AdArte

Introduction

AdArte (A Transformation-Driven Approach for Recognizing Textual Entailment) is based on modelling entailment relations as a classification problem where the single T-H pairs are first represented by a sequence of edit operations (i.e., deleting, replacing and inserting pieces of text) called transformations needed to transform T into H, and then used as features to feed up a supervised learning classifier to classify the pairs as positive or negative examples.