Research On Intelligent Judicial Judgment Prediction Method Based On Multi-information Fusion

The rapid progress of artificial intelligence,natural language processing technology and the publication of legal case documents not only provide a good data foundation and technical precipitation for the research of intelligent justice,but also promote it to become a research hotspot in the field of natural language processing.In view of the multiple tasks included in intelligent justice,the characteristics of judicial data,and existing research difficulties,a multi-task learning model was proposed.On the basis of the multi-task learning model,its integrated chapter structure information,legal text information and charge keyword information,which can not only improve the effect of intelligent judicial judgment,but also solve the current problems in intelligent justice.This article mainly includes the following three aspects of research content.1.Intelligent judicial judgment prediction model based on multi-task learning model and chapter structure information.Firstly,considering the close relationship between the two tasks of charge prediction and law recommendation,we adopt a multi-task learning method to jointly model the two tasks of charge prediction and law recommendation.Secondly,because legal texts are paragraph-level texts,in order to take chapter structure information into account,we integrated chapter structure information on the basis of a multi-task learning model.The experimental results show that the modeling method of multi-task learning is better than single charge prediction method or law recommendation method.In addition,the chapter structure information is also effective for judgment prediction.2.Intelligent judicial decision prediction model with law information.Referred to the relevant law before the judge judges the case,it can be seen that the information of the laws has a positive effect on the prediction result of the intelligent judicial judgment.Therefore,we further fuse relevant law information on the basis of the multi-task learning model that integrates chapter structure information.Fusion methods include vector similarity,pre-training model,and attention mechanism,among them,the attention mechanism method is proposed to reduce the error propagation problem of the first two methods.3.The prediction model of intelligent judicial decision based on the keyword information of charge.The problem of confusing charges in the research of intelligent judicial judgment prediction is the difficulty of judicial judgment research.We found that although the text descriptions of the confusion charges are very similar,the keywords can effectively distinguish the confusion charges,so we propose to integrate the crime keyword information to solve the problem of confusion charges.Firstly,we use a variety of methods to extract the accusation keywords from the legal text and automatically build the accusation keyword table.Secondly,we integrate the charge keyword information into the above model.The experimental results show that the charge keywords have played a very effective role in solving the problem of confusing the charge,and effectively reduce the misjudgment caused by the confusable charges.