Recognizing Relation Expression between Named Entities based on Inherent and Context-dependent Features of Relational words

This paper proposes a supervised learning method to recognize expressions that show a relation between two named entities, e.g., person, location, or organization. The method uses two novel features, 1) whether the candidate words inherently express relations and 2) how the candidate words are influenced by the past relations of two entities. These features together with conventional syntactic and contextual features are organized as a tree structure and are fed into a boosting-based classification algorithm. Experimental results show that the proposed method outperforms conventional methods.