Intelligent label generation and retrieval system based on AI resume

Abstract

The invention belongs to the technical field of computers, and particularly discloses and provides an intelligent label generation and retrieval system based on AI resumes, which comprises the steps of dynamically extracting and optimizing skill labels through an intelligent label generation mechanism, and combining core skills with extended associated skills to ensure that a matching process covers more comprehensive skill dimensions; by integrating external data sources in real time to calculate popularity and association strength of skills, weight values are dynamically updated, it is ensured that candidate sets are focused on current industry hotspots, and practicability and representativeness of tags are enhanced; by binding a unique identifier and optimizing a label candidate set, standardized database storage is realized, rapid query and expandability are supported, a subsequent retrieval process is simplified, and data consistency and traceability are ensured; through fusion of an artificial feedback mechanism, a label candidate set is directly intervened to carry out weight addition, deletion or adjustment operation, a closed-loop optimization system is constructed, and the flexibility of label generation and the controllability of a user are remarkably enhanced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of computers and relates to an intelligent label generation and retrieval system based on AI resumes. BACKGROUND

[0002] In the field of resume processing and job matching, traditional systems face many challenges. The core problem lies in the low efficiency and insufficient accuracy of resume and job description matching. The resumes submitted by users often contain a large amount of text information, but existing methods cannot accurately extract key skill items, resulting in a time-consuming and error-prone resume screening process. For example, the employer needs to identify qualified candidates from a large number of resumes, but the diversity and non-standardization of resume text make it difficult for automated tools to capture the subtle differences in actual skills such as professional skills, work experience, and educational background. At the same time, the job description input by the user may contain ambiguous or industry-specific terms, and traditional parsing algorithms cannot effectively handle these dynamic content, causing the matching results to deviate from the actual needs. This ultimately leads to delays in the recruitment process, insufficient candidate pools, or misjudgment of qualified personnel, affecting the quality of enterprise recruitment and user experience.

[0003] To address these issues, traditional solutions mainly rely on simple keyword-based matching or semi-automatic manual review. In the keyword matching method, the system uses a predefined word library to scan the resume text, extracts common terms such as programming languages or certificate names, and directly compares them with keywords in the job description. This method usually combines rule engines or basic machine learning models to generate matching scores based on word frequency statistics. In addition, the semi-automatic solution involves manual intervention: recruiters manually review resumes, mark potential skill points, and supplement information through database queries.

[0004] The above traditional methods provide preliminary automation support in early applications, but are essentially limited to static rules and fixed data sets, and cannot adapt to rapidly changing industry needs. Specifically, the following disadvantages exist: 1. Insufficient matching accuracy, as the keyword method ignores contextual semantics and skill interrelationships, leading to mis-matching or omissions. For example, skill items mentioned in the resume may be processed in isolation, failing to recognize their combined value with related technologies, resulting in the filtering of excellent candidates.

[0005] 2. Lack of real-time dynamic adjustment capability, relying on static word libraries or historical data that cannot respond to current job market changes, which makes the matching results lag; at the same time, the manual review process introduces subjectivity and inconsistency, leading to prolonged processing time and increased costs.

[0006] These disadvantages collectively exacerbate system inefficiency: in large-scale resume processing, error rates rise, resources are wasted, and ultimately recruitment efficiency and user satisfaction are compromised. Therefore, innovative methods are needed to address these deep-seated problems. SUMMARY

[0007] To overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purposes, the present application proposes the following technical scheme: an intelligent label generation and retrieval system based on AI resume, comprising the following contents: an initial data set generation module, obtaining a to-be-processed resume and a user retrieval condition, generating an initial data set containing a core skill label and a user expected label set.

[0008] An associated skill query module, based on the core skill label in the initial data set, queries and calculates the associated strength value using an industry knowledge graph, and generates an associated skill set.

[0009] A dynamic weight calculation module, combining real-time job requirement data and the associated skill set, calculates the demand heat value and dynamic weight value of each skill, and generates a dynamic weight label candidate set.

[0010] A label optimization module, which adjusts the dynamic weight label candidate set by fusing artificial feedback data, generates an optimized label candidate set.

[0011] A data storage module, which generates a unique identifier for the to-be-processed resume, and stores the optimized label candidate set and the unique identifier in the label database.

