MT Post-Editing

Machine translation (MT), also known as automated translation or instant translation, is the translation of a text performed by a computer. Human involvement comes before in the form of a software programmer to make machine translation possible. There are three types of machine translation systems:

Rules-based: They use a combination of languages rules and dictionaries for words.
Statistical: They learn to translate by analyzing massive amounts of data for each language pair.
Neural: They learn to translate through one big neural network, which is composed of multiple processing devices modeled on the brain.

Since machine translation cannot produce a quality product, human involvement is most needed (again) in the form of an MT post-editor. Post-editing is the process in which a translator/editor amends an automated translation to eventually present an acceptable final product.

Companies that develop and/or use machine translation claim that it increases productivity, so translations can be delivered faster by just post-editing them than by starting them from scratch.

While I agree that it can be comfortable to translate a text that has been divided into segments, and the use of a translation memory can be helpful, I disagree with the previous statement.

A translation memory is a database that stores segments that have been previously translated for future use. That means that you’ll be provided with a suggested translation every time you come across a word, sentence or paragraph that has been translated in the past.

In my mission statement, I acknowledge that translation software may help translators. For instance, I’m proficient in OmegaT and SDL Trados Studio. But I often prefer to work straight on Adobe Acrobat, Microsoft Word or Excel.

The main problem that I see with computer-aided translation tools is that pre-translations, or translation suggestions, are often misleading or just wrong. This can actually double the amount of work that you’d perform if you started from scratch.

MT Post-Editing

Humans can create translation matches and add as much information as possible, but a machine is never going to be able to understand and/or infer context. A machine doesn’t have the versatility and uniqueness of a brain. Language was created by humans. Translators should never have an automatic, default translation for any given word. A machine can replicate as much as we teach it to do, but some things are just not transferable. Professional translation requires more than turning words from one language into another.

That’s why I don’t support the claim that MT post-editing reduces time investment in translation. But why are some people advertising the quickest turn-around for as cheap as possible? Like any high-quality product, translation requires a certain investment of time to ensure precision and accuracy. And that time has to be paid accordingly. Otherwise, it’s a no-win situation: neither the client nor the professional will enjoy the best experience.

It actually seems that MT post-editing is helping some companies to make more profit by paying language professionals less.

I’m very grateful for technology. I always welcome any advance that will help improve the quality of my work as a language professional. I agree that a software creator, developer and/or seller needs to make money. Although it can be very fulfilling to volunteer for charity and non-profit organizations, who doesn’t want to be paid for his/her work?

What I don’t like is that some companies are taking advantage of translation software to earn more by paying professionals less. After all, those companies are still getting paid the same money by their end clients!

In some cases, it’s even getting worse: people are actually starting to pay to work. Not only you have to get software X to work with company X. Company X doesn’t even guarantee that you’ll get work from them! On top of that, people on behalf of company X are organizing workshops or seminars to get certified and/or learn how to use software X, plus all the add-ons that, of course, cost extra.

Besides believing that post-editing pre-translated text actually takes longer than translating from scratch, I’m certainly not willing to become the corrector of a translation machine.

I don’t take on MT post-editing assignments. And I don’t think you should either unless you specifically want professional translation to disappear. You may believe that post-editing is the best thing ever, a new niche for translators. You might even be organizing webinars to convince people to use software X so you can make some money on it.

But ask yourself: is it worth hurting the livelihood of thousands of professionals?

Share your Words