Machine translation solutions have taken the world by storm. With promises for faster turnaround, lower costs and near-human quality levels, it’s easy to see why many businesses ask for it upfront. And it’s only gaining in popularity and sophistication as the technology advances.
While some speculate that this technology could one day replace human linguists altogether, we disagree. (We’ll leave the action-packed machine takeovers for the movies, thank you.) In fact, just because you use machine translation, it doesn’t mean professional linguists aren’t involved. It really depends on your content, target languages and quality expectations.
You may be wondering if it makes sense for you to use some kind of machine translation solution for your localization initiatives. It certainly has its appeal. However, machine translation may not work for everyone. So before you jump on the bandwagon, you’ll want to fully assess your unique situation.
1. Do you have an ample volume of content to translate and train engines?
Your company leaders probably wouldn’t run the business without any metrics or benchmarks to measure performance. Why would your translation program be any different? Translations are a major undertaking for many companies, but oftentimes it’s discovered many companies don’t actually know how their program is performing or even how much is actually spent on translations across the entire organization.
Just like a car needs regular checkups and maintenance to keep it running efficiently, your translation program needs similar diagnostics and tuning to keep it operating optimally. Advantages of globalization are only realized when you can benchmark and see the results of your efforts—globalization consulting services can do this for you.
2. What content types and languages does machine translation work best for?
Without a doubt, machine translation can save you a significant amount of money over time—but it helps to have a lot of high-quality content.
On average, our clients that use this technology see around 10 to 25 percent cost savings on new words. While it does come with a setup fee, the returns from machine translation can quickly offset that—especially if you have a large volume of content to translate. We’ve even helped some of our clients get 75 percent cost savings on new words thanks to the high volume of content that’s pumped through the process.
Lots of previously translated content helps us train your engines as well. The more high-quality content you have, the better the machines can learn your style and word choice preferences. They’re just clever that way.
We can take your translation memory, style guides and glossaries or any other multilingual materials you may have and use them to make the engines smarter. We also continually monitor the output to ensure that the engines keep cranking out quality substance over time. We can still train your engines if you don’t have these resources, but it may take longer for the machines to produce the level of quality you’d like.
There are several machine translation solutions you can choose from. The route you pick depends on the intention of your content. For instance, with any public-facing content we recommend a full post-edit, where a professional linguist looks over the raw output for any grammatical and style errors.
The coolest part? Machine translation with post-editing has been proven to produce quality that’s just as high as the traditional translate-edit-proof process, performed by professional linguists.
You can also choose a fit-for-purpose post-edit—where a linguist only corrects critical errors like grammar mistakes—and pure machine translation with no editing. Fit-for-purpose post-editing and pure machine translation solutions work well when you need only the gist of a message. An example of this is translating social media comments or online reviews.
If you’re undecided on which way to go, there’s no need to despair or admit defeat like you’re in some sort of evil robot sci-fi film. We can do a content suitability audit and look at all your target languages, content types, domains, quality expectations and more. From here, we’ll suggest a translation workflow that best fits your needs.
As you can see, you’ve got many aspects to consider with machine translation. If you are still debating whether it’s the right solution for you, plan to attend our free webinar led by me, Karina Martinez Ferber. I'm a Solution Architect at AMPLEXOR with a focus on language technology including machine translation solutions.
Do you use machine translation as part of your translation workflow? What made you decide it was for you?
➡ Read this article in French
➡ Read this article in German
Karina is Solution Manager Global Content Suite at AMPLEXOR. She is based in Berlin, Germany.
Karina Martínez Ferber 5min read 29/05/18