Similarity between Pairs of Co-indexed Trees for Textual Entailment Recognition