Research On The Entity Relationship Extraction Of Cases In The Field Of Legal System

With the sustainable progress of the process of constructing the legalistic cause,and relying on the rapid development of computer and Internet technologies,a large number of real-life cases from the legal field news reports were presented on the Internet.The information about entities and the relationship between entities in cases of the legal field can be arranged together to construct the relationship network.But sometimes in the relevant reports important information useful for quickly browsing and sorting out ideas is always overwhelmed by the unrelated information.Therefore,it is worth our research to extract the valuable information about the relationship between entities in cases of the legal field.The extraction and integration of these useful information requires the use of information extraction techniques.Given the situation of the existing methods to extract the valuable information about the relationship between entities in cases of the legal field,there are many problems with the existing methods,such as scarcity of high-quality annotation corpus,and the limited size of data sets of cases with no noise in legal area,and the data sets acquired by the traditional distant supervision method contains a large number of noisy tagged sentences,as well as the artificial feature engineering research is not yet fully mature,and so on.In terms of the current situation of labeling noise of distant supervision method,the approach to filter the noises in the data sets of cases in the legal area utilizing reinforcement learning techniques is proposed in this paper,and then an approach to o extract the relationship between entities in cases of the legal field based on convolution neural network in the noise-removed corpus,is also put forward,and the method of extracting the relationship between entities based on co-train is figured out.Finally,drawing entities relationship networks based on the graph database is put forward.The following research works are mainly accomplished:(1)The noise-filtered model based on reinforcement learning techniques.Given the existing high-precision annotation corpus of cases in the legal field are scarce,the model to filter the noises in the data sets of cases in the legal area utilizing reinforcement learning techniques is proposed in this paper.Firstly,prepare a scalable corpus of cases in legal area,and by the means of knowledge base,construct automatically data sets using distant supervision method.Then filter the sentences with the most possible wrong sentence tag and pick out the sentences with the most correct sentence label using the idea of reinforcement learning techniques to promote the accuracy of the correct sentences marking of annotation data sets on the acquired annotation data sets with noises.(2)The relationship classification model based on convolution neural network model.Many existing traditional supervised learning methods of entity relationship extraction largely rely on the features of entity relationship classification provided by extra natural language processing tools.It leads to form errors cumulative effect.Therefore,in terms of this question,an approach to o extract the relationship between entities in cases of the legal field based on convolution neural network is put forward.It can automatically learn the classification features of the relationship between entities and extract the relationship between entities in cases of the legal field based on convolution neural network in the noise-removed corpus,combining the special field features of news coverage of cases in the legal field,is able to improve largely the effect of extraction.(3)Representing and displaying entities relationship networks based on the graph database.The entities relationship networks in legal field is a complex model,which include entities and semantic relations between entities.But traditional database to save these relationship between entities and express these complex relationships could not meet the present application needs.In this paper,an approach based on Neo4 j to save the entities relationship networks is proposed,which can represent and display relationship between entities in legal field by the graph database.