Job-Person Knowledge Extraction Method Based on Large Language Models

Abstract

In the data extraction process for job-person recommendation systems, the data structures present in various forms including structured, semi-structured, and unstructured formats, which leads to significant difficulty and low efficiency in manual knowledge graph construction. Although deep learning techniques have advanced the automation of knowledge graph construction in this domain, they still require substantial amounts of high-quality manually annotated supervised data, thus limiting their practical applicability. To address these issues, this paper proposes a named entity and relation extraction method based on large language models combined with prompt engineering, operating with limited high-quality manually annotated data. We design multi-role prompts and a multi-round chain-based iterative extraction pipeline, complemented by a multi-dimensional knowledge alignment strategy incorporating regex filtering, Wiki-data entity linking, and FuzzyWuzzy spell correction, effectively improving knowledge graph construction efficiency. Finally, comparative and ablation experiments conducted on a public English resume dataset validate the effectiveness and practicality of our method in job-person knowledge extraction and knowledge graph construction.