Source text quality and computer-assisted translation

Abstract

This paper presents a project aimed at establishing whether, with the aid of CAT tools, it is possible to obtain a target text where the presence of two or more translators is not perceived by the reader. The project involved two Italian students, who were required to translate from Italian to English the tutorial of Instant Developer, a programme used to create Web applications. Before beginning the translation, a translation memory and a termbase were created. First, the texts of Microsoft documentation on Visual Studio 2006 were downloaded and aligned using SDL Trados, then the relevant terminology was extracted with Trados Multiterm. In this way, the students started working on the text with a common translation memory and termbase which were updated daily at the end of each translation session. Due to faulty expressions in the source text, the translation memory proved to be almost useless, as segments were hardly ever found. On the other hand, the termbase was frequently used to obtain and check terminology. However, this was not enough to produce a homogeneous target text, so that a cross-revision was required. The partial failure of this experiment was largely due to the faulty expressions in the source text, which greatly compromised the efficacy of the CAT tools. For this reason, it can be said that the effectiveness of CAT tools should be acknowledged but not overestimated as, even in specialized texts, there remains a wide margin of interpretation and rephrasing which requires the presence of a human translator.