Translation Memory Alignment (TMA) is a method by matching same text (sentences, phrases, words) using alignment tools. It creates a database of text segments in two different languages. TMA allows for clients to provide two files (source and target languages) representing the same text and align the sentences matching them to generate a translation memory (TM). These aligned bilingual memories allow clients to reuse identical and similar text segments that have previously been translated, to aid human translators in translation projects. A TM is invaluable when translating and localizing great volumes of content. A term or sentence once translated doesn’t need to be translated again. It is automatically substituted in a new document. The translation process using TM preserves the linguistic consistency of corporate documents across platforms and versions. This proves to be invaluable in translating content that includes many repetitions like technical specifications, financial statements, and legal terminology. With TM as a base, translation of documents goes much faster. Instead of starting from scratch with translation, TMA allows you to build a repository of translation segments that are saved and can be utilized immediately.

Here are some reasons why clients order TMA services:

-Translations were done manually not using a TM system.

-Translations were done by a translator or an LSP who did not deliver a TM.

-Translations were done by a translator or an LSP but the quality was bad.

-Translations were done but final changes were made during the DTP process. There was no TM for the finalized version.

Most CAT tools offer their own alignment feature, but there are other specialized tools in the market. The TMA process is completed using an alignment tool.

What is the TMA process?

The TMA process begins with two files – one with the source text, the other with the corresponding target text in a different language. To create a leverageable TM, you first need to pool as many of your original files and translated equivalents as possible. Then, these 5 steps are applied to create your TM resource:

1. Segment Extraction

All text segments (basically sentences) are extracted from the source and target files to create a bilingual database with original text and the corresponding translated text.

2. Segment Alignment

All segments are aligned using software solutions to pair segments in the source and target files based on their placement, content, etc. The process is highly automated and enables very fast processing, much faster than humans and with high precision.

3. Human Editing

A native linguist reviews the results of the paired segments to ensure they really match. Confirmation or corrections are made at this step.

4. TM Creation

Lastly, redundant segments that have no matches in neither source or target are deleted, and then the bilingual text segments are exported to .tmx format or any other format you may require (e.g. .xliff, .csv, etc.).

When your translated material is not available in TM format, typically TMA is done to create files like .xliff and .tmx - both of which are .xml files amongst the most popular in the industry.

With TMA it is possible to create a reliable TM. As the linguist works on a new translation using this TM, it is standard to apply penalties to any TM matches. A certain percentage is deducted automatically, for example if the penalty is set at 1%, any 100% matches become 99% matches. This information is provided as a reference to the linguist as "aligned". These 99% matches are rechecked to confirm the content and if necessary corrected during the translation process to ensure quality results.

How can you improve TMA results?

If you have the proper preparations, you will be able to improve your alignment accuracy following these points below:

-Make sure source and target files have the same file format.

For example, a PDF file and a Word file have different information imbedded within the files. This difference makes it difficult to match segments automatically. Source and target file formats should match to ensure the alignment process has the best accuracy.

-Make sure the source and target files are the same version.

A source file is often updated including extra information or has deleted text because it was redundant, after the previous translation was done. The content of the source and target file should match for maximum efficiency. If both files do not match completely, the alignment process becomes more complex. Also, layout differences contribute to poor alignment results.

-Make sure your TMA project is performed by a native linguist.

To increase accuracy, the linguist can check each segment to approve the automatic matches or fix incorrect matches creating a reliable TM. Including a native linguist in TMA helps to ensure the alignment is 100% correct. This careful step is especially helpful when dealing with documents that misaligned during the TMA process.

-Make sure to use a TMA tool that generates a quality report.

A TMA tool with internal algorithms can indicate how accurate an alignment was. It can measure automatically how many of the segments have matched accurately. This quality report helps identify broken segments, orphan segments, and unnecessary segments for the linguist to fix or delete as necessary.

Not all TMA are created equal.

Why use TMA?

TMA helps build a repository of bilingual segments which are then saved as a Translation Memory (TM). TM can provide consistency in translation, lower cost, increase productivity, and preserve the previous translation of your documents. This TM can be saved and utilized whenever necessary for your translation projects. There are many benefits to have a TM for your translation projects. idioma® offers TMA in 70+ languages. We can provide TMA and translation services as a package with our professional native linguists. Starting from scratch is OK but if you have previously translated documents, you can leverage them to create amazing results for your future translation projects.

Need help with TMA? Need a translation quote?

If you have any questions, please leave us a message or contact us at info@idioma.com.