Translation Memory Solutions

and Management, LLC


The Translation Memory Program

Understanding a Translation Memory Program




Are you a translation memory program user?


Do you understand how it works?


What is a translation memory program?

A translation memory program is a computer program that has the ability to create a database by binding two sentences, in two different languages, together. The program, instead of working with single words, divides the document by sentences, called segments. The first sentence is in the source language, and the second sentence is in the target language. The target sentence is the translation of the source sentence in a given pair of languages.

All translation memory programs, regardless of the brand, work according to the same principles.



A translation memory is NOT a translation dictionary

The translation memory database is created from scratch by the translator, and the content and accuracy of the memory are entirely the responsibility of the translator. It greatly improves the quality and accuracy of the memory if the translator works with an editor.

After finishing the first translation job, the program keeps in the database all the segments and corresponding translations for a given pair of languages. In subsequent jobs for the same pair of languages, the program has the ability to compare the text in the new document, segment by segment, with the source segments already in the database, and show those segments and corresponding translations if it finds any matches. Matches can be 100% (exactly the same source segment and therefore the same translation) or a percentage of the sentence matches (that means, part of the sentence is the same – the minimum value most often used is 70%, but this number can be changed at the translator’s discretion).

Understandably, as more translations are done with a specific memory, the more the program will “learn” and the bigger the database will be.

The advantage of the translation memory is to speed up the translation job by bringing up sentences or parts of sentences that have been previously translated. Therefore, the translator only can take real advantage of the translation memory in two situations:

1. If the subject matter and the sentences of the new document are similar to a previously translated document, or

2. If the same sentence repeats in the document; the translator only needs to translate the sentence once, and the translation memory inserts the same translation for the repetitions.

IMPORTANT: If a sentence has the same words compared with a previous translation, but they are scrambled, the program will NOT recognize it as having been previously translated. In order to be recognized as a whole or partial previous translation, the words need to be in EXACTLY the same order. Some translation memory programs do recognize when there are all the same words in a different order, but this often falls below the aforementioned “match percentage” threshold.



Types of documents that can be used

Often, text that needs to be translated is written in a Microsoft Word® document. A translation memory program can use this document without any conversion.

But, just as often, the text that needs to be translated is from a desktop publishing program such as Quark Xpress®, Adobe InDesign®, Adobe PageMaker®, CorelDRAW®, etc., or from websites or other types of documents and programs. A translation memory CANNOT directly handle the text inside of those documents.

A few years ago people used to copy the text from those documents, paste it in a Microsoft Word® document, send the Microsoft Word® document for translation and then paste it back in the original document for the final translated version.



Exporting text from different Desktop Publishing

Computer programs for translation

Today, however, the text can be exported with the use of filters into a text document format usable by the translation memory program, and after the translation has been done, it can be imported back into the original document with the translated text.

The advantages of using these filters are fast exporting of the text and fast importing of the translated text, requiring only slight retouches in the style after import; thus, the document is ready for delivery much faster.



The Translation Memory Program and the Translator

Another advantage of translation memory programs is that, the translator does NOT need to own several programs (like Quark Xpress®, Adobe InDesign®, Adobe PageMaker®, CorelDRAW®, etc.) in order to be able to perform his/her job. With Microsoft Office® (Word®, Excel® and PowerPoint®) and the SDL Trados® translation memory, the translator can accomplish all types of written translation requests.

NOTE: Most of the previous considerations apply only to Trados 2007. The new Trados Studio 2009 handles some of the tasks in a different way.



Handling Different File Formats

Trados Translator's Workbench is the part of Trados 2007 that handles translation memory files and maintenance.

Trados Translation Memory can directly use only:

Microsoft Word files (can be used directly with the Microsoft Word program and Trados Workbench®, without the need for conversion).

With Trados Tag Editor combined with Trados Workbench, the following file formats can be used:

XML files – HTML files – Xpress tags (Quark Xpress file export) – .inx (Indesign Interchange Format) – .XLFF – Microsoft Excel® files – Microsoft PowerPoint® files.



Spell Checker and Tag Checker

If you have a spell checker installed in Microsoft Word®, Trados Tag Editor® can be set up to use it and spell check as you progress in your work, and also to check if you are missing any tags in a given segment. The spell checking and tag checking are done when you close the segment.




WinAlign is part of the Trados software. It allows you to create memory translation databases by pairing previously translated texts from 2 different files, one being the source language file and the other the target language file.




MultiTerm is a terminology database that allows for systematic management of approved terms in both source and target languages.



Translation Memory Management

The great convenience of keeping memory translations updated and edited is that when they are used, new translations will have fewer edits, it will be less time consuming to fix and it will thus be cheaper.

To perform the memory translation updates, you can use Filtering and Find and replace.



Sending a Memory to a Translator

A translation memory is composed of 5 different files. If you need to send a translation memory to an outside translator, it is not very convenient to send 5 files of unwieldy size. It is much more practical to export the memory to a text file for sending to the translator. The Trados Workbench can export the memory with different formats. The format that seems to cause the fewest database corruption problems is TMX 1.4.

Note: With the Corporate version of the program, the memory only can be exported if it is opened exclusively.



Analyzing the Original File

The translation memory also has a tool that allows you to make an analysis of the file before it has been translated, comparing it with the database in the memory, and giving you a report on how many words need to be translated and how many are already translated and in memory.



Cleaning the File

During the translation process, a bilingual file is produced (the file has segments in the source language and segments in the target language). At the end of the translation, the file needs to be cleaned, which means removing the original language and leaving only the target language.


SDL TRADOS® is a registered trademark of SDL

Trados Tag Editor® is a registered trademark of SDL

Quark Xpress® is a registered trademark of QUARK

Adobe InDesign® is a registered trademark of ADOBE

Adobe PageMaker® is a registered trademark of ADOBE

CorelDRAW® is a registered trademark of COREL

Microsoft Word® is a registered trademark of MICROSOFT

Microsoft Office® is a registered trademark of MICROSOFT

Microsoft Excel® is a registered trademark of MICROSOFT

Microsoft PowerPoint® is a registered trademark of MICROSOFT