Auxiliary Sentencing Technology Of Criminal Cases Based On Legal Knowledge

As the problem of "more cases and fewer people" in the judicial field continues to intensify,the pressure of judicial workers continues to increase.In order to balance the relationship between the increasing number of cases and unbalanced and insufficient judicial manpower and judicial justice,the text focuses on the issue of auxiliary sentencing in judicial decision-making.It combines artificial intelligence and domain knowledge to analyze cases,and gives full play to the advantages of combining computer science with other disciplines.Our work strives to create a transparent and fair judicial sentencing,which is of great significance for maintaining judicial justice and building a federal society.This article starts with data processing in the field of judicial sentencing,and processes the judgment document data set and the CAIL2018 data set separately.It mainly includes data set acquisition,data cleaning,effective information extraction and text segmentation.Afterwards,in order to facilitate the expression of text data in the computer system,this paper uses multiple word vector models to conduct distributed vector coding of words,and selects the most suitable word vector model and its parameters for this task.In view of the inseparable relationship between the sentence judgment of the case in the judicial judgment and the laws,this paper designs and implements a model of auxiliary sentencing for criminal cases based on a multi-channel attention mechanism.It uses three channels: basic information of the defendant,fact description of the case,and articles,and vectorizes its text through a multi-head self-attention mechanism.After that,the document aggregator is used to fuse the information of multiple parties,and the basic information of the defendant and the fact description of the case are used to query the laws and regulations involved in the case,so as to achieve accurate sentencing according to the articles.The experimental results show that the model has certain advantages in terms of sentence prediction.Through the analysis of the model,we know that the model still has deficiencies in training efficiency and interpretability.Based on these problems,we propose a rule-enhanced auxiliary criminal sentencing technology based on the model of auxiliary sentencing for criminal cases based on a multi-channel attention mechanism.The model uses the sentencing rules to scale the sentence prediction results.It uses a hierarchical attention mechanism to express the facts of the case and the basic information of the defendant.At the same time,it uses a multi-task joint learning mechanism combined with forward feedback to learn the sentencing prediction results to the sentence prediction task.The experimental results show that the model is much higher in efficiency than other comparative models,and still has a greater advantage in the accuracy of sentence prediction.At the same time,in addition to the sentence,the output of the model contains the sentencing rules and calculation methods.The model has good interpretability for the sentence prediction results,and is convenient for users to understand.It has laid the theoretical foundation for practical application.