Research And Implementation On Construction Method Of Knowledge Graph In Tourism Domain

With the rapid development and wide popularization of computer and Internet technology,Internet has become one of the largest platform for people to acquire knowledge.How to extract useful information from mass data in Internet is one of the primary issues in the fields of information retrieval and data mining.In 2012,Google Incorporation proposed the concept of Knowledge Graph,which transformed traditional text searching in Internet into graph searching about entities,attributes,values,and their relationships.The research on Knowledge Graph construction has important application value for information management in Internet and knowledge acquisition.This paper focuses on the methods to construct knowledge graph in tourism domain,including the methods of attribute knowledge expansion and attribute value fusion,and designs and implements a Chinese knowledge graph in tourism domain bas ed on multiple data sources.The task of attribute knowledge expansion is to expand attributes and attribute values of entities.The task of attribute value fusion is to fuse attribute values extracted from different data sources in order to get the most reliable attribute values.For attribute knowledge expansion,this paper proposes and implements an approach based on words field,improves the method based on supervised learning,proposes and implements a hybrid attribute knowledge expansion technique.Within the method based on words field to expand attribute knowledge,the method in this paper increases the range of acquired knowledge,thorough giving a weight to each word related to attributes and values in words field and expending attribute knowledge by using key words and the search engine.Within the method based on supervised learning to expand attribute knowledge,this paper integrates multiple classifier and improves the classification accuracy.The advantage of hybrid attribute knowledge expansion method is that it extracts more and better <entity,attribute,value> triples.To the task of attribute value fusion,this paper proposes and implements a method based on learning to rank.It transforms the task of attribute value accuracy sorting into a traditional task of document sorting in the search engine,and filter effective attribute values using supervised learning.The corpus in this paper are selected from the online encyclopedia and the search engine.For attribute knowledge expansion,the experimental evaluation measures are Precision,Recall,F-Measure and Accuracy.For attribute value fusion,the experimental evaluation measures are MAP and NDCG.The results on experiments show the validity of these methods.Compared to each method alone,hybrid attribute knowledge expansion method increases the range of triples obtained,and has good accuracy.Finally,this paper constructs Chinese knowledge graph in tourism domain with obtained triples.Further,the Chinese knowledge graph constructed by this paper can be used in many areas such as question-answering systems,information retrieval,and be helpful to solve the interoperation problems of the Semantic Web.