Paraphrase and translation: the importance of being close.

Abstract

This article explores the concept of paraphrasing within computational linguistics, seeking to enrich its understanding by drawing parallels with translation studies and especially machine translation. It highlights the existence of two distinct yet related tasks: paraphrase generation and paraphrase detection, as well as points out the many (sometimes implicit) contact points in evaluation in both translation and paraphrasing. We claim that the concept of near-synonymy or near- equivalence is a shared concern of both disciplines, and its formalization should be pursued.