A Translation Memory (TM) is a tool used during translation that basically saves all the words and sentences that have been translated as source and target segments.
A translation memory is an essential software tool in localization that stores segments of a text that have been previously translated for them to be leveraged or ‘re-used’ in another translation. Any new text is then compared with the strings that have been stored in the TM from previous translations. The results of the analysis are divided into four components:
- 100% match (perfect match): These segments are an identical match to a translation that has been stored in the translation memory.
- Fuzzy matches: These are segments which are similar but not identical to a previous translation, meaning that a translator may have to do a little work to edit the translation but not as much as if it were from scratch.
- Repetitions: These words/segments are repeated throughout the text, meaning that they only have to be translated once by the translator and the TM will replace the repeated text with the translation.
- No Match (new words): These are words which are completely new to the TM and therefore the translator will have to translate them from scratch. Each new translation is stored in the TM for future projects, meaning the more you work with a chosen vendor, the better value you will receive.
Since the 1990s, translation memory (TM) has become a core tool within the translation industry, aiding translators, allowing localization companies to maximize consistency for their clients and reducing costs to the client. While nowadays MT (Machine Translation) is also integrated as part of the whole localization process, TM still has a role and a place in the localization workflow.
Now, anyone who is not familiar with localization may have come across a lot of localization jargon during a typical localization project. Usually the first stumbling block to fully grasp this process is understanding the acronyms, job descriptions and tools we use in the localization industry.
What is CAT?
One of those acronyms is CAT, referring to a Computer Aided Translation Tool. CAT is a very broad term to describe software used to aid in the translation process. There are a wide range of tools that may be used under the term. Briefly speaking, the CAT team in a LSP is formed by localization engineers who manage the translation memories; analyzing the files, leveraging content, and making sure the files are ready for the translation team.
There are several CAT tools in the market, like Memsource, XTM, MemoQ, SDL Trados Studio, Smartling, to name a few. Each of them has their own translation memory standard, so it is useful to have a translation memory exchange file format that can be opened in any tool.
What is a TMX file?
This is where TMX (another acronym) comes in. TMX simply stands for Translation Memory Exchange. This is a text-based file created in an open XML standard containing translation units, formed by source and target segments and additional attributes that allow the identification and processing of each unit. This file format is what LSPs use for exchanging translation memories.
The screenshot below shows the content of a typical TMX file. The strings in the blue square form the translation unit, marked within <tu> tags. The attributes show that the translation memory was created by SimulTrans’ Engineering team on May 27th, 2020. The translation segments are marked within <tuv> tags, each with a four-letter code indicating the source and target language – in this example, English (United States) and Spanish (Mexico).
Should you migrate your translation memories?
Let’s imagine that you are currently working with a localization service provider (LSP) and are considering switching to another LSP. One of the most crucial steps to consider is whether you should bring your previous translation memory with you.
It might be worth it if:
- you have translated a significant amount of content
- content is still relevant to the product/service you sell
- translations are of good linguistic quality
- translations have been validated by your own internal stakeholders
- there are Brand Names, Trademarks, Logos or specialized terms already translated and approved
In those cases, using your translation memories will decrease your costs immediately, as the leverage will be high and you will benefit from consistency.
It might not be worth it if:
- you are getting bad quality translations from your current translation service provider
- the translations inside your TM belong to products that are obsolete
- the translations are mostly marketing oriented and to do with past campaigns
In those cases, you might not get a lot of useful leverage, and so it might not be worth to migrate it since you are considering switching translation vendors anyway.
Either way, I believe that you should get your own translation memories. After all, it is your company's intellectual property, and you should always have a copy of them, whether you use it or not.
Once you make up your mind, request the Translation Memories in TMX format from your current translation agency. There are other formats in the market, but this is the best suited for moving.
Whatever your reasons are, find out how easy it is to move…