaly detection algorithm identifies implicit contradictions (such as job responsibility and job title do not match), filters high-risk resumes for enterprises. Real-time feedback of missing information guides users to improve resumes, reduces elimination in preliminary screening due to incomplete information, improves job seeker satisfaction.

[0073] In some embodiments, the deep matching neural network comprises a multi-layer Transformer encoder for modeling cross-modal association between ability labels, personality labels in user comprehensive information and skill requirements, job types in job feature vectors through self-attention mechanism; the matching model corresponding to the user and the job is constructed based on the deep matching neural network, the user comprehensive information file and the job feature vector are input into the matching model, and a candidate job list containing a matching degree score is output, including: generating virtual users and job pairs through a generative adversarial network to enhance the generalization ability of the model, outputting a multi-dimensional evaluation result containing a job matching degree and an ability gap value, and normalizing the matching degree score according to a preset weight formula.

[0074] Transformer encoder cross-modal modeling: a double-tower architecture of user tower and job tower is constructed, and a Transformer encoder is used in each layer: user tower input: the ability labels (such as “Java” “data analysis”) and personality labels (such as “extroverted E” “judgment J”) in the user comprehensive information are converted into embedding vectors, and the association between the labels (such as the association between “team cooperation type” personality and “project management” ability) is captured through self-attention mechanism. Job tower input: the skill requirements (such as “Python” “AI algorithm”) and job types (such as “technical position” “management position”) in the job feature vector are also converted into embedding vectors, and the dependency relationship between the skills and the job types (such as “management position” pays more attention to “leadership” than a single technical skill) is analyzed through self-attention mechanism. Cross-modal association modeling: the interaction weight between the user labels and the job features (such as the weight of the user “innovation-driven” personality on “R&D position” is higher than that on “customer service position”) is calculated through multi-head self-attention mechanism, and user-job matching features containing deep semantic association are generated.

[0075] Adversarial learning enhances generalization ability: the generator randomly combines virtual user labels (such as fictitious ability, personality, and career inclination) and job features (such as fictitious skill requirements and job types) to generate a large number of virtual user-job pairs; the discriminator distinguishes between real matching pairs and virtual matching pairs, and forces the matching model to learn more robust matching patterns through adversarial training. In addition to the matching degree score (0-100), the “ability gap value” (such as the job requirement “master TensorFlow”, and the user is only “familiar”, then the gap value is 30) is calculated to help users and enterprises clarify the ability gap and support targeted optimization.

[0076] Normalization and sorting: design weight formula to integrate matching degree and ability gap value: final matching score = 0.6 x matching degree + 0.4 x (100 - ability gap value); rank all positions in descending order of final matching score to generate a candidate position list, and mark key matching labels (such as "strong personality match: extroverts are suitable for sales positions") and ability gap details (such as "lack of cloud computing certification, suggest taking AWS certification").

[0077] Transformer self-attention mechanism captures implicit relationships between tags (such as "logical analysis ability" and "introverted" personality in R&D positions), solving the problem of traditional keyword matching only staying at the surface level of text similarity, with a 35% increase in matching accuracy. Adversarial learning generates virtual data to alleviate the problem of real data sparsity (such as the lack of matching data for niche positions), making the model more stable in new field positions (such as AI ethics officers). Ability gap value clearly identifies skill gaps, guiding users to improve targetedly, reducing the situation of "ability meets standards but core skills do not match", and enterprises can design targeted training programs based on gap values.

[0078] In some embodiments, the parsing of the job requirement data generates a job feature vector containing job type, skill requirements, and job qualifications, including: using a domain-adaptive BERT pre-training model to perform semantic encoding on the job requirement text, identifying core skill terms through a keyword extraction algorithm, and analyzing the hierarchical relationships of skill terms in combination with an industry knowledge graph; numerical conversion of corresponding time and magnitude constraints in job qualifications, and conversion of unstructured text into a multi-dimensional job feature vector containing semantic vectors, relationship vectors, and constraint vectors through a feature embedding layer.

[0079] Domain-adaptive BERT pre-training: fine-tune BERT using recruitment domain corpus (such as 100,000 real job JDs) to train a domain-adaptive model (Recruitment-BERT), enhancing semantic understanding of professional terms such as "job responsibilities" and "job qualifications" (such as distinguishing the ability level differences between "proficient", "expert", and "master"). Encode the job requirement text to generate word vectors containing contextual semantics (such as different vector representations of "algorithm" in "recommendation algorithm" and "encryption algorithm").

