Paraphrase Identification Using Textual Entailment Recognition

Paraphrase Identification is a challenging and popular research area in Natural Language Processing Applications, including document summarization, information retrieval, question answering etc. In this paper, we focus on paraphrase identification using textual entailment recognition. In the proposed method Paraphrasing is seen as a bidirectional textual entailment. The predicate argument structures are used to represent the text and distant supervision technique for mining the indirect facts from these texts. The proposed method uses a probabilistic network for judging the confidence of each fact and a word sense disambiguation module is used to check the similarity between two sentences.