This dataset consists of 812 verb pairs (v1,v2) which were manually annotated as entailing or non-entailing. 225 pairs were labeled as 1 (right-hand-side verb entails left-hand-side verb) and 587 pairs were labeled as 0 (right-hand-side verb does not entail left-hand-side verb). We generally followed the rule-based approach for entailment rule annotation, where a rule v1 –> v2 is considered as correct if the annotator could think of reasonable contexts under which the rule holds (Dekang and Pantel, 2001; Szpektor et al., 2004).
- Download the dataset
- References: Learning Verb Entailment from Diverse Linguistic Evidence. Hila Weisman, Jonathan Berant, Idan Szpektor and Ido Dagan. To appear in Proceedings of EMNLP 2012.
- Contact: Hila Weisman, weismah1 @ cs.biu.ac.il