Human Translators are Still Irreplaceable Despite the New Technology
Language translation software developers have made great strides to improve the accuracy of machine or AI-based translation. However, it’s clear that there’s no way human translators become dispensable. The following reasons or examples should make this point even clearer.
- The Google Translate Hack
Just this month, Google Translate graced the news for acting erratically as it skewed translations into insults against Russia. The world’s leading search engine’s web-based translation tool referred to Russia as the Lord of the Rings evil fictional area “Mordor,” called its foreign minister Sergey Lavrov a “sad little horse,” and translated Russians into “occupiers.”
These insults emerged when doing translations from Ukrainian to Russian. Obviously, they have been inserted by those who have been protesting Russia’s annexation of Crimea and the country’s military operations in Ukraine. Google quickly corrected the wrong translations and explained that it happened because Google Translate does translations without human intervention. Everything is automatically done based on a base translation library and after scanning and analyzing the patterns found in a multitude of relevant documents (in hundreds of millions).
It’s unsure how “hackers” managed to make Google Translate churn out the insulting translations but what they likely did was some form of “Googlebombing,” wherein they purposely create a deluge of misleading content on the web (in their target language) bearing numerous target keywords and links that are bound to influence the translations generated by Google Translate. This method has been used by political activists to make the system present racist words for the White House and by overly enthusiastic fans to affront sports teams.
If the world’s governments relied on Google Translate for official or even for casual communication, this glitch would have led to a slew of diplomatic protests at the very least or even a ridiculous declaration of war if the one on the receiving end of the insulting mistranslations happened to be a country like North Korea.
2. YouTube’s Language Translation
Well, this one’s quite obvious and something you can test yourself. Just try loading a video on YouTube that features conversations in some foreign language then activate the captions and have it translated to your language of choice via YouTube’s built-in caption translation function. Your likelihood of getting good translations is probably as good as your chances of winning in the lottery. If it’s any consolation, you may not feel bad trying YouTube’s translator, as it will likely make you laugh hard over how the erratic translations come out.
This is not to say that it’s wrong for YouTube to have added this function. The introduction of new technologies for people to use is always something worth welcoming, even when the actual benefits appear only rather rarely. YouTube’s translation, somehow and sometimes, manages to yield a few understandable or sensible translations. It’s worth noting that the translation problem starts in how the captions are generated from the speech in the video. These speech-to-text translations are obviously far from accurate. Even spoken words that are clearly audible and intelligible don’t always translate to the right captions. As such, the resulting language translations also become erratic. This only goes to show that current speech-to-text technologies are still far from becoming significantly accurate.
3. Ili, the Wearable Translator
Make no mistake; Ili by Logbar (maker of one of the world’s first smart rings) is truly impressive. The idea and even the way it translates spoken words appear to have been well thought out. The viral video promoting the product even stands out although some may think it is somewhat inappropriate. The issue with Ili, however, is that it is limited to just a few languages. As of now, it can only work with English, Japanese, and Chinese. Of course there are plans to add more languages in the future but no timeline has been given yet. Moreover, you can’t help but notice how the reps of Ili at CES 2016 had to explain that they couldn’t give a proper demonstration because of the noise on the CES show floor. The device apparently can’t properly separate ambient noises from the sounds that should be translated.
4. Skype’s Real-time Translation
Last year, Skype introduced its real-time text and voice translation function that was met with mixed reactions. Some have been optimistic about it while there are those who think it isn’t anything more than the ordinary language translation services already available. The translations produced may help enable some degree of understanding for certain languages like Spanish, Persian, and German. They are not even 70% close to what a human translation can produce but at least they allow users to get the grasp of what the other party wants to say. On the other hand, just like in the case of Google Translate, Skype’s translation function also finds it difficult to do translations in more complicated languages like Hungarian, Hindi, and Turkish.
5. Manuals and Instruction Slips that Come with Products from Non-Major Non-English-Speaking Companies
You may have witnessed this yourself. Many of the packaging, labels, product documentations, instruction slips, or manuals of products imported from countries that are not native English speakers come with texts that appear to have been translated by machine translators. It’s either their translations are too literal or they don’t make sense at all. These translations could have only been made by “non-thinking” translators that know nothing about the nuances of language. If the companies who printed these labels or manuals actually paid for human translators to do such a hideous job, perhaps they need to ask themselves what they are doing with their money.
Human translators will remain irreplaceable because machine translators are still less than reliable. The technology for converting words spoken in different accents and idiosyncrasies into text or directly to translated audio is just not accurate, and those that are close to becoming precise, are limited to a very few languages — Yet, they still have a hard time when the language complexity rises.
Machine translators still don’t have the ability to take inputs from a speaker’s expressions, gestures, and cultural background to produce more contextually accurate translations. Moreover, partly crowd-sourced translation algorithms like the one used by Google Translate become terribly erroneous. Google’s attempt to make translations close to being contextually accurate, for instance, fired back as malicious parties tried to influence the algorithms.
Something like this will never happen with competent professional human translators.
