Personalized EntityRank-Based Entity Linking with DBpedia

Entity linking is the task of linking the textual mention in a document to the referent entity in the existing knowledge base. This step is important for many tasks, such as text understanding, semantic search, and question answering over knowledge base. This task is challenging due to name ambiguity and word polysemy. In this paper, we propose an approach to link the entity mention in the document to the DBpedia knowledge base entity. We measure the semantic similarity between a document and a candidate entity by comparing their personalized EntityRank vectors, and then score the candidate entities for each mention combined with other local features. Evaluation on three different datasets shows that our approach outperforms state-of-the-art methods.