[0012] A matching degree calculation module, which calculates the aggregation weight of the intersection skill according to the user expected label set in the initial data set and the optimized label candidate set, and generates a resume matching degree.

[0013] An adaptation result generation module, which generates a job adaptation result containing a unique identifier, a resume matching degree and a pass status based on the comparison between the resume matching degree and a preset qualified threshold.

[0014] Compared with the prior art, the present application has the following advantages: (1) The present application dynamically extracts and optimizes skill labels through an intelligent label generation mechanism, combines core skills and extended associated skills, ensures that the matching process covers a more comprehensive skill dimension, avoids the semantic loss or context neglect problem caused by isolated keyword processing in traditional methods, and makes the matching result closer to the actual application scenario.

[0015] (2) The present application dynamically updates the weight value by calculating the heat and association strength of the skill by real-time integration of external data sources such as recruitment platform requirements, dynamically optimizes the process by automatically filtering high-weight skills through algorithm, ensures that the candidate set focuses on the current industry hotspots, enhances the practicality and representativeness of the label, and this dynamic support for continuous self-improvement enables the label candidate set to respond to market changes and automatically adapt to emerging skill trends, thereby enabling the system to maintain high efficiency in long-term application.

[0016] (3) The application realizes standardized database storage by binding unique identifiers with optimized label candidate sets, supports fast query and scalability, simplifies the subsequent retrieval process, and ensures data consistency and traceability. At the same time, by fusing the artificial feedback mechanism, the user can directly intervene in the label candidate set to add, delete or adjust the weight operation, and these feedbacks are used to update the knowledge graph and recalculate the weight, building a closed-loop optimization system, which significantly enhances the flexibility and user controllability of label generation. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0018] Figure 1 The figure shows the schematic diagram of the system module connection of the application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the application will be described clearly and completely in the following with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only some embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0020] Please refer to Figure 1 The application proposes an intelligent label generation and retrieval system based on AI resume, which includes an initial data set generation module, an associated skill query module, a dynamic weight calculation module, a label optimization module, a data storage module, a matching degree calculation module and an adaptive result generation module.

[0021] The initial data set generation module, the associated skill query module, the dynamic weight calculation module, the label optimization module, the data storage module, the matching degree calculation module and the adaptive result generation module are connected in sequence.

[0022] The initial data set generation module acquires the resume to be processed and the user retrieval condition, and generates an initial data set containing core skill labels and user expected label set.

[0023] In a preferred embodiment, the step of generating the initial dataset containing the core skill label and the user desired label set comprises: receiving a resume file uploaded by the user, the file format including PDF or Word document, parsing the content of the resume file through natural language processing technology, identifying the skill items, work experience paragraphs and education background descriptions in the text, extracting them as independent text elements to form the core skill label.

[0024] The core skill label contains skill proficiency levels.

[0025] Receiving user input position description text, and identifying key terms as user desired skill labels through keyword extraction algorithm to form a user desired label set, the user desired skill label being a unique string element to avoid duplication.

[0026] The user desired label set contains skill proficiency levels required by the position.

[0027] The core skill label and the user desired label set are encapsulated into an initial dataset.

[0028] The core skill label refers to the specific description of skill items, work experience and education background extracted from the resume text, with the attribute of a string list, each skill item representing an independent technical ability, and the setting basis being based on the pre-trained model of the open source NLP library (such as spaCy), which is fine-tuned and trained using 5000 public resume samples to ensure parsing accuracy.

[0029] The user desired label set refers to the keyword set extracted from the position description text, with the attribute of a unique string set, and the setting basis being the text processing rules input directly by the user, which are based on the term frequency-inverse document frequency algorithm.

[0030] The initial dataset refers to a data structure containing the core skill label list and the user desired label set array, stored in JSON format, containing two fields: the core skill label list and the user desired label set array, for subsequent data transmission.

[0031] The entire process ensures that the parsing from the resume file to the generation of the initial dataset is executed in sequence, realizing the conversion of the input file and user input into a unified data carrier.

