Everyone has seen embarrassing or funny translations on menus, t-shirts and advertisements. Laminated salad? Shrimp to the garlic? These types of mistakes usually originate from poor computer translations, and human translators can easily fix these errors. This is where machine translation post-editing (MTPE) comes in.
Although machine translation is much better nowadays than it was in the early days of computers, it’s still not perfect. When accuracy and quality is important, machine translation post-editing makes sure that automatic translations sound clear and natural.
What is machine translation post-editing?
Machine translation post-editing refers to the process of human translators editing text that was translated by a computer. They go through the text manually and correct mistakes to make sure the translation is accurate and readable.
Many good translation engines now exist, such as Google Translate, DeepL and Microsoft. They can translate large amounts of text in an instant, and this process is immensely helpful in the translation industry. For people who need to communicate often in multiple languages, this technology saves them a lot of time and money.
But sometimes machine translations are not sufficient, and human translators and post-editors need to make sure translations are high quality, clear and accurate. This is what post-editing in translation is.
Post-editing of machine translation: processes and applications
After a computer automatically translates a document, post-editors work on it to correct mistakes and improve readability. They look for major errors, garbled wording and ambiguous meanings, and they fix these mistakes manually. They often need to consult the original document to understand what the text should say, so post-editors should know both the original and target languages in order to do a good job.
When a human does post-editing translation work, they use tools to help them work faster and better. They often use translation glossaries, which provide definitions for key terms and describe how certain phrases should be translated. They might also use a translation memory, which saves terminology and phrases for reusing throughout the text. This process saves time by allowing translators and post-editors to reuse certain constructions and translations.
To make sure the writing style, formatting and word choice are consistent within a document, translators and post-editors also use style guides and localisation guides that instruct them on how to handle certain aspects of the work, such as spelling, punctuation and formatting.
Types of machine translation post-editing
There are different types of MTPE, or machine translation post editing. Light post-editing involves only minor changes to a machine-translated text. For this type, post-editors correct grammar and spelling, but only if the changes are necessary for readers to understand the text. They rewrite confusing sentences and fix any mistakes that the machine introduced, but they do not edit to improve the overall style of writing or to use different translations, unless necessary for accuracy.
Full machine translation post-editing, on the other hand, involves an in-depth process. These post-editors also correct mistakes, but they make additional changes to improve style, tone, and cultural and linguistic considerations, such as double meanings and idioms.
With light post-editing, the final text might sound a bit repetitive or unnatural, but the writing is clear enough that readers can understand what it means. The point of this level of machine translation post-editing is to maximise efficiency and quality.
With full post-editing, the final text sounds natural and is not repetitive or awkward. This level of post-editing takes longer and is more expensive, but it is necessary when quality is the most important factor, such as in legal translation.
Best practices for successful machine translation post-editing
If machine translation alone is not sufficient for your project and you know you need machine translation post-editing, there are several aspects to consider to avoid post-editing issues in translation work.
- Start with clear, clean text in the original language. Machine translations work better with text that has simple vocabulary, straightforward grammar and no mistakes. You can clean up your document by removing typos, avoiding passive voice, simplifying complex wording and using short sentences. The cleaner the text going in, the cleaner the text coming out.
- Use a good machine translation engine. Many machine translation engines exist, some of which are quite good. Others, however, do not produce good results or work better with certain languages than others. It’s a good idea to do research to choose the most effective translation engine for your project. A good translation will save you a lot of time and money in post-editing and corrections.
- Work with experienced post-editors who specialise in your field and language combination. Ideally, post-editors should be familiar with the original language in addition to the target language. And just as with human translations, machine translation post-editing results are more accurate and efficient if post-editors have experience in the applicable field.
Human vs machine translation
You might be wondering why people use machine translations at all if their outputs need post-editing. The reason is because it saves time and money. Using a good machine translation engine does the initial bulk of the work. If the machine output is high quality, post-editors can correct the text much faster than a translator can translate the text starting from scratch.
However, there are times when human translators should be used from the beginning. If the writing is highly complex, it might be better to avoid machine translations because the output can be too ambiguous or inaccurate. When this happens, it can take post-editors a long time to correct it, or it might even be impossible to make the text sound natural. When accuracy and quality requirements are of utmost importance, such as with legal or medical documents, consider using specialised translation by humans rather than machines.
If you’d like to know more about what post editing machine translation is, check out Berba’s post-editing services. They have experience converting machine translations into human-like quality for many different language combinations.