UAIC Participation at RTE-7

This paper describes the fifth participation of the UAIC textual entailment engine at the RTE shared task. Our approach is rule based and makes use of the notion of predicational semantics for detecting entailment. The system used is based on the system built for the RTE-6 challenge, which we have further modified and improved for the current task. The system works by attempting to match every entity in the hypothesis to at least one entity in the text, but for this version of the systems, the matching process is predication driven, which is to say we used predicates in T and H as pivot points for determining matching entities. Matching is achieved by using extensive semantic knowledge from suck knowledge bases as DIRT, VerbNet, WordNet, VerbOcean, Wikipedia and the Acronym database.