Decision Making And Decision Support In Translation

One reason why computer-based support systems have so far contributed little to translation practice beyond ordinary text processing and database handling tools is that few computational linguists are aware of the linguistic issues which underlie the most cumbersome components in translation. A translator is frequently faced with genuine decision making; the frequency depends on the text type and the language pair, but typically this sub-task stands for the major eeort and time consumption in professional translation. Translators need decision support more urgently than aid to retrieve and operate on pre-established lexical and grammatical facts. The matter is discussed against the background of the ongoing large-scale schemes for translating the bulk of existing European Community legislation into the languages of the countries now candidating for membership in the Community. Tools have been or are being designed for the translators, revisors, legal experts and language consultants involved in these schemes. One is described in a project note presented to COLING 94 (Karlgren et al, 1994). \Distorsion" Desired In a discussion at the COLING Conference on Computational Linguistics, about schemes to evaluate experimentally diierent machine translation systems, Alan Melby warned against the \distorsion" introduced if participant system designers were permitted to see the source texts beforehand and amend their systems accordingly. He was right, of course: when claiming full automation, it is cheating to modify the instrument ad hoc. One could look at it also the other way: Since every text is unique anyhow, one of the most important features of a translation system in practice is how readily it can be thus \distorted".