Recognizing Textual Entailment Using Lexical Similarity

We describe our participation in the PASCAL-2005 Recognizing Textual Entailment Challenge. Our method is based on calculating ?directed? sentence similarity: checking the directed ?semantic? word overlap between the text and the hypothesis. We use frequency-based term weighting in combination with two different lexical similarity measures. Our best run shows 0.55 accuracy on the test data, although the difference between our two runs is not significant. We found remarkably different optimal threshold values for the development and test data. We argue that, in addition to accuracy, precision and recall are valuable measures to consider for textual entailment