Research And Implementation Of Domain Question Answering System Based On Deep Learning

With the advent of the era of big data,the traditional search method based on keywords and returning a series of web pages gradually fails to meet people's demand for information.Question answering system,to a certain extent,makes up for the shortcomings of search engines,and is currently a research hotspot in the field of natural language processing.In recent years,the continuous development of deep learning technology has brought many breakthroughs to the intelligent question and answer field.Therefore,the question answering system based on deep learning has become the hottest research direction in natural language processing.This thesis aims to build a domain-specific question answering system using deep learning technology.First of all,the deep learning model is used to complete the intent recognition and slot extraction in the question comprehension,and then the question is transformed into a structured query that can be understood by the knowledge base,and finally the answer is retrieved from the constructed knowledge graph of movie domain.The specific research content is as follows:Firstly,this thesis introduces the background and significance of this topic,briefly expounds the classification of question answering system,puts forward the domain-specific question answering system based on knowledge graph,and introduces the development status of question answering system at home and abroad,as well as several implementation methods of question answering system based on knowledge graph.In the next two chapters,this thesis begins to study the question comprehension part of question answering system based on deep learning.In the second chapter,this thesis proposes a convolutional neural network model based on Word2 Vec to realize the intent recognition of questions.Firstly,the word vector model is trained to complete the distributed representation of words.Then the CNN model is built to extract the question features and complete the intent classification of the questions.Finally,the effectiveness of the construction model is proved by comparison experiments with other classification models.In the third chapter,this thesis proposes a Bi-LSTM-CRF model for the problem of slot extraction.This model can solve the problem that LSTM can't learn context information well and Bi-LSTM can't carry out label constraint.The experimental results show that the Bi-LSTM-CRF based sequence labeling model effectively improves the accuracy of question sequence labeling,and thus achieves the purpose of improving the accuracy of question slot extraction.After completing the research on the understanding of questions based on deep learning,Chapter 4 begins the requirementanalysis and outline design of this question answering system,presents the system framework and gives the module partition.In the fifth chapter,this thesis designs and implements a film domain question answering system based on the knowledge graph which uses the WeChat Public Account as the terminal,and introduces the implementation process of each module in detail.The actual operation results show that the system can accurately answer the relevant questions raised by the user in real time,which proves the feasibility of the designed algorithm.