Search

CN-121981592-A - College student capability assessment and matching method, device and product based on AI large model

CN121981592ACN 121981592 ACN121981592 ACN 121981592ACN-121981592-A

Abstract

The application discloses a college student capability assessment and matching method, equipment and a product based on an AI large model. The method comprises the steps of collecting job hunting data of a target college student, wherein the job hunting data comprise resume texts, skill certificates and interaction records, carrying out multi-dimensional capability assessment on the target college student by utilizing an AI model based on the job hunting data, generating an assessment result, wherein the assessment result comprises capability level labels and assessment scores, extracting information units used for supporting the assessment result from the job hunting data, calculating initial confidence degrees based on the information units, dynamically updating the confidence degrees of the assessment result, and dynamically updating the current confidence degrees by adopting an index sliding average algorithm when new evidence is received, wherein the information units extracted from the job hunting data are classified into one or more evidence types of certificate evidence, project evidence, text evidence, interaction evidence and behavior evidence. The embodiment of the application can improve the resume matching precision of college students.

Inventors

  • Yang Diexuan
  • WANG LIN
  • ZHENG DONGSHENG

Assignees

  • 前锦网络信息技术(上海)有限公司

Dates

Publication Date
20260505
Application Date
20251222

Claims (14)

  1. 1. The university student ability assessment and matching method based on the AI large model is characterized by comprising the following steps: collecting job hunting data of a target college student, wherein the job hunting data comprises resume texts, skill certificates and interaction records; Based on the job hunting data, performing multidimensional capability assessment on the target college students by using an AI model, and generating an assessment result, wherein the assessment result comprises a capability grade label and an assessment score; Extracting information units used for supporting the evaluation result from the job hunting data, calculating initial confidence coefficient based on the information units, and dynamically updating the confidence coefficient of the evaluation result, and dynamically updating the current confidence coefficient by adopting an index sliding average algorithm when new evidence is received, wherein the information units extracted from the job hunting data are classified into one or more evidence types of certificate evidence, project evidence, text evidence, interaction evidence and behavior evidence; generating a professional ability vector, a learning ability vector and a psychological diathesis vector of the target college student based on the evaluation result and the confidence coefficient, wherein each ability vector comprises a grade label, an evaluation score and the confidence coefficient which represent the grade of the evaluation result; Analyzing the recruitment requirement of an enterprise to generate an enterprise preference vector, wherein the enterprise preference vector comprises a professional ability grade label, a learning ability grade label, a psychological diathesis grade label and weights corresponding to the required ability of the enterprise; And carrying out matching calculation on each capability vector of the target college student and the enterprise preference vector, and outputting a matching degree result.
  2. 2. The method of claim 1, wherein the multi-dimensional competence assessment comprises at least assessing professional competence based on professional courses, project experiences, and skill certificates, assessing learning competence based on academic context and skill acquisition records, assessing mental quality based on textual descriptions and interaction records; wherein, for the professional ability assessment process includes: constructing a professional knowledge graph, and dynamically updating core courses, skill requirements, item types and certificate authentication information of each professional; calculating quantitative scores of knowledge depth, skill breadth and practice thickness based on course relevance, certificate level, project depth and roles; Obtaining a final score of the professional ability based on the quantized score, and obtaining a professional ability grade by comparing the final score with a threshold; Wherein the calculating of the project depth includes: abstracting the items into multi-layer nodes based on the item participation degree; Calculating the shortest paths of the project nodes and the skill nodes in the knowledge graph, and increasing the project depth if the shortest paths are not larger than a preset value; If the preset keywords appear in the automatic analysis resume, the project depth is increased.
  3. 3. The method of claim 1, wherein outputting the enterprise preference vector based on the enterprise recruitment demand comprises: extracting skill entities, degree adverbs and soft quality requirements from text, voice or historical position descriptions through an entity extraction model; judging emotion polarities of the enterprise on various capacity requirements, and mapping the emotion polarities into different demand weights; the enterprise preference vector is generated based on the extracted entities and the corresponding weights.
  4. 4. The method as recited in claim 1, further comprising: Dividing groups according to gender, universities and territories, and calculating the difference of matching opportunities; when the difference exceeds a threshold value, an alarm is triggered, and the weight in the matching algorithm is automatically adjusted to ensure the fairness of the group.
  5. 5. The method of claim 1, wherein when a conflict occurs between different evidence sources of the same capability dimension or the initial confidence level of the tag is below a preset threshold, triggering an interaction verification mechanism, collecting supplemental information by the dialogue intelligent robot, and updating the evaluation result and the confidence level based on the supplemental information.
  6. 6. The method of claim 1, wherein the matching degree is calculated using a multi-factor matching algorithm based on the capability label and the enterprise preference vector, and the matching score MatchScore is calculated as: Where Wi is the weight of the ith dimension, TAGMATCHI is the label level matching degree of the candidate and the enterprise in the ith dimension, prefCi is the preference coefficient of the enterprise in the ith dimension, and aging attenuation i is the time attenuation factor related to the candidate.
  7. 7. The method of claim 1, wherein the initial confidence level C 0 of the evaluation result is formulated as: , wherein, A data characteristic value representing evidence of the i-th type, Representing the weight of the evidence of class i.
  8. 8. The method of claim 1, wherein the current confidence level The calculation formula of (2) is as follows: The method comprises the steps of (1) setting beta as an attenuation factor, evidenceScore = evidence quality x time attenuation x source weight, wherein the evidence quality is determined according to the detail degree of evidence and whether verifiable specific data or links are contained or not, the time attenuation follows an exponential attenuation model, the attenuation coefficient is set according to the capability category of the evidence, the source weight is set according to authority classification of an evidence source, and the source comprises an official verification interface, an enterprise system, an autonomous uploading file and a third-party webpage.
  9. 9. The method as recited in claim 1, further comprising: if the college students finish relearning within the half-life period of knowledge, the method is as follows Giving learning ability compensation, wherein T is the time from last time of knowledge mastering, T is the half-life of knowledge, and the shorter the half-life of knowledge is, the faster the update rate of knowledge is; And establishing a certificate devaluation curve, and regulating the certification of the certificate by 10% when the annual passing number increase rate is more than 50%.
  10. 10. The method as recited in claim 1, further comprising: for candidates of the composite professional background, calculating a composite score by adopting a knowledge graph sub-graph fusion algorithm: composite score = max (major score, 0.7 x minor major score); The composite score is taken as the professional ability score.
  11. 11. The method as recited in claim 1, further comprising: Each feature original record uses a blockchain side chain to store a hash value and stores a resource locator in MySQL; Clicking on the tab pops up an evidence timeline supporting viewing, screening, and exporting PDF reports.
  12. 12. The method as recited in claim 1, further comprising: when positive and negative evidence pairs appear in the same dimension capability evaluation, calculating a Bayesian factor K; if K <1/3, the confidence is reduced by 0.2 and marked as 'evidence conflict'; and if no new evidence is added within 30 days after the conflict, the corresponding label is automatically frozen and manually rechecked.
  13. 13. An electronic device, characterized in that the electronic device is a terminal device or a server, the electronic device comprising a processor and a memory storing computer program instructions, the electronic device implementing the method according to any of claims 1-12 when executing the computer program instructions.
  14. 14. A computer program product comprising computer program instructions which, when executed, implement the method of any one of claims 1-12.

