Research And Implementation Of Intelligent Assistant System For Case Settlement Based On Deep Learning

As a research hotspot in the research of machine learning algorithms,deep learning is a machine learning technology that simulates human brain neurons and has a hierarchical architecture.In recent years,deep learning techniques have made breakthroughs in sub-field applications such as speech recognition,picture recognition,and natural language processing.With the support of big data,deep learning technology has developed rapidly in the field of court intelligence,which has increasingly demonstrated its significance and broad application prospects.Combining deep learning techniques with various aspects of trial practice can greatly enhance the intelligence of the people’s courts.Level of development.At present,the problem of fewer people in the court case is very serious.In 2016,the number of criminal judges per capita in the grassroots courts was more than 200 per year.Therefore,it is urgent to improve the efficiency of judges’ trials and speed up the trial of cases.Because deep learning technology has shown great advantages in text analysis,the use of deep learning technology to assist judges in completing case trials has become one of the key research issues in the construction of smart courts.Based on the above requirements and background,this thesis builds a intelligent assistant system for case settlement,introduces deep learning technology to recommend the case hearing result information for the judge,and assists the judge to conclude the case.The system not only reduces the pressure on the judges,but also avoids the occurrence of different judgments in similar cases.In this thesis,the deep learning technology is researched.The CNN model is used to construct a regression model based on neural network.The CNN-RNN fusion model is used to construct a text classification model based on neural network.The specific work of this thesis has the following two points:1)This thesis builds a intelligent assistant system for case settlement,which helps judges judge the case through incremental training of deep learning models.The system includes three subsystems:sample import and model initialization subsystem,trial decision subsystem,and case trial model incremental training subsystem.First,the system completes the construction of the case trial model through model configuration,helping the model manager to complete the initialization of the case trial model under different categories.Then,the system can recommend the trial result information according to the entered case trial information when the judge conducts the case hearing,and assist the judge to complete the case conclusion.At the same time,the judge can correct the recommended abnormal information,and the revised information will subsequently modify the model library.2)This thesis verifies the application effect of the intelligent assistant system for case settlement.Application verification includes verification of the regression effect of the sentence and verification of the classification effect of the settlement method.For the regression effect verification,this thesis compares the application effect of CNN with ANN and SVM,and finds that the CNN used in this thesis has a significant improvement in the accuracy of the recommended sentence.For the classification effect verification,this thesis compares the CNN-RNN fusion model with the traditional machine learning model,and finds that the fusion model used in this thesis has a significant improvement on the recommendation accuracy of the settlement method.