CN-121981691-A - Intelligent resume screening method, system, equipment and medium based on large language model

CN121981691ACN 121981691 ACN121981691 ACN 121981691ACN-121981691-A

Abstract

The invention discloses an intelligent resume screening method, system, equipment and medium based on a large language model, wherein the method specifically comprises the steps of analyzing job descriptions through the large language model based on initial weights, and extracting key skill requirements; analyzing the matching degree of each resume, generating a matching degree score through key skill requirements, determining a resume screening result according to the matching degree score, collecting adjustment operation and marking reasons of HR on the resume screening result, dynamically optimizing weight distribution through a reinforcement learning model according to the adjustment operation and the marking reasons, automatically adjusting skill priority according to a post requirement prediction model by utilizing the optimized weight, identifying the depth matching relation between candidate skill and post requirements in the resume screening result based on the skill priority, and evaluating potential indexes of the candidates for adapting to new posts according to the depth matching relation and industry talent flow data. The invention realizes the intelligent resume screening with high efficiency, accuracy and dynamic adaptation, and improves recruitment quality and efficiency.

Inventors

CHEN SHUWEI
LI TIANPENG

Assignees

安徽三七极域网络科技有限公司

Dates

Publication Date: 20260505
Application Date: 20251211

Claims (10)

1. The intelligent resume screening method based on the large language model is characterized by comprising the following steps of: acquiring job descriptions drawn by a demand department, distributing initial weights for all requirements, analyzing the job descriptions through a large language model based on the initial weights, and extracting key skill requirements; Scanning a resume library, analyzing the matching degree of each resume, and generating a matching degree score according to the key skill requirements; Determining a resume screening result according to the matching degree score, and collecting adjustment operation and labeling reasons of the HR on the resume screening result through an HR feedback interface; According to the adjustment operation and the labeling reasons, combining the interview passing rate and the post-recording performance data, and dynamically optimizing weight distribution through a reinforcement learning model; automatically adjusting skill priority by utilizing the optimized weight according to the post demand prediction model and combining industry trend data; based on the skill priority, identifying a depth matching relation between the skill of the candidate in the resume screening result and the post requirement through a skill path analysis model, and evaluating the potential index of the candidate adapting to the new post according to the depth matching relation and the industry talent flow data.
2. The method according to claim 1, wherein the obtaining job descriptions formulated by the demand department and assigning initial weights to the requirements, analyzing the job descriptions by a large language model based on the initial weights, extracting key skill requirements, specifically comprises: Acquiring a job description document and a weight configuration table provided by a demand department, wherein the weight configuration table comprises initial weight values required by various skills; based on the weight configuration table, inputting the job description document into a pre-trained large language model, and performing word segmentation and semantic analysis on the job description through a semantic understanding layer of the large language model to obtain a semantic analysis result; based on the semantic analysis results, key skill requirements are extracted, and the key skill requirements are mapped to a standardized skill library through a skill classifier.
3. The method according to claim 1, wherein the scanning the library of resumes and analyzing the matching degree of each resume, generating a matching degree score by key skill requirements, specifically comprises: scanning a resume library to obtain resume documents; The technical entity, the work experience entity and the education background entity in the resume document are extracted through the entity identification technology, the association relation and the time sequence among the entities are determined, and the resume analysis result is obtained; and comparing the resume analysis result with the key skill requirement to obtain a matching degree score, wherein the multi-dimensional index comprises skill coverage, experience matching degree, skill depth and weight coincidence degree.
4. The method of claim 1, wherein determining the resume screening result according to the matching degree score, and collecting adjustment operation and labeling reasons of the resume screening result by the HR through the HR feedback interface, specifically comprises: Determining a multilevel threshold based on the historical screening data and the job importance level, and determining a resume screening result level by comparing the matching degree score with the multilevel threshold; Presenting screening result grades to the HR through a result display interface, and collecting the adjustment operation of the HR on the screening result through an HR feedback interface; Based on the adjustment operation, the HR selection is prompted or the adjustment reason is input through the annotation guiding module, and the input adjustment reason is normalized and classified through the reason classifier.