[0032] For example, the user uploads a resume file containing "proficient in Java programming, 5 years of development experience, bachelor's degree in computer science", the text parsing algorithm extracts the skill item as "Java programming", the work experience as "5 years of development experience", and the education background as "bachelor's degree in computer science", and combines them into the core skill label list ["Java programming", "5 years of development experience", "bachelor's degree in computer science"].

[0033] Meanwhile, the user inputs the position description text "Recruit senior Java engineers, need Java and Spring framework experience", and the keyword extraction algorithm extracts the keywords "Java" and "Spring framework", forming the user expectation label set {"Java", "Spring framework"}.

[0034] The initial data set is encapsulated as a JSON object {"core_skills":["Java programming", "5 years of development experience", "Bachelor of Computer Science"], "expected_skills":["Java", "Spring framework"]}. This example directly verifies the extraction logic of core skill labels and user expectation label sets consistent with the set basis.

[0035] The association skill query module queries and calculates the association strength value based on the core skill labels in the initial data set using the industry knowledge graph, generating an association skill set.

[0036] In a preferred embodiment, the step of generating an association skill set includes: for each core skill label, matching the skill nodes directly associated with the core skill label in the industry knowledge graph as the association skills.

[0037] According to the pre-defined rule base, the association strength value of each association skill is calculated based on the skill co-occurrence frequency and semantic similarity.

[0038] Integrate all association skills and their corresponding association strength values to generate an association skill set.

[0039] Wherein, the industry knowledge graph refers to a pre-constructed structured skill association database, with attributes as a graph data structure, nodes representing specific skill strings, and edges representing skill association relationships, and the set basis is based on 10,000 public position historical data training and construction, and the data sources include recruitment platform API interface.

[0040] Association skills refer to other skill items directly associated with core skill labels in the industry knowledge graph, with attributes as strings, representing technical capabilities or knowledge fields.

[0041] Skill co-occurrence frequency refers to the proportion of skills appearing simultaneously in historical data. By analyzing the historical data set, the number of times two skills appear simultaneously in all resumes or position descriptions is counted, and divided by the total number of appearances (or total number of resumes / position descriptions), to obtain the skill co-occurrence frequency.

[0042] Semantic similarity refers to the semantic distance measure of skill names. It is obtained by calculating the cosine similarity or other similarity measures between two skill name vectors, and the result is normalized to the range of 0 to 1. For example, use a pre-trained word embedding model (such as Word2Vec, GloVe, etc.) to obtain the vector representation of the skill name, and then calculate the cosine similarity between these vectors.

[0043] The association strength value represents a quantitative evaluation score of the correlation between skills, ranging from 0 to 1, and is set according to a predefined rule base defined based on a combination of skill co-occurrence frequency and semantic similarity calculation rules, where the co-occurrence frequency weight accounts for 60%, and the semantic similarity weight accounts for 40%.

[0044] The association strength value calculation formula is , where is the association strength value, is the skill co-occurrence frequency, is the semantic similarity, , and are the weights of skill co-occurrence frequency and semantic similarity, respectively, accounting for 0.6 and 0.4, respectively.

[0045] The predefined rule base refers to a database engine that stores calculation rules, with properties such as a rule set containing the weights of skill co-occurrence frequency and semantic similarity, and the setting basis is a standardized rule set defined by industry expert experience.

[0046] The associated skill set refers to a data structure that integrates all associated skills and association strength values, stored using a key-value pair data structure, with the key being the associated skill name and the value being the association strength value score, used for subsequent data transmission.

[0047] The entire process ensures sequential execution from initial data set extraction to associated skill set generation, realizing the transformation of core skill labels to extended association data.

[0048] For example, from the initial data set example, the core skill label list ["Java programming", "5 years of development experience", "Bachelor of Computer Science"] is extracted.

[0049] For the "Java programming" label, query the industry knowledge graph to match associated skills including "Spring Framework", and calculate the association strength value based on the predefined rule base: the co-occurrence frequency is 80%, the semantic similarity is 90%, and the combined calculation gives the strength value 0.85.

[0050] Similarly, for the "5 years of development experience" label, the matching associated skill includes "project management", and the strength value is calculated as 0.75.

[0051] The integrated association skill set is {"Spring Framework": 0.85, "Project Management": 0.75}. This example directly verifies the consistency of the industry knowledge graph matching, association strength value calculation, and association skill set integration logic and setting.

