The practical implications of a minimum machine translation unit
- Sint-Amandsberg Fédération Internationale des Traducteurs april-june 1997
- p. 138-150
- Volume 43, Issue 2, 1997 v. 43, n. 2 .
incl. ref.
A great deal of speculation dominates the translation industry with regard to the effectiveness of (MT) Machine Translation, or translation software. This project investigates the conclusions of Bennet (1994) about the size of the UT (unit of translation), based on the raw translations of a sample text as produced by four competitive PC programs. These programs are all transfer systems, which employ a minimum UT, such as a single noun phrase. The sample text is an authentic business correspondence text. A linguistic analysis of the four translations is performed. Results of the analysis show that numerous errors are committed which require the intervention of the professional translator. This research concludes that, for this type of text, a transfer system is not cost-effective because it will still require extensive human editing. The semantic errors particularly demonstrate the need to emphasize research towards the development of translation software which incorporates a larger UT.