5. The method according to claim 1, wherein the dynamically optimizing the weight distribution by the reinforcement learning model based on the adjustment operation and the labeling reason and combining the interview passing rate and the post-recording performance data comprises: according to the adjustment operation and the labeling reasons, combining the interview passing rate and the post-recording expression data to construct a state space of the reinforcement learning model; Determining available actions of the reinforcement learning model based on a state space, and constructing an action space, wherein the action space comprises the steps of increasing the weight of a specific skill according to a preset step length, reducing the weight of the specific skill according to the preset step length, reallocating weight proportions among related skills and maintaining the current weight configuration unchanged; Taking the reduction degree based on the HR adjustment frequency as short-term rewards, taking the lifting amplitude based on the interview passing rate as medium-term rewards, taking the improvement condition based on the performance of the staff for recording as long-term rewards, taking the smoothness degree based on the weight adjustment amplitude as stability rewards, and constructing a rewarding function; Based on the state space, the action space and the reward function, training the reinforcement learning model through the strategy network, and converting the optimal action output by the reinforcement learning model into actual weight adjustment through the weight updating engine.
6. The method of claim 1, wherein the automatically adjusting skill priorities in combination with industry trend data based on the post demand prediction model using the optimized weights, specifically comprises: according to the historical recruitment data, the post description data and the industry report data, a post demand prediction model is constructed by adopting a time sequence prediction algorithm; Analyzing skill demand trend of the current post through a post demand prediction model based on the optimized weight; The method comprises the steps of collecting industry trend data, integrating optimized data, skill demand trend data and industry trend data through a weighted average algorithm and correlation analysis, and generating a comprehensive demand index; And calculating skill priority according to priority adjustment indexes based on the comprehensive demand index, wherein the priority adjustment indexes comprise the importance of the current weight, the urgency of post demand prediction, the influence degree of industry trend data and the relevance among skills.
7. The method according to claim 1, wherein the step of identifying the depth matching relation between the candidate skill and the post requirement in the resume screening result by the skill path analysis model based on the skill priority, and evaluating the potential index of the candidate for adapting to the new post according to the depth matching relation and the industry talent flow data specifically comprises the steps of: Adopting a graph neural network algorithm, taking skills as nodes, taking the relevance among the skills and a transfer path as edges, and constructing a skill path analysis model; based on the skill priority and the resume screening result, identifying the depth matching relation between the skill of the candidate and the post requirement through a skill path analysis model; acquiring industry talent flow data, and calculating potential indexes of candidates adapting to new posts according to multi-factor indexes based on the depth matching relation and the industry talent flow data; wherein the multi-factor index comprises skill adaptation degree, industry flow trend, learning and development potential and environment adaptation capability, and the potential index comprises potential score, skill gap analysis, adaptation suggestion and risk prompt.
8. An intelligent resume screening system based on a large language model, which is characterized by comprising the following specific steps: The first screening module is used for acquiring job descriptions drawn by a demand department, distributing initial weights for all requirements, analyzing the job descriptions through a large language model based on the initial weights, and extracting key skill requirements; the second screening module is used for scanning the resume library, analyzing the matching degree of each resume and generating a matching degree score according to the key skill requirements; The third screening module is used for determining resume screening results according to the matching degree scores, and collecting adjustment operation and labeling reasons of the resume screening results by the HR through the HR feedback interface; the fourth screening module is used for dynamically optimizing weight distribution through a reinforcement learning model according to adjustment operation and labeling reasons and combining interview passing rate and post-recording performance data; The fifth screening module is used for automatically adjusting skill priority according to the post demand prediction model and the industry trend data by utilizing the optimized weight; And the sixth screening module is used for identifying the depth matching relation between the candidate skills and the post demands in the resume screening results through the skill path analysis model based on the skill priority, and evaluating the potential indexes of the candidates adapting to the new post according to the depth matching relation and the industry talent flow data.
9. A computer device comprising a memory and a processor and a computer program stored on the memory, which when executed on the processor implements the intelligent resume screening method based on a large language model of any one of claims 1 to 7.
10. A computer readable storage medium, having stored thereon a computer program which, when executed by a processor, implements the intelligent resume screening method based on a large language model of any of claims 1 to 7.

