Manual and automatic corpus compilation: A case study for legal translations

This paper compares two offline corpora for legal translations: the first one composed manually and the second automatically. In order to highlight the advantages of the one and the shortcomings of the other, if any, a translation project was carried out with two experienced translators. The participants translated two different supply contracts into their first language (Italian) by using the manually- and automatically-composed corpora. The manual corpus was qualitatively satisfactory and helpful, despite the lengthy building process. The automatic corpus was considered less accurate, albeit larger. The paper corroborates literature findings, as it is argued that legal corpora need not be large to be effective. It will also remark the usefulness of corpus analysis training before engaging in corpus-based translations.