Should you use AI or humans for your Chinese writing translation?

Updated: Aug 24, 2018

Artificial Intelligence (AI) is disrupting pretty much every field of human endeavor at the moment, including language. The word on the street is that AI will soon become better than us at every single job we do right now and much sooner than that, it would even be awkward to consider humans for your Chinese writing translation.

Yes, you read that right. That big contract you need translated for the China deal? You can translate it right from within your office or even better, your home. You don’t need to engage any pesky humans to do the job for you. AI will get it done for your just as well as any human would.

But is there a lot of truth to this? Are you really better off turning your business and legal documents over to AI for translation?

The truth? No. You’re not better off. And this article shows you just why you should stick with humans for your Chinese writing translation, instead of AI.

Can AI really do well at Chinese writing translation?

Okay, we’ll be upfront about this. AI is really good at translation and interpretation. As it’s doing to many other industries around the world, AI has extremely great potential to revolutionize the translation industry, and it’s well on track to doing so.

IBM has launched an earpiece that can translate languages in pretty much real time (say hello to the Babel fish). Microsoft recently announced that it has developed an AI that can translate Chinese as well as a human. The AI was used to translate news articles and was adjudged to have produced the same quality as a human.

Google too are not lagging behind. Their Neural Machine Translation, which mimics the processes that a human brain uses to identify and translate language, is also set to rock the sector. Using this process, Google was able to achieve AI translation results of 5.0, just below the average human’s score of 5.1.

So, AI is actually pretty good at the job of language translation. Even more, AI has a frightening capacity to learn and grow. Strategies like Natural Language Processing, Neural Machine Translation and Unsupervised Machine Learning have been adapted to help AI learn from its past actions and become radically better in only a short while.

Who does translation better? AI v Humans

Now that we’ve established that AI is pretty good at translation, we’re itching to set up the ultimate translation showdown between AI and humans.

In the blue corner, wearing the ritzy shorts laced with digital circuits is AI. In the red corner, red for all the ketchup and fries consumed during translation jobs, is the human rep. Who wins in this ultimate grudge match?

Thankfully, we don’t need imagination to supply us the remaining details of the match. A showdown between AI and humans was set up in February 2016 and needless to say, the punches flew far and fast.

The contest, which was set up by the Sejong University, Seoul, in collaboration with the International Interpreters and Translators Association of Korea saw a team of human translators face off against Google’s Neural Machine Translation AI.

Unfortunately, the contest didn’t end so well for AI. Once again. Yes, you read that right. The 50-minute competition which involved the translation two never before translated Korean texts, one literary and one non-literary, ended with the humans as victors.

According to the reviewers, about 90% of the AI translated text was “grammatically awkward”, definitely not the sort of translation you’d want for your important business and legal documents.

Why you should stick with Humans instead of AI

Clearly, AI is good, but it’s not as good as humans. At least not yet. There’s a lot of reasons why you’re better off sticking with humans. Here’s 4 of them.

Language and its processing is extremely complex

If you’ve followed the discussion in this article closely, you’ll see that AI-powered translation is basically a mimicry of how the human brain translates language.

However, the process of how humans decode and translate language is pretty complex and the truth is we haven’t even figured out a whole lot of it.

So much happens from the moment a human reads a sentence in one language to the moment they settle on an acceptable translation in another language. AI, as yet, cannot take all these stages into account.

This is why AI still struggles with a lot of translation that we might even consider elementary. Examples like these hilarious AI translation fails show that professional human translators are still your best bet to quality translation.

Context is still king

Context is everything and unfortunately, it’s one thing that AI just doesn’t get. The art of translation does not consist in just interpretation of single words into their target language equivalents.

A translator has to read the source document carefully (and understand it), determine its context, interpret the text into its carefully chosen target language equivalent and rearrange as necessary so the translated document reads essentially the same as the source.

Fail to determine context and the results could change drastically. This Quora user shared his experience of some AI translation of an online test gone wrong. In the English version of the test which was to be translated into Lithuanian, (M) was used for Most and (L) for Least.

It turned out that the word for Most in Lithuanian starts with an “L” and the word for Least starts with an “M”. Needless to say, the fail rate for that question was stratospheric.

Machines don’t always “get” language

Generally, language is a reflection of human thought. Human thought is both rational and emotional. It’s the emotional part that often proves a sticking point because it doesn’t always “make sense”.

AI is programmed to follow rules. There’s always a “right” outcome. It simply achieves its set tasks by making sure it ends up with a right outcome consistently. But if there’s anything we know about language, it’s that there’s not always a right outcome.

There are many English words that don’t have direct equivalents in other languages. What translators do with these words is to find the best collection of words that can satisfactorily translate the meaning in that word, in the target language.

AI finds this very difficult to do right now. That’s why a lot of AI translation can end up being grammatically awkward and socially inadequate.

Language is not constant

Every day, old words are being used in new ways. Language, being an entirely human phenomenon, changes as regularly as the sun moves across the sky.

The pace and scope of this change can be quite astonishing. What made sense yesterday may not necessarily make sense today, especially slang words and professional jargon.

This is a changing reality that AI is not entirely equipped to cope with right now. It’s all too easy for AI translation to use grammatically correct words that a local professional would not touch with a long stick.

Bottom line: AI is really good. But it remains only a really good human tool for now

So, no matter how you look at it, AI is pretty good at translation. However, it’s not ready to go solo just yet.

It remains a very useful tool though. A wide range of AI solutions are available to help professional translators do their work better and with the constantly evolving landscape of AI, these tools will keep improving.

AI continues to be a major driver of disruption in every area of the world. But when it comes to Chinese writing translation, you’re better off sticking with flesh and blood for now.

38 views0 comments

Copyright © 2018 Foundation Language Consultancy Limited. All rights reserved.