The Research On Hierarchical Classification Of Legal Questions Based On Deep Learning

With the development of society,laws and regulations have been increasingly perfected,and more and more legal-related incidents have been encountered in people's lives.When people seek for legal support,it is important to understand the legal areas in which they ask the questions.When people ask questions on the forums or ask people who have legal backgrounds,they will need a certain amount of waiting time before they can receive a reply.This often delays the time for solving problems.Therefore,it is very meaningful to classify these legal issues automatically and efficiently.Text classification based on deep learning is a relatively new method in the study of text classification.In the study of text classification combined with deep learning,deep learning gradually shows its unparalleled charm.Similar to human beings,deep learning has the capability of automatic learning through extracting features in the continuous training stage.Compared with traditional text classification methods,it can train more efficiently and obtain more accurate classification result.First of all,it bases on the analysis and introductions of the deep learning and the text classification,and selects the classification of Chinese legal questions as the research background and then studies the hierarchical classification of legal questions based on deep learning.First,we propose a multi-task Convolutional Neural Network(CNN)for classification of Chinese legal questions with trainable word embedding where coarse grained classification is the main task and fine grained classification is the side task.Second,we develop a hierarchical classification model which takes the output of coarse classification as one part of the input for fine grained classification.We find that the side task can improve the accuracy and efficiency of the classification in a certain extent.Our experiments on the entire Chinese Legal Questions Dataset(LQDS)demonstrate the effectiveness of the proposed approach.Next,we verified the hierarchical classification method mentioned abo ve for datasets with a hierarchical label structure.We designed two experiments.First,we designed a hierarchical classification model based on LSTM on our original dataset LQDS.Second,we combined CNN-non-static and our hierarchical classification method on the classical sentiment analysis dataset(SST).Finally,the effectiveness and versatility of our hierarchical classification method are proved by experimental comparison.And our LSTM-based hierarchical classification model achieves the best results of both coarse-grained classification tasks and fine-grained classification tasks so far.