Translating Microbes: the Future of Automatic Translation?

In the sci-fi series Farscape, beings from different planets could communicate amongst one another by consuming “translating microbes” which allowed them to understand other languages. Might we one day see something along these lines?

Currently, various companies specializing in the area are developing the future of machine translation. What advances might we expect in this field in the near future?

Researchers at the University of Amsterdam are using artificial neural networks to assist machine translation systems to translate words based on their context, which is something that current systems are quite limited in performing. Neural networks are paradigms of learning and automatic processing that are inspired by how animals’ and humans’ nervous systems work, and they are widely used in the field of artificial intelligence. The benefits of this advance would be particularly evident in translations from languages with simpler grammar, such as English, into languages with more complex grammar, such as German or Russian, where machine translations tend to yield more errors.

Another advance within this field involves a variant of machine translation we’re already familiar with. Google created software which analyzes images and “translates” them into sentences. One of the examples used by Google to demonstrate this new system is that of “two pizzas sitting on top of a stove top oven,” a sentence generated from an image showing precisely that. The scientists at Google say they based their work off the recent advances in machine translation to come up with this software.

Lastly, another novelty that will likely exist in the near future is an application currently being developed also by Google, with which two people speaking different languages will be able to chat in real time. The application is being designed for tablets, with a keyboard at each end, where each person would write in their language and that content would be translated in real time using Google Translate and delivered to the other participant. The patent is already two years old and it’s not known when the app would be launched.

Meanwhile, more than just a few companies continue trying to overcome the huge current challenge in machine translation: achieving output text that sounds more natural and less robotic.