Research And Implementation Of Campus Question And Answer System Based On Deep Learning

Questions and answers are very common and very important be haviors in people's daily life.As the technology and network become more and more important,more and more people begin to pay attention to how to let the machine automatically capture the answers they need in the massive information.Driven by this demand,the intelligent question answering system come into being,and has gradually become one of the hottest topics in the field of natural language understanding.Whenever talking about questions and answers,people ofte n think of another technology---search engine.Most search engines are based on keywords,which can neither identify the semantic information of the sentence,nor return the answer directly to the user,so the question answering system is more in line with people's daily needs.The key to the question and answer system research is to make the computer understand the meaning of human speech,that is,recognize the semantic information of the sentence.However,the current system based on thesaurus or regular question-answering system can not capture such semantic information,so we need to achieve by means of deep learning.Deep learning is a mo re complicated neural network,the typical deep learning includes Convolutional Neural Network and Recurrent Neural Network.CNN can extract the eigenvectors of sentences and RNN can understand the contextual information of sentences.They all help with the semantic understanding of sentences.This paper implements a campus question answering system for the field of Jilin University.The overall process of the system is: First,build a campus question and answer library for Jilin University,and the question and answer database contains many questions and answers.Then,when the user asks a question,the semantic similarity of questions and questions in the question and answer database is calculated.Finally,the answer to the question with the highest degree of similarity is returned to the user,and the relevant questions are recommended to the user.Among them,calculating the semantic similarity between sentences is the most critical step in the question answering system.This paper adopts the deep learning method to deeply study the sentence semantic similarity.Based on deep learning theory,the main work of this paper is as follows:(1)Two kinds of sentence semantic similarity algorithms based on deep learning are studied,which are based on convolutional neural network similarity algorithm and similarity algorithm based on recurrent neural network.(2)By combining the network model of CNN and RNN,a similarity algo rithm based on full information feature extraction is proposed.The algorithm uses a bidirectional cyclic neural network to obtain the context information of the words,and uses the convolutional neural network as the input to extract the feature vectors o f the sentences.Finally,the angle cosine of the two sentence vectors is calculated as the similarity.(3)In order to construct a Chinese corpus used for similarity model training,a set of semi-autonomous sentence similarity annotation rules was developed in this paper.Standard similarity score points were used to divide various typical semantic relationship intervals and improve manual work.Efficiency and accuracy when tagging corpora;(4)The sentence semantic similarity algorithm based on the full information feature extraction was applied to the question answering system to implement a campus question answering system for the Jilin University field.It achieved good results in answering questions related to Jilin University.