A new era for translation memory: smart matches, accelerated reviews and much bigger impact

On November 5, at DeepL Dialogues Berlin, we announced the launch of translation memories on DeepL. It means you’ll be able to leverage previously approved translations every time they’re relevant on our platform. But that’s not all. Because we’re connecting translation memories to DeepL’s unique AI capabilities, they can do far more for you. Here’s why that matters.

The promise of translation memory 

Human memories are remarkable things, and they play a vital role helping people operate effectively. They’re recalled instantly whenever they’re needed. They guide our behavior. They help us perform tasks automatically, without having to think about them. Through all of these things, they help to make us who we are.

Translation memories are meant to do something similar. The idea is that they remember when a piece of text has been translated before, and recall the translation so you can use it. This should reduce errors and keep translations consistent. It should help make sure your brand always comes across the same way. Above all, it should save your localization teams time.

The expensive tricks that traditional translation memories play

The problem is, translation memories struggle to live up to this role. Too often, they are a memory you can’t rely on. Teams may spend less time translating similar passages of text. However, that saved time gets swallowed up checking that translation memories have been applied in the right way. Are the translation memories up to date? Do they really fit the context in which they’re being used? These are the questions human translators end up asking themselves, time and time again.

What’s repetitive and time-consuming for translators quickly becomes an expensive risk for organizations. If they’re using an external Language Service Provider, they often need to pay for every correction. When translation memories are clumsy and outdated, the costs quickly add up. Localization projects face unpredictable delays because time savings don’t materialize. Operational efficiency suffers.

There’s also the danger that translation memories end up preserving mistakes that editors have missed in the past. If they do, those mistakes will be repeated every time the translation memory is used. Translation memories are only as accurate and as useful as the original translations that they remember.

Transforming the value of translation memories with AI

These problems occur because translation memories aren’t dynamic, intelligent, AI-powered assets. They are static reference tools. They rely on people to manage and update them. They need human reviewers to check they are up-to-date and check that their recalled translations work in a new context. This all takes time.

When you connect translation memories to AI, things start to change. That’s what DeepL has done by making translation memories a core element in the DeepL Customization Hub. It means that, when you upload a translation memory file to DeepL, you can use it in far more ways and save much more time.

Precision memory matching you can trust

Because of the precision and accuracy of DeepL’s Language AI, you can be confident that the exact matches DeepL detects really do fit the previous translations in your translation memory. This confidence cuts down review time. In many cases, no human review is required at all.

Cut out time on post-editing corrections and adjustments

DeepL doesn’t just alert you about potential, partial “fuzzy” matches in translation memory. It works with you to improve those matches, making it easier to apply them more widely, and dramatically reducing post-editing time while keeping you in control. This means translation memories can be used for far more passages of text. Every time they are used, they need far less input from humans.

With DeepL, you set customizable thresholds to determine how close a match needs to be for a translation memory to be used. DeepL’s smart match capabilities then use AI to adapt the memory for you. DeepL resolves discrepancies, adjusts grammar to fit, and updates variables like dates and other items. With proactive suggestions like these, post-editing time and cost disappears. Translation memories do far more of the work for you.

Combine translation memory with a team of customization features

On DeepL, translation memories don’t work alone. They combine with three other powerful customization features within DeepL Customization Hub: glossaries, style rules and style profiles. Together they create translations that are absolutely right for your business and require much less editing time. 

DeepL glossaries set the translations of specific terms so that your expertise and brand voice always come through. Style rules enforce your preferences around tone, phrases, formatting and more. Style profiles add precision to how you use translation memories. You can specify exactly which translation memories are right and relevant for each project.

The AI control layer that’s as instinctive and immediate as human memory

Your translation memory on DeepL might use the same TMX file as on other platforms, but DeepL’s AI transforms what the memory is capable of. It helps you operate with amazing efficiency, while staying true to your brand voice and who you are. It saves so much time that everything feels seamless. Very like a human memory, in fact.

Discover more of what translation memories can do on DeepL here.