Neural Machine Technology and the Translation Industry

By Atlas LS

The global translation industry is worth $40 billion, which is certainly nothing to scoff at. That said, there’s bound to be a big shakeup on the horizon. Machine learning could make serious inroads into the translation industry with something known as neural machine technology.

Translation experts believe that within three years neural machine technology could handle as much as half of the work currently done by flesh-and-blood translators. If these predictions prove accurate, then neural machine learning could alter the careers of some 500,000 human translators and tens of thousands of translation agencies.

Others believe that neural machine learning will simply augment rather than replace human translators. Two years ago, translation technology could give a serviceable idea of what a text had to say, but most nuances in the text were lost in translation. Translators know the importance of idioms and subtle variations of meaning, especially in the context of business negotiations and conference translation services.

There are experts in the translation industry who believe neural machine learning has advanced beyond the benchmark set two years ago. Neural machine learning might now be on pace with human-quality translations in a number of different fields.

“…many of the documents that are translated by neural machine learning need to be ironed out by a human translator.”

Giving more credence to the idea that machine learning and AI will augment rather than replace human translators, though, many of the documents that are translated by neural machine learning need to be ironed out by a human translator. Subtle variations and different connotations in different cultures can make all the difference.

Top companies like Google and Amazon have put a lot of stock in neural machine learning, so this nascent technology is bound to be groundbreaking. The most technologically advanced machine learning tools use something called neural machine translation, or deep learning.

These companies will feed huge amounts of data on multiple languages into the neural net algorithms (i.e., computer programs, or commands) so that the artificial intelligence and neutral nets can glean as much meaning from each language as possible.

The really interesting part to all of this is that the neural machine technology works much better in certain applications than in others. In other words, the technology has a long way to go before it starts to displace human translators by the tens of thousands.

Here’s an example: machine learning can be a useful tool in the tourism industry for translating between English and German. Unfortunately, this same technology has many shortcoming when it comes to translating from, say, English to Japanese in the tourism sector.

The difference between a good and bad neural machine technology translation ultimately boils down to the complexities and idiosyncrasies involved in learning and toggling between two languages, both of which might have cultures and expressions that are so different as to defy the computer algorithm’s best efforts.

So, where is this all headed? In the future, human translators may be tasked with rating and correcting translations done more quickly by AI. There will also be many applications wherein human translators are more adept.