Legal Case Retrieval Method And Application Based On Deep Learning

A legal case is a judgment document,which contains important information such as the cause of action,the facts,the opinion of the court,and the judgment result.Because of its authority and rigor,it can provide an important reference for legal workers.The legal case retrieval system is one of the most commonly used tools by legal workers.It can help legal workers quickly retrieve the required cases from the huge legal case database and improve work efficiency.There are two main problems in the current legal case retrieval system.First,deep learning-based retrieval models have achieved significant performance improvements,requiring a large amount of manually labeled data to optimize deep model parameters.However,legal cases have the characteristics of strong professionalism and long content,which makes manual data annotation expensive and time-consuming.Second,legal case retrieval requires a lot of legal expertise,and the current legal case retrieval model aims to learn the semantic information of legal cases and cannot effectively utilize legal expertise.In view of the above problems,this thesis mainly conducts the following researches:1.Proposing a legal case retrieval model based on the prompt paradigm.In order to reduce the dependence of the deep retrieval model on a large amount of labeled data,the method based on the prompt language paradigm is applied to the deep legal case retrieval model,and the retrieval task optimization objective is converted into a form consistent with the language model pre-training objective,and a multi-task Prompt templates and multi-stage training strategies to improve the generalization ability of the retrieval model.2.Proposing a legal case retrieval algorithm and model based on legal common sense knowledge graph.Aiming at the problem that the case retrieval model lacks the ability to utilize legal expertise,a large-scale legal common sense knowledge graph is constructed unsupervised by using large-scale legal case corpus,and a query enhancement algorithm based on the legal common sense knowledge graph and an integrated legal common sense knowledge graph are proposed.The legal case retrieval model,thereby improving the modeling ability of the legal case retrieval system for the query text input by the user.3.Developing an online legal case retrieval system.Aiming at the problems of insufficient understanding of input queries and low retrieval efficiency in the legal case retrieval systems on the market,the above-mentioned legal retrieval models and methods are integrated and an online legal case retrieval system is developed.The system can parse the user input query and enhance it with knowledge graph,which has higher recall and precision than the existing legal case retrieval system.