[0052] The dynamic weight calculation module combines real-time job demand data and the association skill set to calculate the demand heat value and dynamic weight value of each skill, generating a dynamic weight label candidate set.

[0053] In a preferred embodiment, the step of generating a dynamic weight label candidate set includes obtaining real-time job demand data from a recruitment platform through a predefined interface, including a plurality of job description text sets.

[0054] The demand heat value is generated by counting the frequency of each skill in the association skill set in the real-time job demand data.

[0055] The dynamic weight value of each skill is calculated by multiplying the association strength value and the demand heat value, representing the quantitative score of the comprehensive importance of the skill. The dynamic weight value ranges from 0 to 1 and has no unit.

[0056] Compare the dynamic weight value of each skill with its preset threshold value to filter skills with a dynamic weight value greater than the preset threshold value to form a dynamic weight label candidate set.

[0057] Real-time job demand data refers to the latest job posting information set obtained through the recruitment platform API interface, with attributes as a text data set, including multiple job description strings, and the setting basis is the real-time data stream interface provided by the recruitment platform, such as LinkedIn or Indeed API.

[0058] The demand heat value represents the frequency ratio of the skill in the job demand, with attributes as a floating-point number, ranging from 0 to 1, and the setting basis is the frequency statistical method, with the formula as the number of skill occurrences divided by the total number of job demands.

[0059] The preset threshold value refers to the screening threshold value of the dynamic weight, with attributes as a predefined floating-point constant, and the setting basis is the 0.5 dividing point determined based on 1000 historical job data analysis, which ensures that the screening result is representative of the industry.

[0060] The dynamic weight label candidate set refers to the filtered skill list, stored in a list data structure, with each element as a skill string.

[0061] The entire process ensures that the dynamic weight label candidate set is generated in sequence from the association skill set, achieving skill weight optimization based on real-time demand.

[0062] Further, periodically analyze the skill trends in real-time job demand data, and automatically update the skill nodes and associated relationships in the industry knowledge graph according to the skill trends.

[0063] Through the continuous updating of the industry knowledge graph by real-time job demand data, the static knowledge base has the ability of dynamic evolution, thereby improving the accuracy and timeliness of the calculation of the correlation strength value, and realizing the synergistic enhancement effect of combining historical knowledge precipitation and instant market demand.

[0064] For example, based on the associated skill set example {"Spring Framework": 0.85, "Project Management": 0.75}, the system obtains real-time job demand data through an API interface: assuming that "Spring Framework" appears 80 times and "Project Management" appears 50 times in 100 job descriptions.

[0065] Calculate the demand heat value: "Spring Framework" is 80 divided by 100, which equals 0.8, and "Project Management" is 50 divided by 100, which equals 0.5.

[0066] Calculate the dynamic weight: "Spring Framework" is 0.85 multiplied by 0.8, which equals 0.68, and "Project Management" is 0.75 multiplied by 0.5, which equals 0.375.

[0067] Set the threshold value to 0.5, and after comparison, filter out the "Spring Framework" skill item to generate the dynamic weight label candidate set ["Spring Framework"]. This example directly verifies that the demand heat value calculation, dynamic weight formula application, and threshold filtering logic and setting basis are consistent.

[0068] The label optimization module adjusts the dynamic weight label candidate set by combining artificial feedback data to generate an optimized label candidate set.

[0069] In a preferred embodiment, the step of generating an optimized label candidate set includes receiving user manual adjustment instructions for the dynamic weight label candidate set. Specifically, user manual adjustment instructions are input through an interactive interface, including three types of operations: adding new skill labels to the candidate set, deleting existing skill labels in the candidate set, or modifying the weight value of skill labels in the candidate set.

[0070] Update the correlation strength value in the industry knowledge graph according to the manual adjustment instructions. Specifically, for the addition instruction, the system adds a new skill node in the industry knowledge graph and sets its correlation strength value based on the user-specified value; for the deletion instruction, the system removes the associated data of the skill in the industry knowledge graph; for the modification instruction, the system updates the correlation strength value of the corresponding skill in the industry knowledge graph to the user input value.

