A Text Expansion Method Based On Entity Association Search

Knowledge graphs(KGs)have rich knowledge and can be extensively used to annotate and enrich text(e.g.news)in order to enhance its comprehension by readers.This requires to map entities occurring in the text to the target entities of the KG and to find so-called semantic associations(SAs)in the KG that connect the target and possibly other entities.In general,search of SAs is expensive and cannot scale to large KGs.Thus,it is common to restrict the search to structurally compact SAs but they often do not exist—they are too small to connect all the target entities.We solve this problem by computing the approximate association of the most salient,but not all,target entities.We first study the query relaxation problem in the context of SA search.We propose the concept of certificate,and prove that only two distance conditions are required to verify the success of the query.According to this,we design a best-first search algorithm based on distance estimation.Furthermore,we propose two heuristics to speed up the algorithm.Experimental results show the effectiveness and scalability of the algorithm.We also propose a context-aware ranking method based on the semantic cohesion and the relevance between SA and text.User experiments show that it is better than the state-of-art context-aware method.Finally,we design a system,which combines query relaxation,association search and association ranking together,in the scenery of news enrichment.