What is Machine Translation (MT)
Machine Translation or MT Is something you probably use more often than you think. Ever use Google Translate to translate a phrase? Or what about that link in facebook posts that says See Translation? Both are examples of machine translation, with AI-driven algorithms parsing the text and serving up the translation. Most of these translation systems these days are what is known as neural machine translation or NMT, which means they can learn and improve over time as more people use them. And they have seen major improvement in the last few years.
Will these robot overlords replace puny human brains?
So, are they a replacement for human translation by experienced translators? The answer is… sometimes. Examples might be when you simply need to get the gist of the content you are looking at on Google or Facebook. However when you’re dealing with complex translation tasks where accuracy and cultural context are critical, machine translation alone, though constantly getting better, falls short.
A little history and the rise of post-editing
Machine translation, as a challenge for software developers, goes back the dawn of computer science in the late 1950s and 60s. Like most computer functionality back then it was extremely slow and required far larger amounts of processing power than most of those early engineers could even imagine. And when you got them, the translations were pretty terrible and/or extremely primitive. Which led to the rise of a translation discipline known as post-edit or post-editing. Post-edit, to put it simply, is a human translator fixing the machine translation to make it understandable and useful.
Machine Translation with post-edit is a win-win for our clients
Fast forward to today where we carry the equivalent of a super computer in our pocket or handbag. A super computer that can connect to NMT services like Google Translate and deliver translations nearly instantaneously. Post-edit of these translations is often a vetting and cleaning up process and post-editing has become a profession or add-on skill for many in the translation community.
Post-editing, along with other translation tools like translation memory software, make the use of machine translation common in our industry. The content is machine translated and then a human, native-speaking translator goes through the entire localization process from start to finish and considers context, accuracy, and relevancy. MT has entered our workflows and it means both cost and time savings for our clients. Post-editing is a form of quality control to ensure that messy machine translations don’t slip through.
Together, this localization process workflow between humans and machines is changing the way people communicate globally, opening up barriers, and enabling businesses to enter new markets faster and cheaper.