A Semantic Similarity Approach to Paraphrase Detection

This paper presents a novel approach to the problem of paraphrase identification. Although paraphrases often make use of synonymous or near synonymous terms, many previous approaches have either ignored or made limited use of information about similarities between word meanings. We present an algorithm for paraphrase identification which makes extensive use of word similarity information derived fromWordNet (Fellbaum, 1998). The approach is evaluated using the Microsoft Research Paraphrase Corpus (Dolan et al., 2004), a standard resource for this task, and found to outperform previously published methods.