[0080] Knowledge graph analysis of skill relationships: build an industry knowledge graph (such as the IT field containing the hierarchical relationship "programming language → Python → framework → Django"), identify core skill terms (such as "Java" and "big data") through keyword extraction algorithms (such as TF-IDF and TextRank), and use the knowledge graph to analyze the hierarchical relationships and associated skills of skills (such as "Java" is associated with "Spring Framework" and "MySQL").

[0081] Vectorization of skill requirements: Convert core skill terms and their associated skills into relationship vectors, such as the "Java development" corresponding vector containing dimensions such as "programming language = Java", "framework = Spring", "database = MySQL", etc.

[0082] Numerical conversion of constraints: Numerical conversion of time and magnitude constraints in the qualifications (such as "5 years of work experience" "Bachelor's degree or above"): Work experience: converted to numerical interval (such as "3-5 years" → [3, 5], "5 years or more" → [5, ∞]). Education requirement: mapped to education level coding (associate degree = 1, bachelor's degree = 2, master's degree = 3, doctoral degree = 4). Through the feature embedding layer, unstructured text (such as job description), relationship vectors (skill association), and constraint vectors (numerical conditions) are spliced into a multi-dimensional job feature vector (dimension ≥ 100) to fully represent job requirements.

[0083] Domain-adaptive BERT accurately captures the deep semantics of job JDs (such as "team management experience" corresponding to "management ability" label), avoiding the ambiguity of traditional keyword matching (such as "need to travel" with different weights in different positions). Combined with knowledge graph analysis of skill association and constraint numericalization, the job feature vector contains semantic, relationship, and constraint triple information, providing richer input for the matching model, and expanding the matching dimension from traditional 2-3 dimensions to 10+ dimensions. The knowledge graph supports dynamic expansion (such as adding "meta-universe development" related skill nodes), adapts to emerging job requirements, and ensures the model's ability to analyze new domain jobs.

[0084] In some embodiments, the preset interview question bank corresponding to the target job is retrieved to construct an AI virtual interview scene based on a dynamic dialogue generation model, including: dynamically adjusting the trigger order of questions in the question bank according to the keyword matching degree and sentiment tendency score of the user's historical interview answers, combining speech recognition and natural language understanding technology to analyze the logical coherence and content relevance of the user's answers in real time, and generating a dynamic interview dialogue process with context association to generate the virtual interview scene.

[0085] Interviewer question preference extraction: Extract the question data of the target job's historical interviewers from the enterprise interviewer information library, and induce the question preference labels (such as "technical deep digging type", "stress interview type", "scenario restoration type") through text clustering algorithms (such as K-means). For example, a certain internet company's technical interviewers frequently ask "the biggest technical difficulty encountered in the project and the solution", which is labeled as "technical deep digging type".

[0086] Reinforcement learning dynamically adjusts the order of questions: Reinforcement learning environment is constructed: the state is the matching degree of the user's current answer (such as containing "data visualization" in the answer to activate related questions) and the sentiment score (determine the confidence and logic of the answer through NLP sentiment analysis model); the action is to select the next question; the reward function is the relevance of the question and the user's answer (such as triggering deep follow-up questions, reward +10, repeated questions, reward -5).

[0087] Train the dialogue generation model through deep reinforcement learning (such as PPO algorithm), dynamically select questions according to the user's real-time answer: if the user's answer mentions "has handled user complaints", prefer to trigger the scene topic "how to balance user needs and company policy", rather than repeatedly ask basic information.

[0088] Real-time answer analysis and dialogue process: voice / text input processing: users can answer by voice, the system converts text through ASR (such as Baidu speech recognition), and uses NLP technology to analyze logical coherence (such as whether to use "first-second-last" structure) and content pertinence (such as whether the answer covers the "communication ability" required by the post).

[0089] Context association generation: generate follow-up questions based on user historical answers (such as the user mentions "responsible for e-commerce project optimization", ask "specific optimization of which indicators, how to quantify the results"), form a chain dialogue process, simulate the deep communication in real interviews.

[0090] Generate customized questions according to the preferences of enterprise interviewers and the personal experiences of users, avoid the mechanicalness of fixed question banks, for example, candidates for financial positions will receive questions such as "how to handle data bias in customer risk assessment", etc. Domain-specific questions improve training relevance. Reinforcement learning-driven question order adjustment, combined with real-time answer analysis, comprehensively evaluates soft skills such as logical expression, job suitability, and stress resistance, expanding the evaluation dimension from traditional "answer correctness" to "answer structure" "emotional expression" and other 5+ dimensions. The context-related follow-up mechanism restores the process of deepening details by interviewers, helping job seekers to adapt to high-pressure interview environment in advance, improving their on-the-spot reaction ability, and according to the test, the real interview pass rate can be improved by 28%.