Description

Intelligent resume screening method, system, equipment and medium based on large language model Technical Field The invention relates to the technical field of artificial intelligence, in particular to an intelligent resume screening method, system, equipment and medium based on a large language model. Background In the competitive employment market today, the high efficiency and accuracy of the recruitment process is critical to the attraction and selection of appropriate talents for the enterprise. The recruitment process typically begins with a demand department drawing up job descriptions based on job demands, and then recruiting Human Resources (HR) departments takes over the role of screening out qualified candidates from a huge resume. However, in the actual operation process, the link faces many challenges, which seriously affects the recruitment efficiency and quality. First, the HR department often faces urgent time constraints that require a large number of resumes to be processed in a specified time. Taking a large business as an example, a recruitment activity may receive thousands of resumes. In such a limited period of time, it is difficult for HR personnel to perform a detailed and comprehensive analysis of each resume, often only with quick browsing and initial screening. The rapid screening mode is very easy to cause the omission of excellent candidates, so that the screening efficiency is greatly reduced, and the requirement of enterprises on talent timely recruitment cannot be met. Second, different HR personnel may have differences in understanding and performing the screening criteria. Because of the lack of unified, explicit screening criteria, each HR may determine whether the candidate is satisfactory based on its own experience and understanding. The subjective difference causes the lack of reliability of screening results, the same resume can be evaluated in different HR hands, fairness and consistency of the whole recruitment process are further affected, and uncertainty is brought to talent selection of enterprises. Moreover, HR personnel have certain human resource management knowledge and skills, but the professional skills for each post are often limited. The requirements of the expertise of the candidates for different posts are greatly different, for example, the software development post needs to master a specific programming language and technical framework, and the mechanical design post focuses on engineering drawing and mechanical principle knowledge. Under the condition that the HR personnel lack of in-depth professional knowledge, the matching degree between skills possessed by the candidates and post demands is difficult to accurately judge, and the screened candidates are easy to be caused to be inconsistent with actual working requirements. Finally, the conventional resume screening method mainly relies on keyword matching, namely, the coincidence degree of the candidate is judged by searching the resume for the content matched with the keywords in the job description. However, this approach has significant limitations. On the one hand, it cannot accurately recognize semantic information. For example, for some words that have similar meanings but different expressions, the keyword matching method may not correlate them, resulting in missing satisfactory candidates. On the other hand, for the judgment of professional skills, it is far from sufficient to rely on keyword matching alone. Professional skills often involve complex knowledge systems and practical handling capabilities, and simple keyword occurrences do not fully reflect the true level of candidates. This limitation severely affects the accuracy of the screening, making it difficult for the enterprise to find talents that are truly suitable for the post. Disclosure of Invention The invention aims to provide an intelligent resume screening method, system, equipment and medium based on a large language model, which realize efficient, accurate and dynamically adaptive intelligent resume screening, improve recruitment quality and efficiency and solve at least one of the problems in the prior art. In a first aspect, the present invention provides an intelligent resume screening method based on a large language model, where the method specifically includes: acquiring job descriptions drawn by a demand department, distributing initial weights for all requirements, analyzing the job descriptions through a large language model based on the initial weights, and extracting key skill requirements; Scanning a resume library, analyzing the matching degree of each resume, and generating a matching degree score according to the key skill requirements; Determining a resume screening result according to the matching degree score, and collecting adjustment operation and labeling reasons of the HR on the resume screening result through an HR feedback interface; According to the adjustment operation and the labeling reasons,