Research On Knowledge Question Answering Method Based On Tourism Knowledge Graph

Today,the Knowledge Graph has become an important tool in the age of big data,widely used in the next generation of search engines,intelligent question and answer,recommendation and other fields.Question answering based on knowledge graph has always been a hot topic in the field of natural language processing.Due to the growing maturity of natural language processing technology and the emergence of high-quality general domain knowledge graph,there are many researches on knowledge answering based on general knowledge graph,and the research results are fruitful.At present,some research thesiss on the construction of tourism knowledge map have appeared in the field of tourism,and even public tourism knowledge map has been available for reference.However,most of the thesiss and published tourism knowledge graph only consider the information of tourist attractions,without considering the possible consumption information such as food,accommodation and purchase in the process of tourism.At the same time,the published tourism knowledge graph has a small amount of data,and the effect of knowledge question and answer research is not good,unable to meet the needs of tourists for tourism.Therefore,how to construct a knowledge graph that can meet the tourism needs of tourists and serve tourists based on this knowledge graph has become a hot issue in tourism research.In view of the above problems,this thesis gives a solution.The main research contents are as follows:(1)Tourism knowledge graph construction.At present,most of the tourism knowledge graph only focuses on the information of scenic spots and ignores the needs of "food,lodging,travel,travel,shopping and entertainment" in the tourism field.In order to better meet the needs of tourists,this thesis expands the knowledge graph of scenic spots and adds information of restaurants,hotels,shopping centers and so on.At the same time,this thesis uses the method combining Bert model and Ratio Distance algorithm to integrate the tourism knowledge graph constiucted in this thesis with the existing route knowledge graph,and further expands the tourism knowledge graph by adding route information.(2)The study of question tourism entity recognition based on BILSTM+CRF+TERA.The BILSTM+CRF model is used in the research of named Entity Recognition.However,when this model is applied in the Tourism field,it cannot accurately identify some entities in the Tourism field.Based on the above problems,this thesis proposes the Tourism Entity Recognition Algorithm(TERA).The algorithm follows the idea of the maximum matching algorithm and identifies the tourism entities based on the fact that most of the tourism entities are nouns or are composed of nouns.The whole tourism entity identification study uses BILSTM to calculate the probability of each word being labeled as each label,uses CRF model to obtain the optimal label combination,and finally uses TERA to identify the unrecognized entity.The experiment shows that adding TERA on the basis of BILSTM+CRF model can better improve the accuracy of tourism entity identification.(3)In this thesis,Siamese LSTM is used to study the question attribute linkage based on the fused attentional mechanism.Siamese LSTM model was used to extract semantic features of candidate attribute sets stored in triples of questions and knowledge graph,and attention mechanism was introduced to highlight key information in sentences.Finally,similarity of semantic features of extracted questions and attributes was calculated,and the attribute with the highest similarity with questions in candidate attribute sets was selected.The experiment shows that the accuracy of attribute linking using Siamese LSTM combined with attention mechanism is higher than that using Siamese LSTM alone.To sum up,this thesis constructs a multi-source fusion tourism knowledge graph,and uses the method combining Bert model and Ratio Distance algorithm to fuse the route knowledge graph to obtain a complete small tourism knowledge graph,including the information of tourist routes,tourist attractions,restaurants,hotels and shopping centers.In this thesis,the model algorithm mentioned in(2)and(3)is used to study Q?A and finally the two methods are combined to realize the knowledge question and answer based on tourism knowledge graph.The experimental results show that the research method and model proposed in this thesis can improve the training efficiency method to a certain extent,and it is suiTab.for the research of Q?A in tourism field.