The Method And Research Of Knowledge Graph Construction For "smart Court"

At the end of July 2016,the General Office of the CPC Central Committee and the General Office of the State Council issued the Outline of the National Informatization Development Strategy,which included the construction of the "Wisdom Court" into the national informationization development strategy.As the judicial organ of our country,the construction of the "Wisdom Court" is of great significance to improving the informationization level of the acceptance,trial,execution and supervision of cases,promoting the disclosure of judicial information and promoting judicial fairness.At present,the case data in the court system is very rich and has a lot of valuable information.However,the current data lacks effective organization,and the key knowledge is difficult to extract,this reasons making it difficult to analyze and utilize.In recent years,the development of knowledge graph technology has provided a feasible solution for the retrieval and analysis of large-scale structured knowledge.The extraction of important structured knowledge from professional and large-scale case data for the construction of knowledge graph can not only solve the search display and storage of large-scale case information in the "Wisdom Court" informatization construction,and it is possible to link important knowledge such as courts,judges,plaintiffs or defendants with case-centeredness.The construction of knowledge graph can effectively organize,analyze and mine the vast amount of valuable information in the court,provide a data foundation for efficient case trial and intelligent division problem solving,and make the "intelligent court" informationization construction better service.society.This paper mainly uses the court judgment document as the data source to construct the knowledge graph for the "Wisdom Court".Firstly,according to the current informationization construction requirements of "Wisdom Court",the knowledge graph structure is constructed,and then the required entities and attribute knowledge are extracted from the data sources such as the judgment documents.After the knowledge fusion,the knowledge graph is formed.For performance reasons,the Neo4 j graph database and MongoDB are used.In order to reduce the maintenance cost of the knowledge graph update,this paper implements the automatic update system,and the system can automatically extract knowledge from new judgments.And then update the knowledge into the knowledge graph.In the process of constructing the knowledge graph,the entity and the attribute labeling data are extremely scarce for the judgment text,and the extraction result with the open extraction tool on the judgment text is not satisfactory.According to the characteristics of the original data,this paper uses different strategies to extract different knowledge.For structured data files,analyze files directly to gain knowledge;For the strong knowledge in the judgment documents.Summarized the rules and using rule-based method to extraction;For weak rules knowledge,extracted by machine learning method.Entity extraction from unstructured data is the key research content of this paper.Based on the sequence labeling method,this paper proposes BiLSTM-CRF model based on entity boundary feature and BERT-CRF model.And constructed training corpus for this method.The experimental results show that both models have relatively good recognition results.