Best Practices for Machine Translation Post-editing
Machine translation is not what it used to be. In its early days, it was clunky, inaccurate, and often unproduced unreliable output to the point of hilarity (and even going viral for, well, the wrong reasons). But as it had more and more input to work with, it vastly improved its translations to the point where they can be reliably used—at least, most of the time.
In fact, it’s evolved to the point where it can be used to efficiently translate entire websites into multiple languages without breaking the bank. But, can it replace human translation?
Quality-wise, human translation leads the way and it might be a while yet before machines reach the expertise and parity of having lived and spoken a language natively. That’s why machine translation post-editing—which combines the best of machine translation and human knowledge—is an indispensable part of cleaning up machine output and attaining the highest translation quality possible.
Let’s dive in.
What Is Machine Translation Post-editing?
Machine translation post-editing (MTPE) is the process of using human expertise to review and correct machine-translated content. It’s an efficient way of combing your content and ensuring they’ve got the highest possible accuracy, since it unites the speed of neural machine translation (NMT) and the fluency of a native speaker.
Make no mistake; machine translation output these days is mostly reliable thanks to artificial intelligence. However, it’s not always perfect. It’s still capable of mistranslations, and it doesn’t always capture nuance or idioms. Post-editing is a surefire way to optimize your content translation and make sure you get your message across immediately in your target languages.
But where is it in the process? It comes after the first step, which is using AI-powered machine translation software to translate your website content. Then, once it produces your output, you can then have your post-editors review the translated content and make the corrections as you see fit. That way, you get to ensure that the intended nuances, voice, and tone stay intact.
Weglot makes machine translation post-editing easy. All you need to do is to go to your Weglot Dashboard and make edits through either your Translations List or via the Visual Editor, which allows you to edit translations directly into a live preview of your website. As a bonus, you can even order translations done by a professional straight from the dashboard.
What’s the Difference Between Machine Translation and Machine Translation Post-editing?
What separates machine translation and its post-editing is that the former is the raw MT output you get when feeding engines like Google Translate or DeepL your source text. It’s an incredibly convenient method to get dependable translations without having to wait months (and stalling your business strategy in the process). After all, these tools supply you with translations within seconds or minutes.
Since machine translation relies on large volumes of data to produce accurate output, that makes it ideal for content that doesn’t need to be engaging, like instructions or other technical text. It’s also great for verifying the translation of a few words. You can even preview how the translated text may affect your website’s layout and design (something called text expansion and contraction).
But if you’re hoping to use the output to persuade an audience, like your website or marketing materials, the output might need an additional layer of quality assurance. And that’s where machine translation post-editing comes in.
Why Do You Need Machine Translation Post-editing?
There are two kinds of machine translation post-editing: full post-editing and light post-editing. As implied by the name, full post-editing is a comprehensive review of the output with detailed edits to ensure the voice is consistent and exactly as you want it. It’s more time-intensive, but it’s the recommended option for content with high traffic.
Light post-editing involves a rapid scan of your output and highlights obvious mistakes. These include misspelled words, wrongly used terms or phrases, and missing punctuation. Though this offers faster results, it won’t be as thorough as full post-editing.
But why would you ever need machine translation post-editing? Here are a few reasons:
It Saves Resources
The biggest draw of machine translation post-editing is that you can sharpen the MT output without spending huge amounts of time or money on it. In the first place, the raw output of MT tools has vastly improved over the years, which means that you often don’t have to make too many edits (depending on the language pairing).
In fact, besides paying for the tool itself, you may not even have to spend a single cent on MTPE. If you have an in-house linguist or native speaker, they can have a good look at the output and clean it up right away. There are also translation management systems equipped with post-editing services that offer all this at affordable pricing.
It’s a cost-effective method that will bring you high-quality results and make a huge difference in the way you connect with your audience.
It’s Time-Efficient for Translating Large Amounts of Text
Naturally, the more you have to translate, the more intensive the process is. But when you have a reputable first layer to work with, much of the manual effort is removed.
MT tools are efficient at eliminating glaring errors like typos and misspellings. That means a quick sweep is often enough to clean up the layer and refine it to make it more presentable. Even if there were bigger adjustments to make, a website translation solution like Weglot has already done the bulk of the legwork for you.
Additionally, MT quality is the best it’s ever been thanks to advances in NMT. That means your translation workflow will be much smoother since it automates the bulk of the process for you.
