SemAnTE is a semantic annotation project carried out on the Recognizing Textual Entailment (RTE) datasets. The annotation scheme used addresses three types of modification that license entailment patterns: restrictive, appositive and conjunctive. These inferential constructions were found to occur in 81.21% of the entailments in the RTE 1-4 corpora and were annotated with cross-annotator agreement of 68% on average.
The corpus contains 2,805 pairs of positive entailments, 2,278 of which were annotated for inferential semantic constructions.
The annotation work is described in:
Toledo et al. Semantic Annotation for Textual Entailment Recognition. In Proceedings of the 11th Mexican International Conference on Artificial Intelligence, Lecture Notes in Artificial Inteligence, Springer-Verlag, 2012.
Work on this corpus was done as part of the research program Between Logic and Common Sense at Utrecht University, supported by an NWO Vici grant.
Annotation SemAnTE 1.0: Stavroula Alexandropoulou and Heidi Klockmann
Research and programming: Sophia Katrenko and Assaf Toledo
Project leader: Yoad Winter
Collaboration: Natural Language Processing Lab (Bar Ilan University, Israel)
To obtain access to the annotated corpus, please fill in the form: