| As the development of AI technology,higher efficiency and automation level are achieved in various fields.For judicial area,"The Smart Courts" is being energetically developed in China,aiming to improve the abilities of case handling for judicial organs of Chinese government.Additionally,"The Smart Courts" can further ensure judicial fairness and efficiency as well as popularizing the law to the public.That is the important values that studies on application of AI in the judicial area.This article mainly includes the following three parts:1.Hybrid deep learning model based on pre-trained model.Pre-training model gets excellent performance on natural language processing tasks benefiting from the supports of novel model architecture,training methods and massive corpus.Therefore,variety of pre-training models are used on feature extraction on legal judgment texts.Charge will be predicted after inputting these features into classification models based on CNN and LSTM.This model is superior to the traditional model.2.The prediction model of prison term based on Event Graph.One disadvantage of traditional Knowledge Graphs is that it ignores the evolutionary laws between events while mainly focuses on the attributes and relationship knowledge of entities.However,Event Graph,of which nodes defined events and lines defined evolutionary relationships,make up for this disadvantage.This article builds the Event Graph through pre-processing texts of the judgment,feature extraction and event relation extraction.Then,the prison term is predicted by calculating the similarity between nodes.3.Auxiliary judgment system based on Event Graphs.With the Application of PyQt,a visualization system is built on the basis of the two models mentioned above.Inputting description of accident while outputting the auxiliary judgment with both predicted charge and prison term.This system will efficiently provide legal assistance to judicial staff as well as the public. |
PDF Full Text Request