One way for me to combine my interests in translation and computers is to study translation memory software packages. A translation memory is in its simplest form a database where a translator may record (usually semi-automatically) old translations for future reuse and easy searches. Although these programs are best classified under computer-aided/assisted translation, one must not confuse them with machine translation programs - translation memory software does not translate anything by itself, whereas a machine translation system actively produces language and translations based on linguistic data, such as grammatical rules and glossaries. A translation memory system leaves all the actual translating to the human translator.
Basically, translation memory records sentence pairs: a source language segment (usually a sentence) is combined together with a target language segment. If an identical (or similar) source language segment comes up in another translation later, the translation memory program will find the previously translated segment and suggest it as the basis for the new translation. The suggestion may be either approved as it is, edited to match the current situation or rejected entirely. Most programs use a "fuzzy matching" algorithm, which makes the programs ever more useful; searches will find not only hundred per cent matches but also segments that only resemble the search phrase to a certain extent.
The advantages of using such software are not necessarily immediately discernible - at first it might seem like the software is more of a hindrance to productivity than an improvement, but as the memories build up, the advantages begin to show: matches found will be ever more accurate and numerous.
There are a great deal of similarities between the various software packages. Usually it is possible to search not only for full segments, but also single words or phrases, which certainly helps the translator to find terminology and keep the text consistent. Also almost all of these programs provide an interface to a terminology management program - usually their own. There are translation memory programs that work together with machine translation software as well. Some work directly within an existing word processor, others (most) have their own editor. All of these packages feature filters for importing and exporting various file formats. Some also provide an alignment tool for adding older translations (not done with the software) into the memory for reuse.
In the table below you can find links to the home pages of several
translation memory software packages that are currently on the market. Almost
all of them let you download a evaluation copy of the software, so if you are
interested, you may try these programs yourself :)
|Trans Suite 2000||Demo version (17 MB)||Cypresoft|
|Déj?Vu||Demo version (13 MB)||Atril site|
|Eurolang Optimizer||No downloadable demo||LANT site|
|IBM Translation Manager||Demo version (11 MB) + additional modules||IBM site|
|SDLX||Demo version (7,6 MB)||SDL site|
|Transit||No downloadable demo||Star site|
|Translator's Workbench||Demo version (7,6 MB)||Trados site|