Legal Text Knowledge Mining For Civil Law

With the rapid development of the artificial intelligence industry,artificial intelligence has begun to be widely used in various fields,such as the medical field,the judicial field,and the transportation field,and it has played an increasingly important role in people’s daily life and work.Among them,the applications in the judicial field include legal question and answer,document production,case analysis and reasoning,etc.However,these applications are more focused on starting from judgment documents and case descriptions,and have less processing and application of the original text of the law.Starting from the civil law text,this article constructs a legal knowledge map based on part of the original legal text,and focuses on extracting legal terms from the original legal provisions based on natural language processing technology and digging out the relationships between terms.This topic is derived from the national key research and development plan "Multi-dimensional evaluation of case handling based on centralized case management and procuratorial disclosure technology research"(2018YFC0830700),to carry out in-depth research on the understanding of civil law texts in the judicial field.There are two major starting points for actual needs.One is to identify specific legal terms in the legal text,and the other is to dig out the relationships between the identified legal terms.In the task of identifying legal terms in legal texts,due to the lack of definitions of legal terms and related data sets,a definition of legal terms for selected laws was proposed,and the corresponding legal terms were marked according to the definitions.data set.According to the legal terms obtained in the selected law,this paper forms a legal term dictionary,integrates the knowledge of the dictionary domain into the legal term recognition model,introduces the pre-training model BERT to learn the text characteristics of the law,and introduces the deep learning model Bi LSTM to obtain the legal text context information,Introduce CRF as a classification model to define categories of legal terms,so as to realize the task of identifying proper names of legal terms.In the task of extracting the relationship between legal terms,based on the understanding of legal knowledge and the definition of legal terms,a sequence labeling method is designed.The training set of relationship extraction is obtained by labeling legal provisions similar to term recognition.Greatly reduce the data collection work.The same model of term recognition is used to train and predict the relational data set,and the result is reprocessed according to the labeling method to obtain the relationship between terms and complete the task of extracting the term relationship.Combining the above methods,this article designs and develops a legal knowledge map system based on partial legal texts,showing the comparison of legal term identification,legal term relationship extraction,and the results obtained after the identification extraction model is processed in foreign laws and Chinese legal terms.