Research And Application Of Intelligent Question-and-answer System For Specific Domain

In the Internet age,search engine has become the main way for people to obtain information.People just need to input the keywords they want to look up,search engines can return to the relevant web pages.However,search engines also have some shortcomings,such as many related pages,cumbersome content,etc.,which need a more simple and efficient way to obtain information,so question answering system came into being.In the face of the huge amount of information in the network,the question answering system in specific domain has attracted much attention because of its high feasibility and wide use.This paper is devoted to the research of intelligent question answering system for specific domains.The main research content is divided into the following three parts.(1)The method of knowledge base automatic construction for specific domain.Firstly,a web page information extraction method based on SVM and text density is proposed,through which the web page text information can be effectively located and automatically crawled;secondly,a method of text topic division and TextRank central sentence extraction is proposed,which can automatically extract the central sentence of text topic and store the central sentence as question answering information in the knowledge base.The method proposed in this paper has achieved good results in experiments and practical applications.(2)Intelligent question answering based on knowledge retrieval in a specificdomain.The Boolean inverted sorting index method is used to filter out the candidate response set from the knowledge base,and then use the similarity algorithm to accurately filter the candidate response set to obtain the best response and return it to the user;this paper proposes a Jaccard based on word vector and position encoding Similarity algorithm.In the experiment,this method not only has higher accuracy,but also has better recognition ability for word order.(3)Multi-round integrated question and answer based on deep learning.This article proposes a multi-round integrated question-and-answer model based on the combination of retrieval and generative methods for small conversations.The experimental results show that the integrated model performs better than the single model to a certain extent,and this model is applied to the chat module of the question and answer system in a specific domain,and has achieved good results.Based on the above research,a domain specific intelligent question answering system is developed.Through the system test,the system can not only answer the related questions in the history field accurately,but also perform well in chatting and meet the basic needs of users.The work of this paper has important theoretical and practical significance for promoting the domain specific intelligent question answering system.