Research On The Recommendation Of Articles Of Law Based On The Content Of The Case

With the rapid development of technology,we are moving towards the information age,which is characterized by the application of technologies such as the Internet,blockchain,and a new generation of artificial intelligence.With the passage of time,the rapid development of artificial intelligence technology is promoting the development of many fields in China,such as medical treatment,service,education,and so on.In the judicial field,artificial intelligence can analyze and process a large amount of data,obtain a preliminary judgment on the case in a short time,help legal practitioners make judgments more convenient and concise,and can also use big data for countermeasures and control,as well as establish a crime spatial index model.In2016,the promulgation and implementation of the "Rules of the Supreme People’s Procuratorate on the Publication of Ruling Text by People’s Procuratorates on the Internet" has brought a significant breakthrough in the development and application of intelligent justice.The establishment of the Chinese Ruling Text Network has solved the problem of data volume that has been difficult to handle for a long time,meeting the preconditions for in-depth learning.Recommendation of legal provisions is an application direction in the judicial intelligence system,which refers to recommending relevant legal provisions through analyzing the specific content of the case and combining relevant legal provisions knowledge under given case conditions.In this thesis,the problems in the field of recommendation are studied as follows:(1)According to the uneven distribution of the diversity of legal cases,the method of data screening optimization is adopted.Through the selection and sorting of data,the top 5 types of law cases with the largest amount of data are selected.Meanwhile,the number of law cases involved in the case is screened,and some controversial and extremely tedious cases are removed.(2)In view of the low accuracy of similar law cases,the method of law recommendation based on the content of the case was adopted,and the Bert-At CNN model was proposed.The feature vector of the case was extracted through the BERT model,and the Text CNN model was improved to integrate the Attention mechanism into the Text CNN model.The Attention mechanism can efficiently draw out the most essential feature data from each dynamic feature vector,and then perform weighted fusion to bolster the "attention" of essential feature information,thus augmenting the precision of the model.(3)A law recommendation model based on similar case matching is proposed,and the idea of case matching is adopted.First,similar cases are matched,and then the law is pushed through similar cases.At the same time,the law information can be recommended and the information of similar cases in the past is given.In the experiment of this thesis,it is verified that the model of BERT-ATCNN in this paper can achieve the expected enhancement effect on the judicial data set,and can effectively distinguish the contents of similar articles of law cases,which proves that the proposed model is effective.At the same time,although the law recommendation model based on similar case matching can push similar cases,the accuracy of the law recommendation is still a little insufficient.