As a client of ours, or someone else generally interested in translation services, you may be unaware of the seismic changes we are experiencing in our industry at the moment. For decades we have relied on a technology known as ‘translation memory’. This has cut man hours and costs down by significant amounts, allowing us to work quicker and more cost effectively.
However, new technology has now come along, which threatens to make translation memory superfluous.
Here, we look at the repercussions of this shift and how we, as translation service providers, can ensure the service we provide is always the very best, regardless of the tools we use to help us.
What is translation memory?
Imagine if we had to translate every sentence, every time we complete a job. Even those sentences and phrases that we have seen a million times.
This isn’t how we work. Instead, we use a translation memory tool which stores the current, optimised translation for phrases and sentences when they have been translated once, for use again and again on other documents or projects.
Before this technology was available, projects would take hours, days or months longer than they do now and the costs would be too high for many business customers to consider. Back in those days, we didn’t live in such as global business world, so translation services were not as in demand as they now are.
The introduction of translation memory tools coincided with the increasingly globalised world, leading to a massive increase in the use of translation services.
What’s happened now?
A little while back, our industry started using another form of technology to help accurately translate written material; called adaptive trained neural machine translation (ATNMT). This technology uses the very latest machine translation tools and we use them alongside our human experience and skills to produce accurate and sensitive translations.
So, what’s the problem?
As an industry, after we started using ATNMT, we began to realise that, sometimes, the results it came up with would not correlate with the translation provided for us by the translation memory technology. Most of the time, the translation memory was the more accurate of the two versions, so we used the two tools alongside each other.
Now, we have got to the stage where ATNMT has improved to the extent that it is coming up with more accurate translations than the translation memory tech.
More accurate, doesn’t mean that in the past we have been using inaccurate translation, it just means that we need to undertake a rigorous editing process before sending the end result to the client. This takes time and costs money.
Is this the end for translation memory?
In a nutshell, no.
Every project is different and often, the way we like to work is to balance the cost / benefit of translation memory against the other options available. Often heavy edits from ATNMT are associated to increased costs – even vs. translating from scratch.
No industry is unaffected by technological disruption, and it is important that we understand on a case-by-case-basis the ways tech is impacting us and our clients. It’s also important that we don’t just jump onto a technological trend that doesn’t improve the quality of our translation work, nor the efficiency of delivery and therefore cost to our clients.
We believe that our industry must move with the times and evolve in order to continue to provide the very best service to their customers. If you share our perspective, get in touch with the team at Bubbles Translation today and let’s discuss your project.