Text typology and machine translation: an overview
Douglas Arnold · 1990
The purpose of these introductory remarks is to indicate the main issues relating to text typology that arise in the context of machine translation, from both the theoretical and practical points of view. There is no ‘typology’ or classification of texts to which MT practitioners or theoreticians can appeal, but it is clear that ‘text type’ is one of the critical parameters which can lead to success or failure in MT – for particular systems on particular occasions or in general, or perhaps ultimately even for the field as a whole. It is clear why this should be the case, as Juan Sager emphasised in a paper given at this conference some seven years ago (Sager 1982). Firstly, because appreciating the kind of text one is dealing with (having regard to subject field, style, register, text format conventions and purpose for which the translation is intended) is the first job of a human translator. Secondly, because not every type of text is equally suitable for every kind of MT. There are a variety of reasons for this. The intended purpose of a translation is important: is it to help readers, who are ignorant of the source language, or writers, who are ignorant of the target language? If it is for readers, what kind of information do they want to extract from the text? If it is for writers, what is the purpose of the translation? Is it for publication? Will it have any legal standing? One must be clear about the level of demand for translation for a particular kind of text, and the shortage or otherwise of human translators with respect to that demand. (Setting up and maintaining an MT system involves an enormous overhead, which will be hard to justify unless there is a sufficient volume of material to translate.)