Cross-Lingual Entity Linking And Semantic Query Processing Based On Knowledge Graphs

Recently, people pay much attention to the semantic information. And the techniques of semantic query processing have greatly improved. Knowledge graphs, as an important component of semantic query processing, contain quantities of named entities and seman-tic relations, and provide open knowledge access interface. Knowledge graphs help us determine the relationships between entities in the real world. Chinese knowledge graphs still need to be improved, since they are lack of substances and interrelationships. Com-pared with those English knowledge graphs, which contain rich relations of entities and semantics such as YAGO and Probase, Chinese knowledge graphs cannot support efficient semantic query processing. Aiming at using mature English knowledge graphs to acquire Chinese semantic information, we propose a cross-lingual query framework based on the cross-lingual graph, and based on the graph algorithm to fulfill the structural semantic information of cross-lingual query. Finally, we find a cross-lingual query system based on the frame, which provides the online cross-lingual query service. Besides, empirical studies over a real-life dataset have demonstrated the effectiveness of our methods.Main contributions of this paper are summarized as follows:· We propose the concept of cross-lingual graph and the cross-lingual query frame-work based on the cross-lingual graph. The frame solves Chinese entity mention query problem, which contains the missing cross-lingual link, and the rerank of en-tity disambiguation candidates online.· We propose a classify-rank model and a random walk algorithm based on the relation graph. These algorithm solve these vital problems, which are structural informa-tion improvement for the missing semantic relation in Wiki-Sketch, the mistakenly identifying link attributes, and missing links complement. Wiki-Sketch, which is extracted from Baike set, is a procedural and structured entity set.· We propose a missing attributes complement method by combining cross-lingual entity linking with cross-lingual attribute tag linking. The supplement of 347,124 missing attributes of 80,566 entities in Chinese wikipedia, and 45 million missing links of 600 thousand entities both in Chinese wikipedia and English wikipedia.· A system provides an online cross-lingual query service, and is able to accomplish almost-real-time client query request. The system based on thousands of entities, makes the query self-adaptable with the record of client’s query. Also the feedback mechanism in our system raises the precision of query based on the clients feedback of the query results.This paper studies the construction of cross-lingual query framework, optimizes struc-tural semantic information of cross-lingual graph, and the query efficiency of the cross-lingual query system. Our designed cross-lingual semantic query processing is usful in solving the problem of Chinese semantic information mining and completion.