Event-centric Tourism Knowledge Graph

With the development of artificial intelligence in recent years,knowledge graph has gradually become an important research topic in industry and academia.The concept of knowledge map was first proposed to improve the performance of search engine,but because knowledge graphs contain highly structured and ing knowledge,their application space has expanded from the earliest search engine to question answering system,recommendation system,knowledge reasoning and many other fields,which is a very promising research direction.Large scale knowledge graphs such as wikidate[1],DBpedia[2]and Yago[3]are mostly constructed based on structured data.With the development of natural language processing technology,extracting information from unstructured text data has become an important method to construct knowledge graph.Experts in finance,law,medical and other fields also began to use their prior knowledge to construct interdisciplinary knowledge graph.However,the traditional knowledge graph is entity centered,which describes the static knowledge in the world.As another element of describing the real world,events are often ignored.Event evolutionary graph[4]is a new concept in the field of natural language processing in recent years.Compared with the traditional knowledge graph,the event evolutionary graph takes events as the center,focuses on describing the dynamic relationship between events,and is used to explore the law of events.In the field of tourism,the traditional knowledge graph also models the static knowledge of tourists during the trip,ignoring the events and spatial-temporal relation in the travel.Therefore,this paper puts forward the concept of event-centric tourism knowledge graph,and constructs related applications based on the knowledge graph.The main innovations are as follows:(1)This paper analyzes the shortcomings of existing traditional tourism knowledge graph,and puts forward a new concept of event-centric tourism knowledge graph.Based on the Simple Event Model[5],a new tourism knowledge graph schema is designed.At the same time,the overall framework of automatic construction of event-centric tourism knowledge graph is proposed,and the behavior of tourists is analyzed based on the knowledge graph.(2)Based on the new event-centric tourism knowledge graph,graph convolution network and graph attention network are improved,and the ETKGCN scenic spot recommendation framework is proposed,which is superior to the traditional recommendation methods in real world data sets of Hainan province.(3)Based on the event-centric tourism knowledge graph,this paper designs an seq2seq tourism route recommendation framework,which makes full use of the spatio-temporal information in the knowledge graph and the method of natural language generation,and achieves good results on real world data sets of Hainan province.