Similarity Is Not Entailment — Jointly Learning Similarity Transformation for Textual Entailment

Predicting entailment between two given texts is an important task upon which the performance of numerous NLP tasks depend on such as question answering, text summarization, and information extraction. The degree to are similar has been used extensively a key feature in much previous work predicting entailment. However, using similarity scores directly, without proper transformations, results suboptimal performance. Given set lexical measures, we propose method that jointly learns both (a) non-linear transformation functions for those measures and, (b) optimal combination predict textual Our consistently outperforms baselines, reporting micro-averaged F-score 46.48 RTE- 7 benchmark dataset. proposed ranked 2-nd among 33 systems participated RTE-7, demonstrating its competitiveness over other approaches. Although our statistically comparable current state-of-the-art, require less external knowledge resources.