A Lexicogrammar approach to checking quality: looking at one or two cases of comparative translation

In this paper I take a tried-and-tested methodology in linguistic analysis (the ‘‘lexicogrammar approach’’) and apply it to a particular problem of translation (a comparison of two equivalent phrases in an English translation of a French text). My purpose in doing this is to raise a number of research questions which I believe should be of importance to anyone in the translation business. My first question is very general: between two potentially equivalent translations, is it possible to identify which one is best? The assessment of any translation can often be highly subjective, but there appear to be some areas which are even more di‰cult to ascertain than others. In particular in this paper I examine the traditionally murky category of phraseology. However, I shall attempt to show here that it is possible to evaluate the phraseology of a particular translation in a scientific, almost forensic way, in particular by using corpus-based evidence. By ‘‘corpus-based’’, I am referring here to the use of computer-held electronic archives of texts, whether texts found on the internet or more specifically texts prepared for linguistic analysis by ‘‘tagging’’, or marking-up the corpus. In fact, it has now become the standard position of many empirical linguists (Sinclair 1991, Coulthard 1995, Hunston and Francis 2000, Tucker 2006) that no scientific statements about the linguistic features of a text can be based on introspection or gut-feeling alone, but should rather be supported by the meticulous observation and comparison of contextualised examples from a representative corpus of texts. To many, this might sound impractical and time-consuming, but the methodology of corpus linguistics has become fairly widely-accepted in the field of translation studies (Pearson 1996, Sinclair, Payne and Pérez Hernandez 1996, Bowker 1998, Xiao and Yue 2007). Furthermore, in this paper I show that it is feasible to conduct a systematic corpus-based analysis relatively quickly, especially if the 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40