Research On Key Technologies Of User's Legal Consulting Intrntion Understangding Based On Neural Network

With the development of artificial intelligence technology and the increase of people's demand for legal services,it's an good way to apply artificial intelligence technology to the field of legal services to build a legal consultation and service system with highly intelligence,where people can obtain more convenient and rapid legal services and save more human resources investment in the legal service system.Among this task,understanding and identifying the user's intention of legal consultation is an important part.The accurate understanding and recognition of user intentions can provide constraints by determining feedback categories and narrowing the scope of retrieval,which plays a key role in more accurate information retrieval and result feedback in later tasks.In this paper,we focus on the understanding and recognition of users' intention of legal consultation based on deep neural network.The research is mainly carried out in three parts: data modeling and formal representation of users' consultation text,the algorithm research of users' intention understanding of legal consultation based on deep neural network,and the algorithm research of users' intention understanding of legal consultation based on word vector model that oriented to polysemy.Data modeling and formal representation of users' consultation text is the formal representation of user consulting data,including data acquisition,text preprocessing,and word vector training.After that,users' consulting data is transformed into wordlevel distributed representation,which provides computable input form for subsequent intention understanding tasks.In the research of user intention understanding algorithm based on deep neural network,we propose a user's intention understanding model based on attention mechanism and convolution neural network?The attention mechanism is based on Scaled Dot-Product Attention,and the convolution neural network is used to extract and classify text features.Then the feature distribution is fitted through multi-layer linear network,and the probability of each category is output through the softmax layer.In the experiment of this paper,the accuracy of this model on the user's legal consulting data set is improved compared with that of softmax regression classifier and TextCnn text classification model.The accuracy of this model is improved by 2.29% compared with that of softmax regression classifier.In the research of the algorithm of legal consulting user intention understanding based on the polysemy-oriented word vector model,we analyse the principle of polysemy semantic acquisition in ELMo language model,Bert language model and FastText classification model,and studies the algorithm of user intention understanding on this basis.Experiments show that the accuracy of the word vector model based on polysemy-oriented semantics is relatively high,and the accuracy of Bert model reaches 79.84%.