大语言模型驱动的跨学科知识图谱构建与共性知识发现研究OACHSSCD

[Purpose/Significance]Interdisciplinary integration is vital for addressing complex global challenges,but terminology differences and paradigm fragmentation obscure implicit common knowledge.Existing research relies on explicit connections and fails to identify unassociated commonalities.The paper aims to construct a LLM-driven interdisci-plinary knowledge graph to discover common knowledge and promote cross-field knowledge migration.[Method/Process]The study selected computer science and economics,using the CORE dataset.It applied SciAIEngine to extract core enti-ties,built a"problem-method"semantic system,classified relationships with large language model,and used the HowSim algorithm on a Neo4j-based graph to mine commonalities.[Result/Conclusion]The constructed knowledge graph contains 24 542 nodes and 61 697 edges,from which 19 strong common knowledge pairs were identified.Case verification showed that some models of computer science and economics have cross-disciplinary commonalities.The integrated framework of large language model,knowledge graph with refined semantics and HowSim algorithm effectively realizes implicit commo-nality identification,providing a technical path for interdisciplinary knowledge discovery.