Design And Implementation Of A Law Prediction System Based On Multi-Model Fusion

The case-filing registration system is a major measure in China’s judicial system reform.The provision clearly states that the court only conducts a formal review of the petition,and all those that meet the requirements should be accepted and the case should be closed according to the law within the prescribed time limit In addition,people’s awareness of the rule of law has generally increased,resulting in a substantial increase in the number of cases accepted by the courts.In the traditional judicial trial process,the judge makes a trial conclusion on the case based on the facts of the plaintiff and the defendant.Legal trials require judges to have professional knowledge reserves,and also need to consult a large number of relevant laws and regulations according to the facts of the case,and cite relevant laws and regulations as the basis for judgment.The comprehensive effect of various factors makes the daily work of judges and other judicial personnel face many challenges.The China Judgment Documents Network has realized the electronic record of the national case trial results.Although the Judgment Documents website has its own retrieval function,the returned results are all qualified judgment documents.For users who lack legal expertise,the retrieval is difficult to achieve the desired effect.At this time,we need a legal prediction system to help judicial workers improve their work efficiency and meet the legal consultation needs of the parties.The traditional legal article prediction scheme abstracts the legal article prediction task as a text multi-label classification problem,and uses algorithms such as TextCNN or TextRNN to predict the corresponding legal articles of legal facts.However,due to the different training data sets and tasks,the generalization ability of the algorithm is poor.The main difficulties faced by the law prediction system are:1.The structure of the judgment documents is similar,and the similarity is high;2.There are little differences between different laws;3.Usually,a case will involve multiple laws.In order to solve the above three problems,this paper innovatively proposes a law prediction algorithm based on multi-model fusion,which can determine the final output law according to the threshold value.The specific work of this paper is as follows:(1)The single-model law prediction algorithm model integrating the features of the lawThe traditional law prediction method often ignores the information of the law itself in the prediction process.Therefore,this paper proposes a legal article prediction algorithm that integrates legal article features.This method extracts the text features of legal articles and legal facts through the TextCNN model,which significantly improves the effect of legal article prediction.For the fixed field size of the convolution kernel of the CNN model,it is impossible to model longer sequence information,and the hyperparameter adjustment is cumbersome.This paper proposes a law prediction algorithm based on the LSTM attention mechanism model,which combines Attention to measure the contribution of each keyword to generate a law prediction list.Compared with the TextCNN model,the Fmacro index of this model is significantly improved.(2)Law prediction algorithm based on multi-model fusionIn view of the problem that the single-model law prediction algorithm is not very stable in test results under different data sets and has poor generalization ability,this paper proposes a multi-model fusion law prediction algorithm.,the TextCNN or TextRNN with different parameters are fused by probability addition to achieve the effect of model complementarity.Experiments show that the weighted fusion algorithm Fmacro and Fmicro indicators are better than the single model algorithm.(3)Legal forecasting system based on multi-model fusionBased on the multi-model fusion legal forecasting algorithm,the research results are applied to practical projects.Understand the actual functional requirements through literature review and demand analysis,use Python as the back-end development technology,and use the Uniapp-based framework as the front-end development framework to enable the project to run on multiple platforms,and finally realize the application of law prediction that can be used on multiple platforms.This paper designs and implements a law prediction system based on multi-model fusion algorithm.The algorithm integrates multiple law prediction algorithms such as TextCNN and TextRNN based on fusion law features.When designing the algorithm,a series of experiments were carried out,and the algorithms were compared and analyzed.The results show that the law prediction algorithm based on multi-model fusion has higher accuracy and stronger applicability than other single-model algorithms.Scalability can be further improved in future applications as needs change.