Project of Paraphrase Identification Based on Weighted URAE, Unit Similarity and Context Correlation Feature