Distant supervision for relation extraction using tree kernels
Azad Abad, Alessandro Moschitti · 2015
Abstract. In this paper we define a simple Relation Extraction system based on SVMs using tree kernels and employing a weakly supervised approach, known as Distant Supervision (DS). Our method uses the simple one-versus-all strategy to handle overlapping relations, i.e., de-fined on the same pair of entities. The DS data is defined over the New York Times corpus by means of Freebase as an external knowledge base, which indicates the relations of some of the entities of the NYT text. Our experiments show that our simple approach performs well in this domain with respect to the current state of the art.