Is Machine Translation Every Translator’s Friend or Foe?
“There is no sweet potato, there is no reciprocation.”
This is what you get when you ask Google Translate to translate the Malay proverb “ada ubi ada talas, ada budi ada balas” into English.
Most of us have tried getting something translated using Google Translate and had a good laugh about the outcome. Yes, the proverb above does mention potato and reciprocation, but an accurate translation of it would tell you to “do unto others as you would have them do unto you” — an interpretation that Google Translate missed by a long shot.
Discussion of the rise of machines and AI and jokes about them taking over our world are nothing new. In fact, we are already seeing the intimate intertwining of contemporary society and technology happening, such as with the increasing use of smart homes and Alexa to make our lives more convenient.
We have also seen many jobs traditionally done by humans, such as various stages of an assembly line, becoming obsolete when companies realised they could get machines to do the same job at a quicker pace and with lower frequency of error.
Many fear that machines will soon take over most of our jobs, especially the more mundane and technical ones, but will it take over translation too?
The wonder of technology
There are many advantages to using software to get translation work done, most notable of all being speed.
It is also budget friendly and especially useful for translating large bodies of straightforward, technical text that may get too repetitive for a human to work on. And when you need something translated in half hour at three in the morning, you can trust software to get the job done without a complaint or demanding twice the usual pay.
In addition, machine translation is helpful when you want to understand the gist of, say, an instruction manual written in a foreign language without needing every line to be worded eloquently for you by a professional translator.
While machine translation is notorious for taking a phrase hilariously out of context — as we have just witnessed — it is improving in accuracy and relevance all the time, to the extent that machines can detect and understand some of the more common idioms.
To take another Malay proverb as an example, Google Translate successfully translates “ada udang di sebalik batu” to “there is a hidden agenda” instead of the direct translation of “there is a prawn behind the rock”. This demonstrates technology’s constant advancement in the area of translation.
That said, machine translation still has a lot of room for improvement.
Not quite perfect yet
It’s true that machines can translate a long list of independent words into another language much faster than humans can, but once you string those words into a sentence or paragraph, more elements of a language such as syntax, metaphors, humour and tone are thrown into the mix, requiring a more sensitive evaluation of the source text that machines are not yet capable of excelling in.
Even a computer program that gets most things translated correctly will still miss certain nuances of a language or culture. For example, a machine may translate “bagai duri dalam daging” to “a thorn in the flesh”, which is mostly correct. However, the Malay and English phrases each mean slightly different things.
In Malay culture, “bagai duri dalam daging” means something that troubles your heart, while “a thorn in the flesh” in the English language means something that is a constant source of trouble or annoyance to you — they have similar meanings, but to be troubled is not exactly the same as feeling annoyed about something. These slight differences are the things a good translator should pick up on and take into account when translating a piece of work.
Just for fun, here’s another proverb translated into English:
Air dicincang takkan putus > The chopped water won’t break
In getting the literal meaning right, Google Translate fails to recognise the phrase as a metaphorical expression. There is an English equivalent for this proverb, which is “blood is thicker than water”, but it might take a while more before machines get the memo.
Currently, most translation work done by software, even of simple texts, require human review and revision to ensure that every phrase or sentence is translated within the original context. After all, you cannot fully trust even the most advanced computer to never take a phrase too literally. Even we humans are not immune to making that mistake.
This brings us to the question: Should we even bother utilising computers to translate for us if we have to go through the final work anyway?
Increasing productivity with machine-assisted translation
While leveraging software does not free us entirely from doing the work, it can increase the speed and efficiency of a person’s translation process.
Human translators, no doubt, have the upper hand in producing a translated work that carefully considers the source text’s context, cultural idiosyncrasies and linguistic quirks. However, by utilising computer-assisted translation tools, we can make the quality of our work even better.
One of the things such translation tools can do is to store segments of a translator’s work as translation memory. The translator can refer back to units of the memory anytime to ensure consistency of style and terms used at later stages of translation, sort of like a style guide. This is most helpful when a translator has hundreds of pages to translate and wants to ensure consistency is maintained from beginning to end.
If you are a translator worrying about your relevance in the age of AI-takeover, fret not. Instead of seeing technology as the enemy threatening to rob you of your livelihood, you should see it as an assistant that helps you excel at your job.
Written by Amanda Soo