[0071] The system re-accesses the recruitment platform API interface to obtain the latest real-time job requirement data, calculates the demand heat value of each skill in the optimized skill set, i.e. the number of occurrences of the skill divided by the total number of job requirements, and then based on the updated association strength value and the latest skill demand heat value, re-applies the dynamic weight calculation formula, i.e. the dynamic weight equals the association strength value multiplied by the demand heat value, to calculate the dynamic weight value of each skill, compares each dynamic weight value with a preset threshold, and filters out all skill items whose dynamic weight value exceeds the preset threshold, and integrates the filtering results to generate an optimized tag candidate set.

[0072] Among them, the user manual adjustment instruction attribute is a data structure containing an operation type string and a skill information string or a numerical value, and the setting basis is an interactive interface design standard based on HTML form to capture user input.

[0073] The optimized tag candidate set refers to the generated skill tag list after adjustment, which is used for subsequent data transmission.

[0074] The combination of manual adjustment instructions and real-time job requirement data enables single human intervention to dynamically adapt to market changes, producing more accurate optimized tag candidate sets than isolated adjustments or isolated analyses, and achieving the synergistic optimization effect of human experience and real-time data. The entire process ensures that the sequence from user adjustment input to optimized tag candidate set output is executed, realizing skill weight optimization driven by human feedback.

[0075] For example, based on the example dynamic weight tag candidate set ["Spring Framework"], the user inputs an addition instruction to add the skill "Git", and the system updates the industry knowledge graph, adding the "Git" skill node and setting its association strength value to 0.7.

[0076] The real-time job requirement data is re-acquired through the API interface. Assuming that "Git" appears 30 times in 100 job requirements, the demand heat value is calculated as 30 divided by 100, which equals 0.3; "Spring Framework" appears 80 times, and the demand heat value remains 0.8.

[0077] The association strength value is obtained from the updated industry knowledge graph: "Spring Framework" remains 0.85, and "Git" is 0.7. The dynamic weight is recalculated: "Spring Framework" is 0.85 multiplied by 0.8, which equals 0.68, and "Git" is 0.7 multiplied by 0.3, which equals 0.21.

[0078] The preset threshold is 0.5, and the filtered optimized tag candidate set is ["Spring Framework"]. This example directly verifies that the user adjustment instruction processing, industry knowledge graph update, dynamic weight recalculation, and filtering logic and setting basis are consistent.

[0079] In a further preferred embodiment, the updating of the association strength values in the industry knowledge graph according to the manual adjustment instructions comprises the following steps: in response to the professional judgment of a user on a single resume entity or attribute information, such as an AI engineer, an artificial intelligence algorithm engineer, or attribute information such as a certain skill description, receiving artificial feedback data generated based on the manual adjustment instructions, for example, the user thinks that the expression "AI-related worker" written in the resume is not accurate, and manually adjusts it to "AI engineer", and the system will receive the artificial feedback data corresponding to this adjustment.

[0080] According to the artificial feedback data, the updating operation of the industry knowledge graph is performed, specifically including correcting entity attributes, adjusting the association relationship between entities, or updating the association strength values, such as: correcting "AI engineer" to "artificial intelligence algorithm engineer"; deleting the wrong association between "blockchain engineer" and "financial audit", and adding the skill association between "cloud computing architect" and "Kubernetes"; and increasing the association strength value between "Python" and "data analyst" from 0.7 to 0.9, indicating that they are more closely associated.

[0081] The professional judgment information contained in the artificial feedback data for a single resume is converted into general knowledge representation that can be recognized and utilized by the industry knowledge graph, for example, the user's judgment of the association between the skills and positions in a certain resume is converted into rules or data structures that the industry knowledge graph can understand.

[0082] In all subsequent resume processing processes, the optimized industry knowledge graph is called for entity recognition and relationship extraction, and the association strength values between entities are calculated based on the updated knowledge graph state.

[0083] Through the above steps, the continuous conversion of individual user experience into general model knowledge is realized, so that the calculation of the association strength values of all subsequent resumes can benefit from the professional judgment of the previous users, thereby producing a group intelligence effect and a system self-optimization synergistic effect beyond single correction.

