Demystifying machine translation: Do translators need to worry?

"See translation". Two increasingly visible words in the digital environment. Whether it's about built-in machine translation (MT) apps and gadgets in social media, browsers, or even the soon-to-be-available wireless earplugs connected to smartphones that translate human speech as we speak, machine translation is all around us. Machines are learning and improving exponentially, and attentive translators could be increasingly bothered by the fact that some mystical Joe Bot takes over their potential income. There are a few reasons, why translators don't (yet) need to worry about their remote future and well-being. Not just for general or Shakespearean style translation, but also many other fields.

Will machine translation replace human translators?
Source: AdobeStock.com

Three reasons favoring human translators

1. Machines think logically, while human language is not logical.

Global human language as a generalized form of any human communication is like an organism. It's highly complicated, imperfect, evolving and even somewhat illogical if the intention is to to rewrite and transfer the written word into the rigid, binary world – ergo the language of machines. Over seven billion people with various historical and cultural backgrounds have accumulated such a vast number of fuzzy content to deal with that even with increasing computing capacity and advanced artificial intelligence, today's MT solutions are still unable to deliver satisfying results.

2. MT discriminates.

The more "logical", in-use and exception-free a language is (English and Spanish are good examples), the closer it gets to the thinking, or processing, of MT bots. Frankly, sometimes it's rather disturbing how accurate MT gets with translation into English, but the more complicated rules and exceptions and illogical nuances a language has, the more desperate results the translation apps return. Eventually, it's always the human brain that puts the pieces together and one can probably conclude that the less narrower  your language focus is, the longer you can laugh at what Google and other service bot translators deliver, albeit with some exceptions.

3. MT can slow down the translation process.

Simple, short segments, mainly of general text and even technical stuff, constitute an area where MT can be very helpful if the work-flow of the translation environment is well designed. With well-prepared translation resources, MT can save time for  translators as it can correctly translate a fair deal of content. However, the "fair deal of content" – of course depending on the language – really does not make up (based on our internal data and experience) much more than about 10-15% of the overall text mass. Longer segments are still a challenge for translation engines, and often they deliver unintelligent results that can take a lot long for a translator to straighten out than simply translating such segments from scratch.

Together with translation memories and glossaries, translators have an arsenal of content at their disposal, which of course can help but this can also distract and cause loss of time when the translator tries picking his/her best options among all given resources. Decision paralysis, or work obstruction for lack of a better word, are well-known to translators working with MT.

They need to make numerous decisions for each and every text segment they work on: which fuzzy match would be best for the segment, is the machine proposal acceptable, does glossary proposals interfere, is grammar correct, etc.?

Summing up, we have a paradox where certain types of work-flows with MT intended to speed up the translation actually does the contrary and slows down the process as a result of overinput, human hesitation, a quest for perfection and vagueness and inconsistency that a human trained translator has learned to avoid. 

The main argument talking against MT is the way we interact and think. Yes, machines can beat the human brain in terms of volume and sheer processing power, but until MT actually becomes able to reflect human intuition, work flexibly and with an ability to improvise, impress and excel, real quality – which is what matters most in translation – namely the art of translation will still rest with us humans.

Future of MT

The question then is if there is a future for machine translation. The answer is a definite Yes. It will without doubt become a valuable tool to professional translators working in scientific fields. Once we learn to tame the output and recognize the short-comings, there is a good chance it can boost productivity.