Design And Implementation Of Similar Case Discovery System Based On Network Representation Learning

With the establishment of the China Judgment Web and other judgement publication websites,there are a lot of judgements on the Internet.Finding similar judgements from a large number of judgements quickly and automatically becomes the demand of the majority of law-related practitioners.Similar case discovery aims to find similar judgements.It not only can help judgers to find similar judgements and sentence similar cases similarly for ensuring legal justice,but also can assist defense lawyers to look for the similar judgments to defend the defendant.The earlier similar case discovery ignored the professional knowledge contained in judgements,but judgements belong to the text of the professional field.The utilizing of legal knowledge in professional fields can help to find similar judgements accurately.Accordingly,this thesis proposes a method that combines legal knowledge to find similar judgements.The main work of this article includes:(1)This thesis selects legal terms from Baidu Encyclopedia and identifies the entities in the judgments,and uses the entity link technology to construct a heterogeneous network of legal terms and judgments.(2)A heterogeneous network representation learning method is proposed for the heterogeneous network of legal terms and judgments.We get the vector representation of the judgment document in the network through this method,and according to the cosine similarity of the judgment vector,we get the similar judgment of the judgment.(3)In this thesis,we set up multiple experiments to verify the effectiveness of our method.The results of experiments show that the method proposed in this thesis improves the effect of similar case discovery,and verifies the role of legal knowledge in similar case discovery.(4)Based on the method proposed in this thesis,a similar case discovery system is designed and implemented.After testing,the system meets all functional and performance requirements.In conclusion,this thesis proposes a similar case discovery algorithm based on network representation learning and implements a similar case discovery system which improves the result of similar case discovery compared with existing system.