[0084] The data storage module generates a unique identifier for the resume to be processed, and stores the optimized tag candidate set and the unique identifier in the tag database in association.

[0085] In a preferred embodiment, the step of storing the optimized tag candidate set and the unique identifier in the tag database in association comprises: creating a unique identifier for the current resume to be processed using a globally unique identifier generation algorithm.

[0086] The unique identifier is used as the key, and the optimized tag candidate set is used as the value, to construct a key-value pair data structure.

[0087] The connection tag database is executed to insert data into the database to store the key-value data structure into the database, and after the storage is completed, a data storage signal is output, which is an acknowledgement message containing a unique identifier, completing the persistent storage.

[0088] The unique identifier represents a unique identification code of the current resume, the attribute is a string type, and the setting basis is the UUID generation algorithm based on the RFC4122 standard, which generates a globally unique identifier by combining a timestamp and a random array.

[0089] The tag database refers to a database system for storing tag data persistently, and the attribute is a relational database engine, and the setting basis is that the system architecture uses a MySQL database instance to support standard SQL insertion operations.

[0090] The data storage signal refers to an acknowledgement response of the completed operation, and the attribute is a JSON format object containing a unique identifier field, and the setting basis is the RESTful API design specification.

[0091] The entire process ensures that the input of the optimized tag candidate set to the output of the data storage signal is executed in sequence, and the permanent association storage of the tag data and the resume identifier is realized.

[0092] For example, based on the example optimized tag candidate set ["Spring Framework"], the system generates a unique identifier "9a8b7c6d-1234-5678-efgh-ijklmnopqr".

[0093] The association data structure {"resume_id":"9a8b7c6d-1234-5678-efgh-ijklmnopqr","tags":["Spring Framework"]} is constructed and stored in the tag database.

[0094] The data storage signal {"status":"success","resume_id":"9a8b7c6d-1234-5678-efgh-ijklmnopqr"} is output. This example directly verifies that the unique identifier generation algorithm, data structure association logic, database storage operation and signal output mechanism are consistent with the setting basis.

[0095] The matching degree calculation module calculates the aggregation weight of the intersection skill according to the user expected tag set and the optimized tag candidate set in the initial data set, and generates a resume matching degree.

[0096] In a preferred embodiment, the step of generating a resume matching degree includes extracting all skills in the user expected tag set and all skills in the optimized tag candidate set, and identifying the intersection skills of the two types of tag sets.

[0097] The intersection skill refers to a skill item that is expected to be in both the user expectation label set and the optimization label candidate set, and the attribute is a string list representing the matched skill label, and the setting basis is set comparison logic based on complete string matching.

[0098] For each intersection skill, the dynamic weight value associated with the optimization label candidate set is obtained.

[0099] The dynamic weight values of all intersection skill items are aggregated, and the total aggregation weight value is calculated by summation operation, and then the total aggregation weight value is output as the resume matching degree, which represents the quantitative indicator of the matching degree between the resume and the user expectation, for subsequent adaptation decision.

[0100] The entire process ensures sequential execution from input label set to resume matching degree score generation, realizing quantitative evaluation conversion of resume matching with user expectation.

[0101] For example, based on the example of the user expectation label set being Java and Spring Framework string elements, and the optimization label candidate set being Spring Framework string elements.

[0102] After extracting the skills, the user expectation label set contains Java and Spring Framework, and the optimization label candidate set contains Spring Framework, and the intersection skill is identified as Spring Framework.

[0103] The dynamic weight value of this intersection skill is 0.68, the aggregation weight is 0.68, and then the resume matching degree is output as 0.68. This example directly verifies that the intersection skill identification, dynamic weight acquisition and summation aggregation logic are consistent with the setting basis.

[0104] The adaptation result generation module generates a job adaptation result containing a unique identifier, a resume matching degree and a pass status based on the comparison of the resume matching degree and the preset qualified threshold.

[0105] In a preferred embodiment, the generation of a job adaptation result containing a unique identifier, a resume matching degree and a pass status includes: comparing the resume matching degree with the preset qualified threshold, and performing logical judgment: if the resume matching degree is greater than or equal to the preset qualified threshold, a job adaptation pass signal is generated.

[0106] If the resume matching degree is less than the preset qualified threshold, a job adaptation fail signal is generated.

