Research on Knowledge Graph Construction Technology Based on Intellectual Property Legal Documents

Abstract

Amidst the dual thrust of artificial intelligence development and the construction of legal informatization, knowledge graphs are increasingly applied in areas such as judicial judgment assistance, legal retrieval, and question answering. Addressing the heterogeneity of intellectual property rules and the complexity of case semantics, this paper aims to propose a construction method for a legal knowledge graph in the field of intellectual property. This method establishes an ontological framework covering 84 concepts, 89 relationships, and 105 attributes. It then employs the CasRel and BiLSTM-CRF models to extract entities and relationships from 3,713 intellectual property judicial cases, integrating these with the conceptual-level ontology using the TransE knowledge alignment model. Finally, a visualization platform is built on the Neo4j database to display and manage the knowledge graph, and a knowledge reasoning model is developed using the TransD algorithm to facilitate intelligent question answering. The knowledge graph constructed in this study provides a critical knowledge resource for legal research, case analysis, and decision support.