Research And Application Of Key Technologies Of Chinese Intelligent Question Answering System For Restricted Domain

Intelligent question answering system is an advanced form of information retrieval for natural language processing.Based on a large number of corpus databases,a suitable data model is selected and constructed to realize the dialogue between man and machine,providing users with accurate answers to questions.In recent years,with the continu ous rise of artificial intelligence technology,intelligent question answering has become a hot topic and attracted more and more scholars.Therefore,how to make the computer understand the problem described in the Chinese language and the system can auto matically give a quick and accurate answer to the problem is the most difficult problem.At present,the research work of intelligent question answering system based on Chinese context is still in the continuous development stage,and a mature intelligent question answering system has not yet appeared.For the purpose of improving the accuracy of the system,it is usual to combine specific domain knowledge to assist the computer in understanding the question and answer process.By setting restricted domains and constructing a knowledge base of questions,the positioning of the answers to the questions can be more accurate and efficient.Based on this,this article mainly researches and implements an intelligent question-and-answer system for the "Computer Op erating System" course according to the actual needs of distance network education in colleges and universities.The followings are the main contribution and innovation of this paper:1.Based on the characteristics of the limited domain of the "Computer O perating System" course,for the sake of avoiding the error of question matching in the word segmentation process caused by the phenomenon of "inclusion",this paper designs a character-level neural network termed Enhanced Representation Neural Network(ERNN),this network model uses feature extraction and feature reorganization for matching sentences,which not only avoids the word segmentation step but also improves the accuracy of question matching;2.In order to improve the question answering system’s ability to recognize diverse forms of questions and reduce storage space overhead,this paper uses knowledge graphs and deep learning techniques to design a Knowledge Based Joint Neural Network(KBJNN)for sentence matching,which significantly improves th e accuracy of the question answering system;3.In this paper,the proposed matching algorithm has been verified by experiments,and extra ablation experiments have been carried out.Based on the experimental results,considering performance and robustness comprehensively,this paper implements the "Operating System" question and answer system based on ERNN and the compound similarity algorithm(ERNN+)...