[0107] The unique identifier, the resume matching degree and the job adaptation pass signal or the job adaptation fail signal are integrated into the job adaptation result, and then the job adaptation result is output to the user interface or the external system.

[0108] Among them, the preset qualification threshold refers to the score threshold value for judging whether a resume is qualified. The attribute is a floating-point constant. The setting basis is the industry standard value of 0.6 determined based on the analysis of 2,000 historical job matching data. This value ensures that the screening has practical application significance.

[0109] The position adaptation pass signal indicates confirmation that the resume meets the qualification requirements. Its attribute is a string constant such as "pass", and its function is to trigger subsequent operations such as notification sending.

[0110] The position adaptation failure signal is a negative indication that the resume does not meet the requirements. The attribute is a string constant such as "failed".

[0111] The entire process ensures sequential execution from resume matching score input to adaptation result output, realizing the automated conversion of job adaptation decisions.

[0112] For example, based on the sample resume matching degree of 0.68, the system obtains a preset qualification threshold of 0.6. Comparison operation: if 0.68 is greater than or equal to 0.6, a position matching pass signal "pass" is generated.

[0113] The unique identifier references "9a8b7c6d-1234-5678-efgh-ijklmnopqr" for example.

[0114] Construct the adaptation result as a JSON object {"resume_id":"9a8b7c6d-1234-5678-efgh-ijklmnopqr","match_score":0.68,"status":"Passed"} and output the result to the user interface. This example directly verifies that the threshold comparison logic, signal generation rules, and result data structure integration are consistent with the set basis.

[0115] In a further preferred embodiment, after the job adaptation result is generated, the following steps are further included: comparing the user expected tag set with the optimized tag candidate set, and identifying missing skills in the user expected tag set that do not appear in the optimized tag candidate set.

[0116] Based on the missing skills, the related skills are queried in the industry knowledge graph as recommended learning paths.

[0117] Generate a skills gap analysis report that includes missing skills and recommended learning paths.

[0118] Among them, by combining job adaptation results with industry knowledge graphs to generate skill gap analysis reports, the matching assessment function is expanded to a career development guidance function, achieving the synergistic effect of data reuse and function upgrading.

[0119] The above merely illustrates and describes the concept of the present application, and those skilled in the art can make various modifications or supplements to the described specific embodiments or adopt similar ways to replace, as long as the modifications or supplements do not deviate from the concept of the present application or exceed the defined scope of the present application, and should belong to the protection scope of the present application.

Claims

1. An intelligent tag generation and retrieval system based on AI resume, characterized by: include: The initial dataset generation module obtains the resume to be processed and the user's search conditions, and generates an initial dataset containing core skill labels and user expected label sets; The associated skills query module uses the industry knowledge graph to query and calculate the association strength value based on the core skill tags in the initial data set to generate an associated skill set; The dynamic weight calculation module combines real-time job demand data with the associated skill set to calculate the demand heat value and dynamic weight value of each skill and generate a dynamic weight label candidate set; The label optimization module integrates manual feedback data to adjust the dynamic weight label candidate set and generate an optimized label candidate set; A data storage module generates a unique identifier for the resume to be processed, and associates the optimized label candidate set with the unique identifier and stores it in the label database; The matching degree calculation module calculates the aggregation weight of the intersection skills based on the user's expected label set and the optimized label candidate set in the initial data set, and generates the resume matching degree; The adaptation result generation module generates a position adaptation result including a unique identifier, resume matching degree and pass status based on the comparison of resume matching degree with the preset qualification threshold.

2. The intelligent tag generation and retrieval system based on AI resume according to claim 1 is characterized in that: The steps of generating an initial data set including core skill labels and user expected label sets include: Receive user-uploaded resume files, parse the resume content using natural language processing technology, identify skill items, work experience paragraphs, and educational background descriptions in the text, extract them into independent text elements, and form core skill tags; Receive the job description text input by the user, and use the keyword extraction algorithm to identify key terms as user expected skill tags to form a user expected tag set; The core skill labels and user expected label sets are encapsulated as the initial dataset.

