Reliable translation processes with (or despite) AI?

Artificial intelligence is ubiquitous. Many companies are quickly starting projects to find out whether it makes sense to integrate AI in their own business processes. The expectations are high: The benefits of automation, quicker workflows, and more economic efficiency are enticing.

At Gemino, we’ve already come a long way on the topic of translation and AI. And we can tell you, with or without AI, the following requirements are must-haves for translation processes for technical products, especially in regulated industries.

1. Correct translation of the content in the source text
   Keywords: liability risk & usability
2. Consistent use of correct specialist terminology
   Keyword: usability
3. Consistent and reliable reuse of previous translation
   Keywords: liability risks & less approval effort

How can these points be reconciled with AI? 

Especially considering that, besides delivering extraordinary translation results at first glance, AI (in particular, large language models like GenAI) also involves limits and risks.

You can probably hear it coming, it depends…

Our experience shows:

The key to an efficient translation process that performs as expected is a translation memory system (TMS), aligned with your company’s goals. The TMS is a translator’s daily working environment. Located on a server, it makes translation using correct specialist terminology possible.

And it does so time and again, because all translations are saved in a database and can be leveraged later on, regardless of whether the translator alone is performing the translation or if they are being helped by AI.

This means:
AI and TMS have to be considered together. 

The best-possible integration of AI in the translation workflow depends on optimally designed processes and a suitable system landscape. Regardless of whether you would like to use the corresponding tools yourself in your company or if a language service provider supplies them: The successful implementation of AI translation in controlled processes is only possible if a fully compatible TMS has been selected. We’ve described the aspects that have to be considered to select such a system in a short article, which you can find here.

Conclusion: AI is not the answer to all questions yet. But it can still be advantageously incorporated into the translation process with an acceptable degree of risk.

If you would learn about how Gemino is doing this already, feel free to contact us.

P.S. If you have a TMS selection project in the pipeline, take a look at our web tool TMSreq. This application helps you develop a specific requirements catalog and will save you many hours when preparing tenders and deciding on a TMS.