Construction of Knowledge Graph of Judicial Case Based on LLMs and Embedding Models

This paper constructs the Judicial Case Knowledge Graph (JCKG) dataset based on China Judicial Judgements Online, which fills the gap of knowledge graph research in the legal field. First, we systematically collate the relevant data of Judicial Judgements Online, and then use advanced Large-scale Language Models (LLMs) technology to extract the entities and relationships in the data. Then, through the link prediction experiment, the mainstream knowledge graph learning data set was compared horizontally, and the embedding models such as TransE, ConvE, and pRotatE were compared vertically to verify the validity of JCKG data set. Experiments show that JCKG data set can effectively implement multi-hop inference. Legal Knowledge Graph plays an important role in the intelligent judicial system, which improves the efficiency of case processing and the accuracy of judicial decision-making and reduces the risk of errors. This research provides strong support for the field of legal artificial intelligence and promotes the development of the judicial system in the direction of more efficient and intelligent.