How Do Translation Programs Work?
The fundamentalsBasically, translation programs work by matching up segments of (in our case) German text, usually a clause or sentence in length, with their English equivalents to form “translation units”.
These German translation units are then compared with units generated from previous German texts which you (or another agency translator) have stored in its translation memory, the actual TM. Where it finds matches, the TM recalls the parallel English text and presents it as a possible translation for the current text.
You’ll often hear the term “fuzzy matching” and this means that the TM will also suggest partial translations for translation units based on similar rather than exact past German texts. You can usually set your German to English translation software to defined degrees of fuzziness, e.g. 80 %.
Creating the databaseTranslation programs are clever tools, but their memories - the database of matching texts - don’t create themselves!
Indeed, the translator needs to go through each new translation they do, aligning the German and English text segments and then saving them in the terminology and phrase database. But for future related texts, you do enjoy significant time savings as it then automatically retrieves related segments from past German-English matches.
Many brands will also carry out automatic internet searches for terms, or draw on terminology databases such as Google, Systran, Ultralingua etc. which is a time saving in itself.
.TMX is the universal format for exchanging translation memory files between different German to English translation software, so when you’ve invested in one tool, you should still be able to “communicate” and transfer databases of terminology with others.
Obviously the software is only as good as the translations it uses as reference texts, and the regularity with which it is updated. I’ve received several TM files which include a variety of translations for a set term, i.e. the consistency which is so loudly touted is flagrantly ignored/not achieved.
Translation Programs v. CorporaYour software comes “empty”.
The parallel corpora (corpus = large and structured set of texts) that makes up its database - the actual translation memory itself - is either created by yourself, as you work through your regular German translations and line up all the matching German and English segments and save them, or you download a corpora which has been provided by your client or agency for a particular job.
Some parallel corpora are publicly available. For example, OPUS (open source parallel corpus) http://urd.let.rug.nl/tiedeman/OPUS/ is a collection of translated texts from the web. Many of the corpora it provides are issued by the European Union and European Parliament. This a great terminology source in general, but only relevant as a TM corpora if you are translating on behalf of these organisations.
When do I need translation programs?Language translation software is a boon for translators who work:
a. Frequently for a client in a particular subject area, and/or
b. On long and repetitive translations.
Agencies use translation programs:
a. To ensure consistency where more than one translator works on texts for a client.
b. To ensure consistency over time for regular translations of similar texts (e.g. a large corporation’s annual financial statements).
Drawbacks?Leading brands of translation program can be relatively expensive (500 to 1,000 Euros) and once you have one, agencies that demand their freelance translators use a translation memory may even expect to pay less for a translation – after all, they calculate that you won’t need to freshly translate every word or sentence.
As the translator has to buy the program, pay to attend seminars on how to use it, and still manage the translation, the question of whether you really come out on top when using a more expensive TM can be a bit iffy unless your volumes of translation warrant it.
However, if you haven’t got a TM, it doesn’t mean you can’t use one: agencies often send me TM files which can be used as standard glossaries. If you don’t have the corresponding TM system, you can’t use its full functionality, but you can use it as a glossary (just use the Search tool).
The future of computer-aided translation?We are beginning to see the advent of free, browser based, TM systems (e.g. Wordfast Anywhere) and the popularity of open source translation software (e.g. OmegaT) is growing. The sheer number of different TM systems available makes compatibility more important – and this should, in turn, reduce the hold that any one brand may have over the market.
Which translation program is the best?Here is a list of the most popular TMs on the market and some background information on each. Some are even free. (Hurrah!)