Recognition of Paraphrasing Pairs

Our approach to recognize paraphrasing pairs is to make use of lexical relationships as representations of paraphrases. In this paper, we describe a classification process based on logistic regression as the inference mechanism. In this way, the set of features to be analyzed correspond to dependency trees from semantically equivalent paraphrase pairs. The results of the experimentation conducted show how a model constructed around lexical relationships is a plausible alternative for paraphrasing detection.