Research And Application Of Tourism Question Answering System Based On Knowledge Graph

With the continuous development of artificial intelligence technology,tourism intelligent question answering system,as one of the symbols of the development of tourism artificial intelligence,has been widely used in the field of tourism.In this paper,we use the Knowledge Graph and natural language processing technology to build a tourism question answering system based on Knowledge Graph,which can provide users with concise and accurate answers and can improve the efficiency of information retrieval to a certain extent.The main work of this paper is as follows:We used crawler technology to collect relevant data information about tourist attractions in Guizhou Province from all major tourism websites and encyclopedia websites,proposed an entity alignment method based on address attributes,fused the entity data collected from each platform,and constructed a knowledge graph in the field of tourism in Guizhou province.In the question answering model based on Knowledge Graph,For the lack of static and single of the input characteristic vectors in entity recognition basic model(BiLSTM+CRF),we adopt the output vector of pre-trained BERT model as the inputlayer feature vector of the BiLSTM+CRF model.Meanwhile,in view that the sparsity of corpus text information may affect the performance and effect of the model,we adopt n-best decoding output for the CRF layer,resorts the output candidate label sequence after compressed,uses Bi GRU to extract the compressed dense sequence in rearrangement,and realizes the optimization and improvement of the model.The experimental results show that the improved model significantly improves the accuracy of entity recognition in the tourism field compared with the basic model,thus improving the accuracy of the tourism question answering system.For the sentences whose answers cannot be queried in the Knowledge Graph,we collect a large number of Q & A pairs datasets as a standard question-answer corpus to realize the extension of the data source of the question answering system.For the shortcomings of BERT model,in calculating the similarity distance between input question and standard question,it consumes a lot of computational resources and has slow response speed.we propose a SBERT network model based on shared parameters and vectorize all standard interrogatives to realize offline calculation of text similarity.The experimental data show that the calculation rate of text similarity and accuracy of the proposed model are significantly improved compared with the original model.The Intelligent Question Answering Model of this paper is constructed after combining the question answering model based on Knowledge Graph and the question answering model based on standard question answering corpus.Then an intelligent question answering system based on Knowledge Graph is built,and applied to the query of attractions in major scenic spots in Guizhou Province.A large number of experiments show that the scheme is feasible,it has certain reference significance for the development of "intelligent" travel in tourism industry..