Design And Implementation Of A Matching System For Similar Legal Cases

With the development in artificial intelligence algorithm and infrastructure,more and more companies start using artificial intelligence to help to improve their product and service.Traditional legal service fails to be provided for the public due to its high entry bar,heavy workload,and the shortage of participator,which makes artificial intelligence having huge application space.In 2017,State Council of China promulgated the "New Generation of Artificial Intelligence Development",which mentioned"smart courts".In smart courts,artificial intelligence is used to automate tasks including retrieval,review,and generation.Many companies and research teams have entered the field.This project aims at designing and implementing a system for matching judgment documents of similar legal cases,in order to provide judicial precedents for lawyers and judges.This paper firstly introduces the features of current legal-intelligence products,the application of artificial-intelligence in this field and the development of text matching technologies.Then there are three different case-matching models proposed in the paper.The first in based on case elements extracted by programmed rules.The second is a representationbased model.The third is an interaction-based model whose mechanism this paper is the first to propose.In the second and third model,BERT is used.In the experiment,the second model wins the second place in the rank of CAIL2019-SCM,and the third model wins the first place.Both model is of lower computational complexity than the ex-champion.Through ensemble learning technologies,the case-matching model can be of high accuracy and efficiency.In order to serve more people,a system with above-mentioned models embedded is designed,implemented and evaluated in the paper.It includes requirement analysis,design,implementation details and testing of the system.The evaluation shows that this system can manage models,tasks,judicative papers and so on.The system also has fast response,security,scalability,maintainability,and a friendly user interface.Other legalintelligence products can access the system via its application programming interface.By far the system has been integrated into a legalintelligence platform for People’s Procuratorate.