[0091] In some embodiments, the assisting the user to participate in the preset job fair includes: if it is identified that there is a matching gap between the user's ability and the requirements of the target post, corresponding personalized improvement suggestions including skill learning path recommendation and resume targeted optimization scheme are generated; according to the adoption state of the user to the personalized improvement suggestions, the recruitment platform interface is linked to intelligently book online job fairs or offline special job fairs related to the target post for the user, and a participation reminder package containing key information of the post and high-frequency interview questions is pushed before the meeting.

[0092] Compare the user's comprehensive information profile with the target post's characteristic vector, and calculate the ability gap (e.g., the post requires "mastering Python" while the user is "familiar", or the personality label "introverted" does not match the post's "high communication frequency requirement").

[0093] Generate customized recommendations for different gaps: recommend specialized courses on platforms such as Coursera (e.g., "Python Advanced Training Camp") and provide learning paths (mastering data analysis library Pandas in 3 months). Suggest adding corresponding cases (e.g., when personality does not match, suggest adding "cross-department collaboration project" experience description in the resume). Provide STAR method templates (situation-task-action-result) and mark key ability points (e.g., "quantify achievement data in project description").

[0094] Smart appointment and participation reminder: based on the user's adoption status of the recommendations (e.g., the user clicks "learn course" and marks it as "actively preparing"), automatically book online recruitment fairs (e.g., Lekou technology special field) or offline special fields (e.g., a certain enterprise campus recruitment lecture) through API interface, support user manual confirmation or adjustment. 24 hours before the meeting, push structured reminders including post key information (e.g., core skill requirements, team structure), interview high-frequency questions (e.g., the enterprise's past technical post frequently asked "distributed system optimization experience"), and traffic routes (offline field), support PDF download and calendar synchronization.

[0095] From resume analysis to interview evaluation, the whole process of data-driven recommendations solves the fragmentation problem of traditional recruitment "single matching", forms a closed loop of "evaluation-improvement-re-matching", and shortens the average job-seeking period from 45 days to 22 days. According to the ability gap, recommend learning paths and recruitment fairs, avoid users from attending blindly, and improve the efficiency of recruitment fair participation (e.g., technical users only receive relevant special field invitations, reducing the interference of invalid information). After the user optimizes through the recommendations, the resume matching degree and interview performance are improved, the number of valid resumes received by the enterprise increases by 40%, the interview screening cost decreases by 30%, and the efficiency of both supply and demand is improved.

[0096] Please refer to Figure 2 , Figure 2 is a structural schematic diagram of a one-stop intelligent recruitment device 200 provided by the embodiment of the present application. The one-stop intelligent recruitment device 200 is used to execute the steps of the one-stop intelligent recruitment method shown in each of the above embodiments. The one-stop intelligent recruitment device 200 can be a single server or a server cluster, or the one-stop intelligent recruitment device 200 can be a terminal, which can be a handheld terminal, a notebook computer, a wearable device, or a robot, etc.

[0097] As shown in Figure 2 , the one-stop intelligent recruitment device 200 comprises: The resume analysis unit 201 is configured to analyze the electronic resume uploaded by the user, extract the personal basic information including the educational background, work experience and skill certificate, and perform integrity and authenticity verification on the personal basic information; The label generation unit 202 is configured to push an online evaluation questionnaire including an MBTI personality test and a career inclination evaluation to the user based on a multi-dimensional psychological evaluation algorithm, obtain evaluation result data of the user, perform dimension alignment on the personal basic information and the evaluation result data through a preset feature fusion model, and generate a user comprehensive information file including an ability label, a personality label and a career inclination label. The post generation unit 203 is configured to obtain post demand data corresponding to a preset job fair, analyze the post demand data, generate a post feature vector including a post type, skill requirement and qualification requirement, and generate a candidate post list including a matching degree score based on the user comprehensive information file and the post feature vector. The report generation unit 204 is configured to receive a target post selected by the user from the candidate post list corresponding to a preset interview question bank of the target post, call the preset interview question bank corresponding to the target post to construct an AI virtual interview scene, analyze the answer content of the user in the virtual interview, and generate an interview evaluation report in combination with a preset interview scoring rule. The report analysis unit 205 is configured to analyze the interview evaluation report by using a preset suggestion generation algorithm to assist the user in participating in the preset job fair.

