Machine translation (MT) has improved so much over the past few years that it can now complement a translator’s work perfectly: MT can save you an enormous amount of time by providing translations based on its algorithm.
It is no wonder that a whole new translation workflow has been built upon machine translation, Machine Translation Post Editing (MTPE). Many translators and LSPs now specialize in this hybrid method. The workflow revolves around a pre-translated text (typically furnished by a project manager of an LSP), which is then reviewed and edited by a linguist.
In the next section, we give you some tips on how to translate using MTPE, and present cases for which it might be useful. If you would rather jump to the new MT features we’re bringing you in 9.9, feel free to skip the next part.
Machine Translation Post-Editing: How does it Work?
As the name suggests, post-editing of machine translation is needed when a linguist has an MT engine translate a document and then provides post-editing for the machine translation. Translators and LSPs all over the world are taking advantage of this technique since it combines the speed of MT and the knowledge of the human mind.
The advantage of this method is that it can provide translations much faster than using only machines or only a human translator in case of certain documents. Of course, not all types of texts can be worked with this way—in some fields (such as life sciences) using only machine translation would actually be dangerous. Also, it must be noted that some editing is always required when working with a machine translated text. While machines are excellent at coming up with the translation of individual segments, their algorithms are typically not smart enough to always identify the context and choose the most appropriate string for a certain segment.
Let’s see some use cases where working with MTPE is worth it.
How to be Successful at Machine Translation Post-Editing
There are a few things you can do to get the best results when post-editing machine translation. First of all, know that not all types of texts are suitable for MT. You should either refrain from MT or edit your machine translated text multiple times and very carefully when you have any of these situations:
- Texts for advertising/marketing purposes: these are expected to be read by many, and they are often written with the purpose of being witty or funny. MT simply does not (yet!) have the lifelong experience and social context that a human has, so those jokes or meanings can be lost in translation if you use MT.
- Texts requiring expertise in a certain field (e.g. medical, legal, or engineering).
- UX/UI copy: these depend heavily on context; you have to see the entire user interface to be able to decide what words to use exactly.
However, other types of texts (such as blog posts, articles, informal texts) may be suitable for machine translation. Editing these is usually straight-forward, so you can save an immense amount of time and resources.
Post-editing machine translation: our top tips
- Make sure the source text is as suitable for machine translation as possible.
If necessary, make some tweaks or adjustments to the source text to prepare it for an MT engine. Be careful not to alter or omit any meaning from your segments. - Find the optimal amount of editing.
Speed is one of the key advantages to MTPE. If you find yourself altering MT-generated strings out of personal preference (instead of necessity), then using machine translation is not really worth the effort, and it might be more effective to start your project from scratch instead. - Choose the engine that best suits the project.
memoQ supports a myriad of machine translation engines (they are either built-in or can be added manually) for your localization projects. We suggest that you research and identify the engine that best suits the particular text and project before starting translation. - Prepare to edit machine translated and human translated texts differently.
When editing a text translated by a human translator, the errors you have to watch out are mostly grammatical, such as conjugation or subject-verb agreement. You also need to spot typos. On the other hand, when working with a machine translated text, the errors tend to be mostly contextual: make sure that all segments are translated accurately within the context of the entire document.
memoQ 9.9: New Horizons in Machine Translation
Whether you are a translator who wants to take advantage of MT or you work for an LSP using the Machine Translation Post Editing method, we hope you will enjoy the new functionalities in memoQ 9.9. The following features are all about making it a breeze to work with MT.
Machine translated segments are now part of progress reports
When you work on a project with memoQ, you can generate statistics with counts of the words, segments, and characters either in a specific document or the entire project. Machine translation has not been part of these statistics so far—but starting with 9.9, we’ve added a row that shows the number of segments translated with MT, so you can easily see exactly where your target segments come from.
MT Plugins: XL8 Now Supported, TexTra Updated
In addition to updating the TexTra plugin, there is also a new machine translation plugin, XL8, that can be manually loaded and used in memoQ. XL8 AI machine translation is a powerful MT engine that strives to make localization easier, especially in the fields of media and entertainment.
“Powered by Skroll solutions, XL8 removes language barriers by enabling customers to apply neural machine translation to content, audio, video, analytics and more. XL8 instantly identifies languages, checks the output quality of translations, learns from previous translations and analyzes content to make the translation process fast, streamlined and transparent.” (Source: XL8.ai)
The signed plugin package can be downloaded here.
If you rely on machine translation in your work, memoQ 9.9 is just for you. Download it today to enjoy the full functionality and take your MTPE game to the next level.
9.9 release machine translation