Description

College student capability assessment and matching method, device and product based on AI large model Technical Field The application relates to the technical field of Internet, artificial intelligence and human resource technology, in particular to a college student capability assessment and matching method, equipment and program product based on an AI large model. Background In the field of intersecting artificial intelligence and human resource technology, existing college recruitment and talent assessment systems typically work based on preset keyword matching, static resume parsing, or simple rule engines. These systems generally have the following limitations: Firstly, the prior art mainly relies on limited information such as resume text, certificate list and the like submitted by job seekers at one time for evaluation, and lacks multi-dimensional deep and structural analysis on candidates. Even if the labeling capability description is introduced into part of the system, the label generation is mostly based on surface keyword matching, and the evaluation result is solidified, so that the authenticity and dynamic change of the candidate capability can not be reflected. Second, the authenticity, degree of detail, and exaggeration of the content of the input information by conventional systems are difficult to effectively discriminate, so that the capability labels and matching suggestions output by the systems may be based on unreliable or outdated information. For example, a fictional project experience or an unverified certificate may lead to a continuous overestimation of the candidate's ability, resulting in false matching recommendations, whereas the subsequent actual results of the candidate are difficult to incorporate into the assessment system quickly and efficiently to correct previous unilateral decisions, resulting in limited final post matching accuracy. Disclosure of Invention In view of the above, the embodiments of the present application provide a college student ability assessment and matching method, an electronic device, and a computer program product based on an AI big model, which solve at least one technical problem. The embodiment of the application provides an university student capability assessment and matching method based on an AI (advanced technology) university model, which comprises the steps of collecting job data of a target university student, wherein the job data comprises resume text, skill certificates and interaction records, carrying out multidimensional capability assessment on the target university student by using the AI model based on the job data to generate an assessment result, wherein the assessment result comprises capability level labels and assessment scores, extracting information units used for supporting the assessment result from the job data, calculating initial confidence degrees based on the information units, dynamically updating the confidence degrees of the assessment result, dynamically updating the current confidence degrees by adopting an index sliding average algorithm when new evidence is received, classifying the information units extracted from the job data into one or more of certificate type evidences, item type evidences, text type evidences, interaction type evidences and behavior type evidences, generating a professional capability vector, a learning capability vector and a learning capability vector of the target university student based on the assessment result and the confidence degrees, and a psychological capability vector of the target university student, calculating the required enterprise capability of the enterprise level and the required enterprise capability level and the required enterprise level, and the required enterprise level and the required enterprise level. Optionally, the multi-dimensional capability assessment method at least comprises the steps of assessing the professional capability based on professional courses, project experiences and skill certificates, assessing the learning capability based on academic backgrounds and skill acquisition records, assessing psychological quality based on text descriptions and interaction records, and assessing the professional capability assessment process comprises the steps of constructing professional knowledge maps, dynamically updating core courses, skill requirements, project types and certificate authentication information of each professional, calculating quantized scores of knowledge depths, skill scales and practice thicknesses based on course correlations, certificate grades, project depths and roles, obtaining final scores of the professional capability based on the quantized scores, and obtaining the professional capability grades by comparing the final scores with thresholds, wherein the project depth calculation comprises the steps of abstracting the project into multi-layer nodes based on project participation degrees, calculating shortest paths of proje