Partager :

Vidéo Jan 24, 2025

What is Neural Machine Translation

Hello. Welcome to MotionPoint Minute. I'm Verity Coburn.

Today, we'll be talking about neural machine translation. What is it and how best to use it for digital localization?

So neural machine translation.

This is a type of machine translation that’s been around for the last six or seven years, and so the chances are if you’ve used Google translate or Bing translate or another machine translation service you have been using neural machine translation.

Neural just distinguishes it from the older types of machine translation such as statistical and rules based machine translation.

Neural machine translation uses supervised learning to train a neural network on how to translate from one language to another.

So what about the quality of neural machine translation? Well this can vary quite a lot from language to language as it's very dependent on how much source and target data is available to train a model. So for English to Spanish, for example, because this pair is translated more often, there's more data out there, whereas for say English to Norwegian the quality might be lower because the machine has had fewer examples to to learn with.

Nowadays the overall quality is considered good, though, of course, human post editing is recommended for your high visibility content. You can use neural machine translation to your best advantage by using it for your high volume, low visibility content, and then combining it with post editing by human translators for that higher visibility content.

If you use neural machine translation in this way, you can actually lower your translation costs by about fifty to sixty percent.

The amount of work needed by human translators to post edit content versus translating from scratch is around thirty to forty percent less.

So you might be wondering where do large language models such as chat GPT come in?

Tune in to the next motion point minute where we will talk more about that. But that's it from me for today. Thank you very much.