Open Relation Mapping Based on Instances and Semantics Expansion.

Mining the semantics of open relations is an important task in open information extraction (Open IE). For this task, the difficulty is that the expressions of a specific semantic in free texts are always not unique. Therefore, it needs us to deeply capture the semantics behind the various expressions. In this paper, we propose an open relation mapping method combining the instances and semantic expansion, which maps the open relation mentions in free texts to the attribute name in knowledge base to find the real semanics of each open relation mentions. Our method effectively mines semantic expansion beyond the text surface of relation mentions. Experimental results show that our method can achieve 74.4% average accuracy for open relation mapping. © 2013 Springer-Verlag.