Combining Lexical, Syntactic, and Semantic Evidence for Textual Entailment Classification

This paper describes the Emory system for recognizing textual entailment as used for the RTE4 track at the TAC 2008 competition. We use a supervised machine learning approach to train a classifier over a variety of lexical, syntactic, and semantic metrics. We treat the output of each metric as a feature, and train a classifier on the provided data from the previous RTE tracks. As a result, our system is general, easily extensible, and naturally supports both two-way and threeway versions of the entailment task, as well as confidence estimation for the predictions. The results on both the training and the official data are promising, placing our system within the top 30% of all submissions.