Design And Implementation Of Intelligent Question Answering System For Tourism Field

In recent years,with the development of the Internet,the traditional tourism industry has ushered in a new change.Tourist attractions have begun to apply the development mode of "Internet + tourism ".At present,in the field of tourism,it is a challenge that how to obtain information quickly and accurately.In the past,tourists mainly obtain information through traditional search engines,but it cannot meet the needs of fast and accurate.Unlike traditional search engines,the Q&A system directly returns the answers to a question,and can meet the actual needs of tourists.Presently,the data sources of the question and answering system mainly include Q&A pair data,document data,and structured data.Compared with the previous two,the answers returned by the Q&A system based on structured data are more accurate and concise.Traditional relational databases face challenges in structured data storage,but knowledge graphs can solve the above problems.Generally,the question answering system based on knowledge graph uses the knowledge graph as the answer source.The system allows users to obtain the knowledge from the knowledge graph through natural language question.This dissertation aims to implement an intelligent question answering system for tourism field based on knowledge graph.Among them,the basic knowledge graph is based on the tourism knowledge graph NMTKG-1 constructed by Inner Mongolia Autonomous Region Key Laboratory of Mongolian Information Processing Technology from 2018 to 2019.The main research contents of this dissertation are as follows:(1)Expand the knowledge graph of tourism.NMTKG-1 does not cover the peripheral information of tourist attractions.In order to improve the performance of Q&A system,this dissertation adopts a knowledge expansion method based on Baidu Map API.It shows that this method can effectively expand the peripheral information of tourist attractions.(2)An intelligent question answering system based on tourism knowledge graph is implemented.There are two key steps from the natural language question to answer generation,namely named entity recognition and attribute selection.In the named entity recognition task,this dissertation proposes BERT-BLSTM-ATT-CRF model.This model uses BERT model to learn character features of text,and BLSTM model to learn context features of text,and attention mechanism to focus on key information in text,and CRF model to obtain the global optimal output sequence.Experimental results show that it can effectively improve the system performance.Meanwhile,for the task of attribute selection,this dissertation proposes an attribute selection method based on GSA＿SMCNN,which uses the BGRU model to capture the semantic-level correlation between questions and attributes,and the CNN model to capture the literal-level words similarity between questions and attributes.Experimental results show that this method can effectively improve the system performance.(3)Implementation of travel assistant APP.This dissertation provides the travel assistant We Chat applet for the user.The front-end of the APP is developed by Vue.js language,the back-end is developed by Python language,and the tourism knowledge graph NMTKG-2 is stored in the form of Neo4 j database.The functions provided by the APP include travel information Q&A,weather information Q&A,and ticket information Q&A.