An Eccentric Approach for Paraphrase Detection Using Semantic Matching and Support Vector Machine

Paraphrase is the process of writing a sentence in another form. It relates to computing the similarity between sentences which are not lexicographically similar. This paper proposed an efficient method to estimate paraphrase. Specifically this paper defines various word co-occurrence in the sentence measured and its synonyms are also identified using web. By using the algorithm, similarity has been measured. This research focuses on evaluating the paraphrase between the sentences. In this paper, an approach for detecting paraphrase is tested by counting tokens. It processes token matching with the help of synonym of tokens, and produces the result with SVM classifier using Plagiarism detection corpus (PAN). Thus the SVM produces the output as +1 if the given text is paraphrased, - 1 if the given text is not paraphrased. This approach may also be used to identify Plagiarism of documents and to eliminate duplicates in a text repository.