Using Coreference and Surrounding Contexts for Entity Linking

Ambiguity in meanings of words or phrases in a document is considered one of the most primary barriers in natural language processing. In this work, we address the task of identifying and linking mentions of entities into correct referents described in a given knowledge base. To deal with it, we propose a supervised learning method for ranking candidate entities in combination with exploiting a heuristic and coreference relations among mentions in a document. Furthermore, another advantage of our method is its simplicity and effectiveness with using much fewer features than other systems. The results from evaluation on TAC-KBP 2012 datasets show that our combination is efficient and this method has a comparable performance to the state-of-the-art ones.