3. The intelligent tag generation and retrieval system based on AI resume according to claim 1 is characterized in that: The step of generating the associated skill set includes: For each core skill tag, match the skill node directly associated with the core skill tag in the industry knowledge graph as the associated skill; According to the predefined rule base, the association strength value of each related skill is calculated based on the skill co-occurrence frequency and semantic similarity; Integrate all associated skills and their corresponding association strength values ​​to generate an associated skill set.

4. The intelligent tag generation and retrieval system based on AI resume according to claim 1 is characterized in that: The step of generating a dynamic weight label candidate set includes: Obtain real-time job demand data from the recruitment platform through a predefined interface, including multiple job description text sets; Count the occurrence frequency of each skill in the associated skill set in real-time job demand data to generate a demand heat value; Multiply the association strength value by the demand heat value to calculate the dynamic weight value of each skill; Compare the dynamic weight value of each skill with its preset threshold, filter out skill items with dynamic weight values ​​greater than the preset threshold, and form a dynamic weight label candidate set.

5. The intelligent tag generation and retrieval system based on AI resume according to claim 1 is characterized in that: The step of generating an optimized label candidate set includes: Receive manual adjustment instructions from users on the dynamic weight label candidate set; Update the association strength value in the industry knowledge graph according to manual adjustment instructions; Based on the updated association strength value and the latest skill demand heat value, the dynamic weight calculation formula is reapplied to calculate the dynamic weight value of each skill. Each dynamic weight value is compared with the preset threshold, and all skill items whose dynamic weight value exceeds the preset threshold are screened out to generate an optimized label candidate set.

6. The intelligent tag generation and retrieval system based on AI resume according to claim 5 is characterized in that: The updating of the association strength value in the industry knowledge graph according to the manual adjustment instruction includes the following steps: In response to a user's professional judgment of a single resume entity or attribute information, receiving manual feedback data generated based on manual adjustment instructions; Based on manual feedback data, perform updates to the industry knowledge graph, including modifying entity attributes, adjusting relationships between entities, or updating association strength values; Convert the professional judgment information contained in the manual feedback data for a single resume into a generalized knowledge representation that can be recognized and utilized by the industry knowledge graph; In all subsequent resume processing processes, the optimized industry knowledge graph is called for entity recognition and relationship extraction, and the association strength values ​​between entities are calculated based on the updated knowledge graph status.

7. The intelligent tag generation and retrieval system based on AI resume according to claim 1 is characterized in that: The step of associating the optimized tag candidate set with the unique identifier and storing it in the tag database includes: Use a global unique identifier generation algorithm to create a unique identifier for the resume to be processed; Use the unique identifier as the key and the optimized label candidate set as the value to construct a key-value pair data structure; Store the key-value pair data structure in the database to achieve persistent storage.

8. The intelligent tag generation and retrieval system based on AI resume according to claim 1 is characterized in that: The steps of generating resume matching degree include: Extract all skills in the user's desired tag set and all skills in the optimized tag candidate set, and identify the intersection skills of the two tag sets; For each intersection skill, obtain its dynamic weight value associated with the optimized label candidate set; Aggregate the dynamic weight values ​​of all intersection skill items, calculate the total aggregate weight value through summation, and then output the total aggregate weight value as the resume matching degree.

9. The intelligent tag generation and retrieval system based on AI resume according to claim 1 is characterized in that: The steps of generating a job matching result including a unique identifier, a resume matching degree, and a pass status include: Compare the resume matching degree with the preset qualification threshold and perform logical judgment: If the resume matching degree is greater than or equal to the preset qualification threshold, a job adaptation pass signal is generated; If the resume matching degree is less than the preset qualification threshold, a job adaptation failure signal is generated; The unique identifier, resume matching degree and position adaptation pass signal or position adaptation fail signal are integrated into the position adaptation result, and then the position adaptation result is output to the user interface or external system.

10. The intelligent tag generation and retrieval system based on AI resume according to claim 9 is characterized in that: After the job matching result is generated, the following steps are also included: Compare the user's expected label set with the optimized label candidate set to identify missing skills in the user's expected label set that do not appear in the optimized label candidate set; Based on the missing skills, query the industry knowledge graph for related skills as recommended learning paths; Generate a skills gap analysis report that includes missing skills and recommended learning paths.

Images