Study On The Construction And Application Of Tourism Knowledge Map In Xinjiang

With the improvement of people's living standards,tourism has become an energy way for people to broaden their horizons and relax their body and mind.Furthermore,people's demand for tourism services is increasing.They are no longer satisfied with beautiful scenery,but also require a better sense of tourism experience,which requires improving the quality of tourism supporting facilities.Such as accommodation,food,Shop etc.The scenic spot data,hotel data and passenger information of traditional tourism websites need separate databases and data tables to maintain,and it takes a lot of time and manpower to integrate these data.The knowledge map can easily realize the joint analysis of complex data.Xinjiang is located in the hinterland of Europe and Asia,with abundant tourism resources.Every year,a large number of tourists come to Xinjiang every year,which gives the development of Xinjiang's tourism industry a far-reaching potential and broad prospects.At present,there is no public tourism knowledge map that is only aimed at Xinjiang and includes food and housing.The existing products based on tourism knowledge map are basically assisted by knowledge map,with the main purpose of realizing recommendation function.In view of the above problems,this paper has done the following work:(1)Acquisition of unstructured and semi-structured data: Since there is no public data set needed to construct Xinjiang knowledge map at present,this paper uses crawler tools to crawl unstructured and semi-structured data from multiple websites.Unstructured data are mainly 400,000 sentences obtained from Xinjiang travel diaries crawled from websites and processed by duplicate removal,noise removal and clause removal.Among them,4000 pieces of data are manually marked and used as test data sets for named entity recognition experiments.Semi-structured data mainly includes2,895 pieces of information related to tourist attractions,1,405 pieces of information related to hotels and 15,146 pieces of information related to gourmet shops.(2)Attention mechanism named entity recognition model based on multi-source information fusion: Aiming at the problem that manual labeling of training corpus requires a lot of time and manpower,this dissertation introduces natural language processing tools into the Bi LSTM-Attention-CRF model,and proposes an attention mechanism named entity recognition model based on multi-source information fusion.Under the premise of no manual annotation,the existing natural language processing tools are used to annotate the corpus,and the semantic alignment of the annotation results of various tools can reduce the inaccuracy of annotation.(3)The construction of Xinjiang tourism knowledge map covering food and housing: the data with different structured degrees are transformed into triple form by different methods,and on this basis,the data are fused to obtain high-quality knowledge map.In order to display the data of Xinjiang tourism knowledge map more intuitively,this dissertation uses Neo4 J to visually display the triple.There are 39,067 entities and104,790 relationships/attributes stored in Xinjiang tourism knowledge map.(4)Constructing Xinjiang tourism knowledge map question answering system based on knowledge map: The main tasks in the construction of the question answering system include question comprehension,question definition and answer generation.In order to accurately identify the entities in questions,this paper defines five types of entity dictionaries and a hybrid dictionary according to the entities in the knowledge map.In addition to defining 21 kinds of problem types for entity attributes,this dissertation also adds some recommended problem types and maximum problem types.Through the functional test,the question answering system in this paper can better answer the common questions in the field of tourism.