Translation memory and terminology management
Translation memory tools divide a text in the source language into ‘segments’ (which usually correspond roughly to sentences), and store the source segment together with the translated (target) segment as a translation unit. If a previously translated source segment is repeated in the current project (either as an exact or ‘fuzzy’ match), the CAT software suggests the corresponding stored target segment. This can be very useful with repetitive texts, which are usually technical rather than literary in nature.
Machine translation (as carried out by Google Translate and Microsoft Translator) uses masses of bilingual language data and sophisticated software (‘neural machine translation‘) to generate translations that are becoming increasingly useful, although there are issues of confidentiality that may prohibit their use in some areas (such as medical translation). Translation memory and machine translation work best with strings of words that incorporate some context.
However, there is still a place for a simple list of terms in one language together with their equivalent in one or more other languages – in other words a glossary or terminology database (termbase). Such termbases may be constructed by companies or governmental organisations to help manage their own translations and keep their terminology consistent (e.g. Inter-Active Terminology for Europe” (IATE).
But they may also be constructed by translators for storing previously translated terms for future reference. This can be done in the form of a simple spreadsheet, or in a more elaborate proprietary form such as SDL’s Multiterm termbase or ApSIC’s Xbench. Termbases in the form of spreadsheets can be constructed ‘by hand’, simply by entering source terms in one column and the target terms in the corresponding cells of the next column. This tedious process can be bypassed to some extent by automatic terminology extraction tools (such as Xbench, Synchroterm (Terminotix) or Multiterm Extract (SDL), although they all require some human assistance before extraction (e.g importing word lists) or after extraction (e.g. manual clean-up). Some are incorporated into rule-based machine translation tools, e.g. Systran Business Translator (15 languages) and PROMT Professional (5 languages) and are recommended by Uwe Muegge (Senior Tools Translation Strategist at CSOFT) as cost-effective terminology extraction tools for freelance translators and small agencies. Another excellent extraction tool is Similis, which is incorporated into a free translation memory system and can extract accurate bilingual glossaries from TMX files.
Some of the most widely used computer-assisted translation tools are SDL Trados, Wordfast, MemoQ and Across. An extensive list can be found at Atenao’s ‘List of computer-assisted translation software and translation tools‘.
These two searchable online databases of glossaries may be useful: