Home » Articole » EN » Business » Translation » Computer-assisted translation » Translation memory

Translation memory


A translation memory is a database of text segments, and these segments equivalent in another language. It can store and reuse segments of sentence. In this case it is about computer-assisted translation (CAT) rather than machine translation generally used to define the software that does not require the intervention of translators to translate the text, but which are much more approximate.

Advantages and disadvantages

The advantage is that it saves time in highly repetitive texts: the translator must make less typing and less research.

A translation memory is very effective for technical texts, often more than 90% of text is common between a product version and the previous one. It would not be the case for literary texts, which contain little repetition in the source text and where they would be undesirable in the target documents.

Most software using translation memory are not free and are developed for Windows only (Trados Workbench, DéjàVuX, SDLX, Star Transit, MultiTrans Similis MetaTexis), but there are a growing number of Java applications that can work both on Mac OS X and Linux (OmegaT, Open Language Tools Heartsome, etc.). Finally, some newer systems (Wordbee, XTM Cloud) are used from a web browser without any installation.

Some software are not suitable for work in group. Software designers consider TMs should not be used by more than one person. Systems for group work are generally more expensive. Trados Workbench, DéjàVuX, SDLX, Wordbee, Nobodies, for example, offer both versions.

In terms of quality of translation, translation memory systems, if they facilitate overall consistency can be problematic: the sentences are separated, the translator notes eg less repetitions or agreement errors. It is also more difficult, sometimes impossible, to rewrite a paragraph to fit the style of the target language.

Translation memory systems are increasingly imposed on freelancers by the agencies, which apply a sliding scale based on the percentage of correspondence between text and memory. The negotiation of the two parties is difficult and unreliable.

General operation

When he starts working, the translator is launching a program that segments the source text. Where in the database is an identical or similar source segment (the proximity threshold is determined by options), the program provides translation. The translator then has the option to accept it, adapt it, or reject it (and translate itself).

Whether, by choice or lack of correspondence, the translator must translate itself, the database records the translation for the next time.


The abbreviation is TM (Translation Memory) or TMM (Translation Memory Manager). The segments are called TU (Translation Unit). There is an interchange standard TM compatible (in theory) with a very large number of computer-assisted translation, this format is called TMX (Translation Memory eXchange). The translation software is called “Location”. Localization CAT tools are very useful, the repeats are very common.


OmegaT open source translation memory.

  • Multiplatform (requires Java).
  • License: GPL.
  • Developed by OmegaT Project, a group of independent translators and programmers.
  • Requires no intermediate format (XLIFF or PO) and allows direct translation of formats it supports.
  • Accepts standard TMX translation memory.

Open Language Tools free translation memory software.

  • Multiplatform (requires Java).
  • License: CDDL.
  • Originally developed by Sun Microsystems for his work location and made available as open source.
  • Works on the principle of converting the intermediate format XLIFF. The Converter is made available under the same terms.
  • Accepts standard TMX translation memory.

Heartsome Translation Suite Set of commercial software and free multiplatform software.

  • Requires Java.
  • 3 commercial software is XLFEdit, TMXEdit, TBXEdit.
  • XLFEdit is a publisher of XLIFF file and provides similar functionality to OLT. Conversion to XLIFF does not depend on external software and accepts a number of formats that are not accepted by the OLT Converter (particularly stylish Trados RTF). TMXEdit is a publisher of TMX and also offers alignment functions. TBXEdit is a glossary editor.
  • Accepts standard TMX translation memory.

Similis memory software translation business.

  • No platform (requires Windows).

Swordfish Software commercial translation memory.

  • Multiplatform (requires Java).
  • Swordfish is a publisher of XLIFF file.
  • The conversion does not depend XLIFF external software and accepts a number of formats.

Wordbee Translator is a commercial collaborative web space that includes both project management and translation support tools for translation (translation memories, terminology database, word count, etc.).

  • Wordbee accepts standard TMX and SRX.
  • No installation is required.

Wordfast is a software suite for business part (software Wordfast Wordfast Pro and Classic) and free (Wordfast Anywhere), centered around the translation memory.

  • All three tools have incidentally use machine translation.
  • Wordfast enables the exchange of data with the TMX standard.
  • No installation is required for Wordfast Anywhere software online and free. PRO Classic and settled on Windows, Mac and Linux.

Leave a Reply

Your email address will not be published. Required fields are marked *