Post description automatic generation method based on deep learning

By constructing module aggregation degree and boundary jump degree, the prediction loss function of the ProphetNet model is improved, which solves the problems of semantic confusion and boundary ambiguity between modules in job description generation, and realizes high-quality and structurally complete automated job description generation.

CN122133724AActive Publication Date: 2026-06-02XIAN TALENT GROUP RESEARCH INSTITUTE CO LTD +1
6 Cites 0 Cited by

Patent Information

Application Number
CN202610568824.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-04-28
Publication Date
2026-06-02
Estimated Expiration
2046-04-28

AI Technical Summary

Technical Problem

The existing ProphetNet model cannot effectively perceive the semantic boundaries and functional importance differences between modules when generating job descriptions, resulting in cross-module semantic confusion, blurred boundaries, and low quality of core and difficult vocabulary generation in the generated text.

Method used

By constructing module aggregation degree and boundary jump degree, structural weights are extracted to improve the prediction loss function of the ProphetNet model, enabling it to adaptively weight and perceive the semantic boundaries and functional importance differences of different modules.

Benefits of technology

It improves the structural integrity and practical usability of job descriptions, and generates clear and standardized text, reducing the workload of manual writing.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to the field of post description, in particular to a deep learning-based post description automatic generation method, which comprises the following steps of: preprocessing acquired post description text data to obtain a plurality of modules, and adding corresponding module labels to segmented words in each module; according to the probability distribution of vocabularies in each module, constructing a module aggregation degree reflecting semantic purity in the module; a boundary local word sequence between adjacent modules is extracted, information entropy is calculated, and a boundary jump degree is constructed in combination with a module polymerization degree; and constructing a structural weight of each module according to the boundary jump degree and the word number proportion of each module, and improving a prediction loss function of the ProphetNet model by using the structural weight as an adaptive weighting coefficient to obtain an improved ProphetNet model for generating a post description text. According to the invention, the writing efficiency and quality in the recruitment scene are improved.
Need to check novelty before this filing date? Find Prior Art