Research On Intelligent Sentencing Methods Based On Deep Learning

With the advent of the era of big data,the judicial system has been hard to satisfy the growing judicial needs of the people.The upsurge of deep learning have pushed intelligent sentencing to the frontline in judicial field.The sentencing method based on artificial intelligence can absorb the content of the latest sentencing implementation rules and it can give the sentencing results rationally and objectively,which allows judges to make reference when dealing with sentencing cases,and allows more people to conduct judicial supervision.Moreover,intelligent sentencing can achieve significant results in judicial reform.Therefore,research on intelligent sentencing methods based on deep learning emerge as the times require.In order to solve the problems of traditional judicial sentencing,this paper incorporates deep learning into intelligent judicial sentencing and proposes a comprehensive network fusion model based on massive legal documents.We focuse on the three tasks of predicting judicial sentencing and design the task topology of this research.The proposed method combines multiple networks,e.g.,recurrent neural network,convolutional neural network and hierarchical network in the procedure of sentencing prediction based on the pre-trained word vectors.Experimental results justify that this method achieves good performance.The intelligent sentencing method in this article includes three tasks,namely,conviction,law article and the defendant's prison term based on the description of the case.Firstly,the pre-processed legal document data are used to train static word vectors and dynamic word vectors respectively,and then we design the corresponding deep network models include the convolutional neural network,the bidirectional recurrent neural network based on the attention mechanism,and the combined network of the recurrent and convolution layer,hierarchical attention network,etc.In addition,in order to consider the correlation of three tasks,the information of other tasks is embedded in the network of the corresponding task.For the task three,we use the post-classification regression to significantly improve the accuracy of regression.Finally,in order to avoid the limitations of a single model and enhance the effect of the model,this thesis further designs a model fusion based on three-way decision,which enhances the model effect through multiple decision and weighted fusion of the combined model.This thesis conducted a lot of experiments to verify the performance of the algorithm.Experimental results show that the method of sentencing prediction based on deep learning is significantly better than traditional machine learning methods,and the model structures designed in this thesis significantly improve the model accuracy.Finally,the corresponding model is analyzed and selected for model fusion based on three-way decision,and the global accuracy can reach up to 93%.Therefore,the research method proposed in this thesis can effectively predict intelligent sentencing tasks and improve the efficiency of judicial sentencing.