Linking named entity in a question with DBpedia knowledge base

Abstract

The emerging Linked Open Data provides an opportunity to answer the natural language question based on knowledge bases (KB). One challenge of the question answering (QA) problem is to link the entity mention in the question with the entity in the existing knowledge base. This study proposes an approach to link entity mention with a DBpedia entity. We propose an entity-centric indexing model to help search candidate entities in KB. After obtaining the candidate entities, we expand the context of the entity mention withWordNet and Concept- Net, we compute the context similarity between the expanded context and the property value of the candidate entity and the popularity of the candidate entity. Finally, we rerank the candidate entities by leveraging these features. Evaluations are performed on DBpedia version 2015, the evaluation tests show that our approach is promising in dealing with linking named entity in DBpedia.