How much Romanian does Google Translate know? A corpus-informed genre-specific error analysis of English-into-Romanian translations

Abstract

To compensate, even if on a small scale, for the scarcity of investigations of English-into-Romanian machine translation from a corpus-based genre-specific perspective, this case study concerns the translation between these languages, by Google Translate, of texts belonging to two different, but closely-related genres – everyday and newspaper/ news releases language. In particular, it aims to offer a view of the translation errors in the Romanian target texts and, implicitly, of translation quality. To meet this aim, a translation error analysis is performed, starting from Keshavarz’s (1999) very general model of error analysis, and a linguistic error profile is created for each of the two genres taken into consideration. The errors identified are discussed and illustrated with small-scale corpus examples. Since translation errors, affecting translation quality, are the direct consequence of the capabilities of the Google translation platform, the findings of this paper may be relevant for the developers of this platform. They may get a clearer picture of its strengths and limitations and suggest ways of improving it so that it can ultimately provide higher quality translations when working with the English – Romanian pair of languages, with the particular text genres looked at here. It may also contribute to raising translators’ attention to the areas that are potentially problematic in these contexts, in the post-editing stage.