5 reasons why machine translation will never work

5

‘translating just one segment of spoken or written language requires approximately three times as many computer commands as is required to orchestrate the flight route of a guided missile.’

Those were the words of IBM scientists’ way back in 1950, and while it can be argued that technology has since then moved in leaps and bounds there’s no arguing the fact that it takes a lot of computing power to guide a cruise missile, but even then its accuracy is never 100%.

A less than 100% accuracy for a guided missile is by every military standards acceptable, however for language translation where a simple mixup in punctuation (as is in the sentence ‘lets eat, Grandpa’ and ‘let’s eat grandpa’) connotes two entirely different scenarios, suffice it to say that less than 100% is unsuitable.

Frankly, it’s going to take massive computing resource to achieve 100% accuracy. So while the promises of machine translation usurping human translation keep reeling in (we’ve been getting such for over 60 years now), it’s now more than ever certain that this will not be the case and there are many other reasons to corroborate this fact.

There’s a difference between translating words and ideas

Regardless of how complex and sophisticated an algorithm (the basic framework on which machine translation subsist) it is still primarily a line of code using set grammar rules and integrated dictionaries to translate a word string in one language to its equivalent in the other. The meaning of words as they are used in sentences is, however, contextual and dependent on the intent of the speaker. It’s why the word vacuum in the sentences ‘she knew so little that her brain was a vacuum’ and ‘I used my mother’s vacuum to clean the rug’ have entirely different meanings. Machine translations have a hard time figuring out which is which.

Machine translation will behave like machines

And that means a carryover of slight errors in the written text or pronunciation of a speaker. The impacts of such mistakes are, however, more than slight as they could distort the entire context of the translated speech. Sure, it can be said that such errors are no fault of the machine translator, which is true. But that said, it is also pertinent to note that human translators will in most cases spot and rectify such errors if they occurred.

There’s no tone, context or depth – Machines are emotionless

What’s the difference between an acapella and a written script of words? The emotions of the performer. If a machine translator were to translate an acapella, it would most likely render it as an emotionless string of words. This analogy illustrates just how machine translations function in real life. They strip the sadness, happiness, enthusiasm or soul of the speaker out of his/her spoken words

Not every word has a direct (or indirect) synonym in every language.

Lacunae or words that have no direct translation in another language are more common than you think. For example, Kummerspeck, a German word translating to excess weight derived from eating during emotional distress has no English counterpart. A machine translator will most likely translate it into its literal meaning in English, which is ‘grief bacon.’ Nice try, but entirely off the point.

Machine translation can’t keep up with the dynamics and evolution of languages

In today’s world, words, expressions, and phrases are created on the fly, in fact per the Global Language Monitor a new word in English is created every 98 minutes. Their meanings also vary significantly across cultures and demography. ‘Bitch’ is a word a black American could comfortably use to address a female friend in America. Addressing someone by that same word in Scandinavian countries like … and … would most likely be seen as provocative.

In all, machine translation with all its capabilities and (overblown) glory is grossly incapable of deriving actual contextual and semantic meanings to spoken or written content. However, it’s not all doom and gloom for machine translation. In very specific settings and with efficient implementation it can produce rather impressive results – but saying it will usurp human translators that’s just plain old bollocks.

***

If you need a professional translation or interpretation done, with the highest quality and fast turnaround time, we invite you to get a free quote online or contact us now!





About the author

Add comment

By admin