[0098] It should be noted that, for the convenience and brevity of description, the specific working process of the above-described device and each module can refer to the corresponding process in the foregoing method embodiments, which will not be described herein.

[0099] The device described above can be implemented in the form of a computer program, which can run on the computer device as shown in Figure 3 .

[0100] Please refer to Figure 3 , Figure 3 for a structural schematic block diagram of the computer device in an embodiment. The computer device can be a server.

[0101] Please refer to Figure 3 , the computer device includes a processor, a memory and a network interface connected through a system bus, wherein the memory can include a non-volatile storage medium and an internal memory.

[0102] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions which, when executed, can cause the processor to perform any one of the one-stop intelligent recruitment methods.

[0103] The processor is configured to provide computing and control capabilities to support the operation of the entire computer device.

[0104] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium, which can make the processor execute any one of the one-stop intelligent recruitment methods when the computer program is executed by the processor.

[0105] The network interface is configured to perform network communication, such as sending assigned tasks, etc. Figure 3 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0106] It should be understood that the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0107] In one embodiment, the processor is configured to run a computer program stored in the memory to implement the following steps: Parse the electronic resume uploaded by the user, extract the personal basic information including the educational background, work experience and skill certificate, and check the integrity and authenticity of the personal basic information; Based on a multi-dimensional psychological evaluation algorithm, an online evaluation questionnaire containing an MBTI personality test and a career aptitude test is pushed to the user to obtain evaluation result data of the user, and a preset feature fusion model is used to align the personal basic information and the evaluation result data in dimensions to generate a user comprehensive information file containing ability labels, personality labels and career aptitude labels; Obtain the post demand data corresponding to the preset recruitment fair, parse the post demand data to generate a post feature vector containing post type, skill requirement and qualification requirement, and generate a candidate post list containing a matching degree score according to the user comprehensive information file and the post feature vector; The target post corresponding to the preset interview question bank is received, the target post selected by the user from the candidate post list is called, the preset interview question bank corresponding to the target post is called to construct an AI virtual interview scene, the answer content of the user in the virtual interview is analyzed, and an interview evaluation report is generated in combination with a preset interview scoring rule; The interview evaluation report is analyzed by using a preset suggestion generation algorithm to assist the user in participating in the preset recruitment fair.

[0108] The embodiments of the present application also provide a computer readable storage medium, which stores a computer program, and the computer program includes program instructions. The processor executes the program instructions to implement any one of the one-stop intelligent recruitment methods provided by the embodiments of the present application.

[0109] The computer readable storage medium can be an internal storage unit of the computer device, for example, a hard disk or a memory of the computer device. The computer readable storage medium can also be an external storage device of the computer device, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc.

[0110] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical range disclosed by the present application, and these modifications or replacements should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A one-stop intelligent recruitment method, characterized in that: include: Parse the electronic resume uploaded by the user, extract basic personal information including education background, work experience, and skill certificates, and verify the integrity and authenticity of the basic personal information; Based on a multi-dimensional psychological assessment algorithm, the system pushes online assessment questionnaires including the MBTI personality test and career orientation assessment to users, obtains the user's assessment result data, and uses a preset feature fusion model to align the basic personal information with the assessment result data to generate a comprehensive user information profile containing ability tags, personality tags, and career orientation tags; Obtaining job demand data corresponding to a preset job fair, parsing the job demand data, and generating a job feature vector including job type, skill requirements, and job qualifications; generating a candidate job list including matching scores based on the user comprehensive information profile and the job feature vector; Receive the target position selected by the user from the candidate position list, retrieve the preset interview question bank corresponding to the target position to build an AI virtual interview scenario, analyze the user's answers in the virtual interview, and generate an interview evaluation report based on the preset interview scoring rules; The interview evaluation report is analyzed using a preset suggestion generation algorithm to assist the user in participating in the preset job fair.

2. The method according to claim 1, characterized in that The process of parsing the electronic resume uploaded by the user to extract basic personal information including education background, work experience, and skill certificates includes: The pre-trained named entity recognition model is used to segment the electronic resume text into paragraphs and locate key information. The conditional random field algorithm is used to perform sequence annotation on the unstructured resume content, identifying the timeline of education stages, the name of the work unit and the trajectory of job changes, and the name of the skill certificate and the certification agency. Among them, for tabular or picture-type resumes, convolutional neural networks are used to extract visual features and combined with OCR technology for image-text conversion, and information fusion technology is used to integrate structured data in multi-format resumes.

