Comparing non-native and translated language: monolingual comparable corpora with a twist

This paper illustrates a novel methodology for corpus-based investigations, which is being used in an ongoing research project looking into the salient features of two interfacing forms of mediated discourse, namely non-native and translated language. This work focuses on the language pair English-Italian within the framework of translation universals, and the novelty of our approach lies in the adoption of a monolingual comparable perspective which diverges from research paradigms traditionally used in corpus-based translation studies. These typically rely on corpora consisting of texts produced by native speakers on the one hand, and translations into the same language from a range of source languages on the other (cf. Baker, 1995:234). Our method differs from classic approaches in two main ways: (1) non-native written data are added to the equation (cf. Chesterman, 2007:63); and (2) the focus is restricted to a specific language pair, which is analysed in both directions. Our corpus thus includes four core components: translations from Italian into English done by native speakers of English (TR-ENG); texts written directly in English by native speakers of Italian (NN-ENG); translations from English into Italian done by native speakers of Italian (TR-ITA); and, finally, texts written directly in Italian by native speakers of English (NN-ITA). In addition, reference corpora are used as benchmarks for comparison with the non-mediated native norm. The paper describes the advantages offered by this composite structure and discusses corpus design, construction and annotation, focusing on comparability issues specific to the English vs. Italian monolingual sub-components. Preliminary results yielded by our analyses seem to indicate that some of the features observed in translated language, and usually explained in terms of translation universals, are in fact also present in non-native production. We therefore suggest that an extended notion of mediation universals might better capture the actual import of these shared phenomena.