FAUST RULE 4: You should maintain your translation memory!

October 9th, 2020 by Christian Faust - Posted in English

Translation Memories (TM) have been at the heart of all translation applications since the very beginning of Computer Assisted Translation (CAT). Basically, this is a database that stores sentences alongside their corresponding translations for later reuse. This saves time and money, because nothing needs to be translated twice.

Computers can be Rigid Thinkers

Even small deviations in the source text will mean that no exact match can be found within the TM, even if the basic meaning has remained the same. Deviations can be tiny things, like a missing comma or formatting changes, but also include missing or additional words, phrases or changed numbers, etc.

Luckily, CAT tools are great for detecting such deviations, searching the TM for the closest matching previous translations and displaying any deviations between the changed source text and the stored translation clearly for the editing translator. This makes the linguists work much more straightforward, allowing them to concentrate on the actual task of translating.

Language is Powerful

Language is the most powerful tool humans have at our disposal. At the same time, it’s also the finest tool imaginable and our greatest ever achievement. Spoken language is almost an artform in itself: the respective context, voice pitch, intonation, as well as, the raising or lowering of the voice can all produce completely different statements.

The written word is more simply structured. It lacks some of the features of oral articulation. However, even written expression is still so complex that automatic translations cannot deliver error-free results (at least, for the foreseeable future). When just one word more or less, any slight change in word order or a missing comma lead to different meanings, then wit, irony, sarcasm and allusions can understandably wreak havoc!

Functionality

Within translation technology the source text is divided by the program into sentences. Each sentence is referred to as a ‘segment’. A segment need not necessarily be a natural and complete sentence, but can be decided by anything that marks an end: a full stop,  a line break or paragraph end, a page break or defined fields, cells, etc.

Understandably, a translators’ job can be made difficult when individual  sentences are broken up into several segments. This can happen during manual pagination, conversions from unsuitable programs, OCR processes, and sometimes even data transfer from PDF files.

Since the formatting of the source text determines the segmentation, this seemingly small issue can lead to quite disastrous results. If sentences are divided into several segments during segmentation, these then cannot be translated 1:1 and the TM becomes useless. When the sentence is incomplete the TM can’t match segments with old translations!

Context: Matches From 100% to 103%

A TM can only reach its full potential when the degree of agreement is 100% and better. A higher percentage value takes into account the context that the program “recognizes” from the preceding and following segments. If the context is secured, the effort for the new translation is almost zero.

Editors often mean well when they optimize existing texts here and there during a revision. However, it must be borne in mind that in this case the translator will have to make an adjustment. This costs time and money.

So the question is: Does the small improvement represent a necessity? Does it help to make the text more understandable or generate sales? You can almost always do without pure cosmetics.

Use Text Blocks the Smart Way

Common editorial systems use text modules to build up operating manuals. This makes the creation of texts easier and avoids duplicate translations. This procedure can also be used in other environments: Standard texts should be kept in a single version.

It’s also advisable to use simple sentence structure and going for several short sentences instead of inordinately long clauses. Short sentences that refer to a specific subject can be reused much more easily in future versions!

🔴  Faust rule TM Top Tips:

❗️ Make Sure your TM Belongs to You

Don’t panic, you don’t need to maintain your TM yourself. This is difficult to do if there are several departments in your company that order translations. You should be able to rely on your translation service provider to manage this.

But in any case, be sure that the TM built from your company’s translations belongs to your company and not your language service provider (LSP). As your ™ leverages your old translations against each fresh translation, this means you only pay for each translated sentence once. If the LSP refuses to release the TM, it can make it unreasonably difficult for you to switch to another provider should you want to. Worst case scenario, if your TM is kept from you then may have to be rebuilt completely from scratch if you move providers and you could lose a large part of your existing translations. Never pay twice for the same translation, always check!

❗️ Stick with One Translation Service Provider

Translation memories will be of the most benefit to your business when each and every translation can be built upon and your TM can grow with previous translations that are uniquely relevant to your company and needs.This won’t work if you send translation jobs here and there willy nilly. It’s better to choose a single translation provider (i.e. us🤣) – at least for a specific language combination. Not forgetting to follow the previous tip at the same time!

❗️ TMs Collect Garbage

How can it be that useless translation segments accumulate in the TM? The honest answer: it’s hard to say.

Over the months and years, there are entries in every translation memory that ultimately make no sense. Whether it be poor segmentation, incorrect translations, technical developments, new product lines, improved processes or even changed legal requirements, many factors accumulate to cause some segments stored in the TM to become obsolete or simply unusable.

This can make it difficult to quickly and easily reuse stored translation units.

This is why, at FaustTranslations, we create a new translation memory for each customer approximately every two years. The old TM is then used exclusively for reading. Only new translations and checked entries from the old TM are stored in the new translation memory. In our experience, this is the best approach. This way, we ensure the TMs of our customers always up to date.

🔴   The TM goes hand in hand with the Termbase (TB). Check out our Faustregal 2: You should maintain your terminology! It’s worth using these two important elements intelligently, so costs and effort can be reduced considerably, while ensuring the highest quality translation results.

 Faust rule 4: “Maintain Your Translation Memory” to remind you of the importance of TMs as the heart of any translation environment.

Our important Faust rules are now available weekly on our blog! We don’t think much of newsletters and registrations. With us you decide which news and tips you want to read and when. Either drop by our blog from time to time or even easier:  Just follow us on LinkedIn!

 See You Next Time!

« »