Corpus-driven study of translation units in an English-Chinese parallel corpus

It is widely known that texts are not translated word by word but in larger units which are, from the perspective of the target language, more or less monosemous. This dissertation argues that translation units are the smallest such units, and that they can be identified in parallel corpora. It aims to show that these translation units and their target language equivalents can be extracted from parallel corpora and can be re-used to facilitate new translations. The concept of translation units and their equivalents will enable translators to translate competently into languages other than their native language, something not sufficiently supported by traditional bilingual dictionaries. For my exploratory study presented here, I will use the Hong Kong Legal Document Parallel Corpus (HKLDC). This dissertation starts with the definition of the concept of the translation unit and its equivalent and goes on to describe a method of extracting translation unit candidates. These candidates are then validated by further analysis. It will also test the hypothesis that each complete translation unit has only one translation equivalent. Finally, by comparing the translation equivalents extracted from the corpus with those provided by traditional dictionaries, this dissertation will argue that parallel corpora, as the repository of the translation units and translation equivalents, can, by complementing traditional translation aids, facilitate translation.