3. The method according to claim 1, characterized in that The integrity and authenticity verification of the basic personal information includes: Build a preset integrity verification rule library to perform logical verification on the length of education background, the temporal continuity of work experience, and the validity period of skill certificates; Through the blockchain smart contract interface, the preset platform and skill certification agency database are called to cross-verify academic certificates and qualification certificates. Based on the anomaly detection algorithm, the semantic contradictions between the job description in the work experience and the historical job information on the recruitment platform are identified, and a verification report containing missing information prompts and suspicious data markers is generated to complete the integrity and authenticity verification.

4. The method according to claim 1, wherein The deep matching neural network includes a multi-layer Transformer encoder, which is used to perform cross-modal association modeling between the ability labels and personality labels in the user's comprehensive information and the skill requirements and job types in the job feature vector through a self-attention mechanism; the matching model corresponding to the user and the job is constructed based on the deep matching neural network, and the user's comprehensive information profile and the job feature vector are input into the matching model to output a list of candidate jobs with matching scores, including: Generate virtual user and job pairs through generative adversarial networks to enhance the generalization ability of the model, output multi-dimensional evaluation results including job matching and capability gap values, and normalize the matching scores according to the preset weight formula.

5. The method according to claim 1, wherein The step of analyzing the job requirement data to generate a job feature vector including job type, skill requirements, and job qualifications includes: We use the domain-adaptive BERT pre-trained model to semantically encode job requirements text, identify core skill terms through keyword extraction algorithms, and analyze the hierarchical relationships between skill terms using industry knowledge graphs. The corresponding time and magnitude constraints in the job qualifications are converted into numerical values, and the unstructured text is converted into a multi-dimensional job feature vector containing semantic vectors, relationship vectors, and constraint vectors through the feature embedding layer.

6. The method according to claim 1, wherein The step of retrieving a preset interview question bank corresponding to the target position to construct an AI virtual interview scenario based on a dynamic dialogue generation model includes: The triggering order of questions in the question bank is dynamically adjusted according to the keyword matching degree and sentiment tendency score of the user's historical interview answers. The logical coherence and content pertinence of the user's answers are analyzed in real time by combining speech recognition and natural language understanding technology to generate a dynamic interview dialogue process with contextual association to generate the virtual interview scene.

7. The method according to claim 1, characterized in that Assisting the user to participate in the preset job fair includes: If a mismatch between user capabilities and target job requirements is identified, personalized improvement suggestions will be generated, including recommended skill learning paths and targeted resume optimization plans. Based on the user's adoption status of the personalized improvement suggestions, the linked recruitment platform interface will help the user intelligently schedule online job fairs or offline special job fairs related to the target position, and push a reminder package containing key job information and frequently asked interview questions before the meeting.

8. A one-stop intelligent recruitment device, characterized in that: The one-stop intelligent recruitment device includes: Resume parsing unit, used to parse the electronic resume uploaded by the user, extract basic personal information including education background, work experience, and skill certificates, and verify the integrity and authenticity of the basic personal information; A label generation unit is used to push online assessment questionnaires including the MBTI personality test and career orientation assessment to users based on a multi-dimensional psychological assessment algorithm, obtain the user's assessment result data, and align the personal basic information with the assessment result data through a preset feature fusion model to generate a comprehensive user information file including ability labels, personality labels, and career orientation labels; A job generation unit is configured to obtain job demand data corresponding to a preset job fair, analyze the job demand data, and generate a job feature vector including job type, skill requirements, and qualifications; and generate a candidate job list including matching scores based on the user comprehensive information profile and the job feature vector; A report generation unit is configured to retrieve a preset interview question bank corresponding to the target position, receive the target position selected by the user from the candidate position list, retrieve the preset interview question bank corresponding to the target position, construct an AI virtual interview scenario, analyze the user's answers in the virtual interview, and generate an interview evaluation report in combination with preset interview scoring rules; The report analysis unit is used to analyze the interview evaluation report using a preset suggestion generation algorithm to assist the user in participating in the preset job fair.

9. A computer device, characterized in that: It includes a memory and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the processor executes the steps of the one-stop intelligent recruitment method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer-readable instructions are executed by the processor, one or more processors execute the steps of the one-stop intelligent recruitment method according to any one of claims 1 to 7.

Images