SemKer: Syntactic/Semantic Kernels for Recognizing Textual Entailment.
Yashar Mehdad, Alessandro Moschitti, Fabio Massimo Zanzotto · 2009
Abstract. In this paper we describe the SemKer system participating to the fifth Recognizing of Textual Entailment (RTE5) challenge. The major novelty with respect to the systems with which we participated to the previous challenges is the use of semantic knowledge based on Wikipedia. More specifically, we used it to enrich the similarity measure between pairs of text and hypothesis (i.e. the tree kernel for text and hypothesis pairs), with a lexical similarity (i.e. the similarity between the leaves of the trees. The results show the benefit of this added semantic information. 1