Abstract

In this paper, we propose a semantic-based paraphrase identification approach. The core concept of this proposal is to identify paraphrases when sentences contain a set of named-entities and common words. The developed approach distinguishes the computation of the semantic similarity of named-entity tokens from the rest of the sentence text. More specifically, this is based on the integration of word semantic similarity derived from WordNet taxonomic relations, and named-entity semantic relatedness inferred from the crowd-sourced knowledge in Wikipedia database. Besides, we improve WordNet similarity measure by nominalizing verbs, adjectives and adverbs with the aid of Categorial Variation database (CatVar). The paraphrase identification system is then evaluated using two different datasets; namely, Microsoft Research Paraphrase Corpus (MSRPC) and TREC-9 Question Variants. Experimental results on the aforementioned datasets show that our system outperforms baselines in the paraphrase identification task.