Research And Implementation Of Legal Case Recommendation Service System Based On Deep Learning

With the deep integration of artificial intelligence technology and the field of justice,judicial intelligence has been developing continuously.Judicial intelligence can not only assist judicial practitioners to improve their work efficiency,but also better serve the public and promote the convenience of justice.As one of the basic tasks of judicature intelligence,the recommendation of similar case can assist sentencing decision-making and standardize judgment standards to a certain extent.With the advent of the era of big data,legal data is becoming more and more open and transparent.The construction of a recommendation service platform for legal similar case can further promote the intelligence and convenience of justice.Through combining natural language processing technology,recommendation strategy and deep learning,this paper researches and implements a service system of similar case recommendation.Firstly,this paper researches and analyzes the related technologies of natural language processing.Then it performs the function of data segmentation and feature representation for the judgment document data using the rule-based information extraction method,the word embedding and other natural language processing technologies.And according to the characteristics of legal documents and the advantages and disadvantages of the recommendation method,the content-based recommendation method is used to build a similar case recommendation system.In order to further improve the accuracy of the recommendation results and make better use of the semantic information of the documents,a classifier based on deep learning is proposed.According to the description of the case classifying the documents,in more than 100,000 real data,it make the classification experiment contrast using three models including Word2vec_CNN,FastText_CNN,and Word2vec_LSTM,and do the recommendation experiment combined with the case classification model.The experimental results show that Word2vec_CNN had the best classification effect.The recommendation results of similar cases show that fusion classification information get a better recommendation result comparing with the results which only using case descriptions.Finally this article designed and implemented a recommendation service system of legal similar case from an application point of view.The main function of this system can get the similar case recommendation and case classification by entering the case description,also can get the extraction results of each part of the document through the formatting and segmentation of legal judgment documents.