Samuel Murray, Translator - English into Afrikaans | Article index
This short article was written for a special Muratho issue on CAT/MAT. The English translation was done by the Muratho's editor.
The idea of a translator not using computer-assisted methods of translation is almost unthinkable. The advantage that the computer gives the modern translator over his pen-and-paper predecessor lies in the translator having to spend less and less time on repetitive little tasks and having more time available for the actual hard creative work of translation.
Some of these utilities are so widely used that translators often do not even realise that they are doing "computer-assisted" translation. Take a spell-checker, for example - although not infallible, it makes the translator's task a hundred times easier. Or think of a CD-ROM dictionary - thanks to this technology, a translator can source definitions, translations, synonyms and encyclopaedic information in seconds.
The variety of programmes that make computer-assisted (CAT) or machine-assisted (MAT) translation possible is staggering. Since some utilities are bundled as CAT program packages, even translators who regularly use CAT programs are not always aware of all the possibilities. We will consider some of these utilities in the second half of this article. But let us first look at what the CAT program packages usually comprise.
The great attraction of modern CAT programs like Trados, Déjà Vu, TransSuite, Transit and SDLX is translation memory (TM). Translation memory is nothing more than a fancy word for an extremely complicated word list. This word list is stored in a special format and read in a special way by the program so that large chunks of the translation are so to say "automated".
The CAT program uses the TM to suggest translations for phrases in the source text automatically. If the TM contains the entry "rekenaar = computer", for example, the CAT program will suggest the translation "computer" for every occurrence of "rekenaar". If the translator is satisfied that this is the right word, she can accept the translation simply by pressing a key. A clever CAT program will even identify "rekenaars" and "rekenarisering" as near equivalents and suggest translations for them (based on the entry "rekenaar = computer").
Since the TM also contains phrases and full sentences, CAT programs can recognise equivalent or near-equivalent phrases in the source text and automatically suggest previous translations from the TM for these. Just think of how much time can be saved if repetitive sentences (or sentences that correspond to a large extent) are translated automatically in this manner. TMs consist of already translated phrases rather than single words, and for this reason TMs cannot be bought in their final form - no, each translator has to build her own TM up from scratch. For this reason CAT packages usually contain alignment programs, which are able to segment texts that have already been translated and put them together to form a comprehensive translation memory.
In segmentation, the CAT program breaks the text up into sentences or phrases, according to previously set rules. The translator then translates these segments one by one, and when the CAT program comes across a segment that corresponds to a certain percentage with a segment in the TM, the translator can choose to import that translation and to alter it, if necessary. In segmentation, an existing source text and an already translated target text are both segmented and arranged in pairs to form a TM. Since such automatic alignment is rarely likely to be 100% accurate, the translator must read the half-completed TM through quickly and make corrections where necessary. When the translator is satisfied that every target text phrase is aligned with its corresponding source text phrase, the TM is ready for use.
In addition to segmentation, alignment and TM, most large CAT packages can also create word lists and compile concordances.
For a newcomer wanting to acquire a CAT program, the choice is overwhelming and it is difficult to make a choice. With prices ranging from €500 to €900, this is quite an investment. The best-known CAT package is probably Trados, with Déjà Vu close behind.
The bias between MS Word and WordPerfect users, or that between Internet Explorer and Netscape Navigator users, is fairly mild in comparison with the loyalty that Trados or Déjà Vu users exhibit regarding their favourite software. There is thus almost no point in asking a Trados or Déjà Vu user which is the better program.
Trados is currently largely the industry standard, but operates only in MS Word. Déjà Vu apparently has extra functions that Trados does not, and operates in conjunction with any existing word-processing program. Most Trados files can be imported into Déjà Vu, and vice versa.
Not all CAT programs are expensive. The localisation program plaaslikmaker ForeignDesk is openware software; in other words it is free. WordFisher costs only €30 and performs limited pre-translation and TM functions. WordFast costs €170, and handles segmentation, alignment, TM and close-equivalence with ease, as well as dictionary, concordance and context searches. However, both WordFisher and WordFast work only in MS Word.
Spell-checkers: Both Potchefstroom University and Pharos currently offer spell-checkers for MS Word, at around R250. There are, however, also free spell-checkers, like Ispell, that work in any word-processor (even Notepad), as long as one can find a reliable word list.
Find programs: Windows has its own Find function through which any document can be searched for text strings. More powerful find programs are also available, like SeekEasy, which can search compressed files (like zip files) or can locate near-equivalents or in which multiple search strings can be specified. Find programs usually also have search-and-replace functions.
Clipboard memory: When one copies and pastes (Ctrl-C and Ctrl-V), the text is copied to the clipboard. The Windows clipboard can remember only one piece of text at a time, though. A program like Clipomatic (which is free) remembers up to 64 pieces of text, and often-recurring pieces of text can be permanently stored in its memory.
HTML translators: Using a program like CatsCradle (also free) one can translate web pages without any knowledge of HTML. The program automatically extracts the text from the web page for translation and then reconstitutes the translated product.
Term searchers: There are programs that search word lists automatically on line. Examples are Glossy and Xanadu, which both work only with Internet Explorer and are also of any use only while one is linked to the Internet.
Excerptors: A program like Terminology Extractor (which will leave you €100 the poorer) converts any piece of text to a useful corpus in a flash, by arranging words and phrases that recur often in concordance format. Less expensive programs like the WordSmith suite have similar functions, and free programs like WClip enable people who create word lists by hand to eliminate a few laborious steps.
Word counters: Every now and then one comes across a PC on which MS Word or another advanced word-processor is not installed. In such cases simple web-page script can give an approximate word count. Purpose-made word counts (for example only words of three or more letters) can be done by special programs.
Comparers: Although MS Word has a built-in text comparer, the same function is fulfilled by the free program ExamDiff. The program compares two documents and indicates where words have been added or deleted.
Lemma trees: Once can store articles in a tree structure using lemma tree programs like Treepad or MyInfo. Articles (pieces of text) are stored in a tree structure with multiple nodes (lemmas). It is basically a program that as it were ties separate files together into a unit for easy reference, e.g. pieces of legislation or a collection of newspaper articles.
Aligners: TransSuite's aligner can be downloaded free of charge. This can be used to TMs of existing translations, which one can then use like a normal word list, even if one does not have a CAT program.