In order to facilitate the development of an Example-based Rough Translation system, a method is proposed to retrieve such meaning-equivalent sentences from non-parallel corpora using lexical and grammatical features such as content words, modality and tense