It Improves Your Translated Output
MT post-editing is a quick way to take the quality of the raw target text to the next level to be customer-ready. Your customers will know that there was thought and care put into the translated version of your website, as it’s easy to detect unpolished, machine-generated translations.
Why Not Use Human Translation Instead?
It’s clear that there’s no beating the quality and precision of a native speaker. They will always understand the nuances, differences, and preferences of the language better than a machine ever could. But the tradeoff in quality is in the long turnaround time of human translation, not to mention the cost. Depending on the volume of texts to be translated, it could take months to fully translate everything.
That’s why the post-editing of machine translation is the best of both worlds: you’ll have the speed and efficiency of machine translation, and the expertise of a native speaker to come up with high-quality translations. You can’t stall your plans and projects for months just to wait on the different translations you’ve ordered!
Best Practices When Post-editing Machine Translations
To ensure you get your MTPE right, there are a few practices you’ll need to keep in mind in the post-editing process:
Use High-Quality Translated Source Text
Not all MT tools are created equal. Some are better than others for particular language pairs. For instance, in our study on machine translation tools, DeepL ranked the highest for English-Spanish (and vice versa) translations, and Google Translate did the best for German-English (and vice versa). When your source text is as accurate as possible, it’ll make the rest of the editing process a breeze.
Select a Website Translation Solution That Automatically Uses the Optimal MT Engine
Integrating a machine translation engine into your website like Google Translate API is one way to go about it. But using translation management software automates much of the workflow for you.
For instance, you can lessen the manual translation work required by opting for a website translation solution that automatically assigns the best-performing machine translation engine to your language pair—something that Weglot does.
Weglot makes it simple to translate and localize your content. It automatically detects, scans, then translates all your content for you. It even translates new content uploads or edits right away. For instance, if you edit the text in a blog post in your source language, it will immediately deploy the change across all language versions.
To make things easier, it also has translation glossaries. That means it will save the manual translation edits you make and apply them across all your translation projects.
Even better, you can easily order translations from professional translators through Weglot’s Dashboard if you don’t know anyone to verify the target text’s accuracy. You’ll have your translated text in as soon as 2 business days.
Be Aware of Common Machine Translation Mistakes
One thing’s for sure: you’ll get accurate translations from AI-powered machine translation. But it’s good practice to know the common mistakes that are found in its raw output. That way, you can easily look out for them and correct each one so that they don’t slip into the final product unnoticed. These include the following:
- Wrong or missing articles
- Incorrectly translated words
- Additional or forgotten words
- Incorrect punctuation, gender, capitalization, formatting, or order of words
- Words in the original language left untranslated
…and other considerations you’ll need to make for a multilingual website.
When you know what to look for, post-editing becomes much faster to do!
Implement a Uniform Brand Voice
When you have more than one person editing your translated output—whether an in-house team or translation services—be sure to centralize your editorial guidelines, so they’re easily referenced. Be detailed in what your brand style is: how many sentences per paragraph should there be? What is your preferred tone–is it casual and friendly, or brisk and formal? Do you write numbers as digits? What about Oxford commas?
It may seem overwhelming, but when you have standardized editorial guidelines, it creates a framework for all your content to be proofread under. And when you have something to consult, it’ll be easier to know when something should be removed or tweaked to fit your brand.
Edit Thoroughly but Wisely
Though it’s crucial to get translations as accurate as possible, it’s easy to get lost in trying to perfect them. Your priorities should be ensuring the translations retain the same meaning as the original text and removing offensive translations. Remember, the goal is to eliminate as much manual labor as you can!
Another thing to consider—what sounds witty and clever in English may sound bizarre in another language. So watch out for idioms and expressions that could be easily misinterpreted (or completely mistranslated).
Make deliberate changes when you edit the output. Remember – 65% of users prefer reading content in their native language, even if the quality could use some work!
Lastly, of course, is to give the content one last quality check before putting it live. Your TMS is often good enough to get rid of obvious mistakes. But you may spot typos or misspellings you missed the first time, so it’s always a good idea to give it a clean sweep before publishing it!
Conclusion
Machine translation accelerates your localization and translation efforts exponentially. What used to take months is now ready in minutes, and adding a layer of post-editing can completely transform the different language versions of your website. Weglot makes this process easy with its intuitive post-editing interface and machine translation, powered by the leading NMT providers.
Curious about just how fast it’ll translate your entire website? Give it a go through the 10-day free trial.