Our latest news & views
Competitive translation services today are using translation memory systems (TMS) to give their clients faster, higher quality translations. They certainly aren’t risking their clients’ reputations by relying on machine translation. The use of TMS is particularly useful when it comes to specialist materials such as for technical marketing, legal or financial services documents, for example. But what is translation memory and how does it work?
Let’s say you’re preparing to relaunch an updated product or service into Europe or Asia. It’s got all the old favourites from your previous version, with a few more add-ons that existing and new customers will love. Chances are, 80 per cent of the new version contains the same features and benefits as the current one, so you only need to update 20 per cent of your marketing materials.
Using translation memory, your translation team can recycle the existing, pre-approved translations and focus on the new content. So you only pay for the new stuff. Magic.
And it’s equally applicable for new projects – more on that later.
Machine translation uses a machine-based algorithm that translates automatically from one language into another. The results can sometimes be usable. The problem, however, is that they can sometimes be shocking. Ultimately, that’s the problem that makes machine translation unworkable if you want to avoid damaging your brand. For serious brands, it’s never OK that a machine-translated marketing asset goes out to thousands of contacts across the world, using awkward, stilted or just-about-understandable language. It’s never going to entice customers and could end up causing existing customers to disengage – especially if a competitor comes along and nails it.
Like the elephant, translation memory never forgets. It also makes new memories. It learns, stores and suggests translations based on previous translation activity – all in real time.
No matter who is translating it, where or even when, the preferred translation for certain strings, words and phrases are already there. For example, sometimes multiple translators will work on the same project – sometimes even years apart. With translation memory none of that matters. This saves our translators time and keeps our clients’ materials consistent.
We’re big fans of translation memory. As are our clients, who get a faster, better translation service at a price they’re happy to pay.
But great quality translations don’t stop at translation memory. Translation memory facilitates stage one, in conjunction with your glossary or style guide (more on that below). Then follows localisation and creative translation (“transcreation”) before anything leaves our word palace.
Before we look at how to get started on translation memory, let’s quickly see how a translation memory system works alongside a glossary or style guide.
The glossary is a key component of every marketer’s style guide. It gives us specific terminology, or lexicon, for a translation project. This is a different function to translation memory, which consistently translates and updates repetitive content. At the end of every project, the translation memory and glossary are updated and reconciled, giving the next job a head start.
When your project is ready for translation, your project manager will analyse it with translation memory software, identifying the previous translations. If you’re new to our services, we’ll either work with your existing translation memory files (TMX) or – if you don’t have any – we’ll conduct what’s called “translation alignment”. This is when we make use of your existing translation materials to match your source and target language files side-by-side to work out which pairs belong together. This way, we don’t need to build a translation memory from scratch, and we can save time and money on never having to translate the same sentence for you in the future.
Once the translating starts, the software will present the translator with database matches, which they can accept or modify, saving a significant amount of time.