Interlingua machine translation

Hi everyone,

Everyday machine translations are getting better and better, but they still can’t compete with human translations for a number of reasons. One of these is that they often struggle with context, and so deciding which translation of a word that has many meanings to use is problematic.

However Google has been retraining its translation service to use a neural network¬†to analyse entire sentences rather than individual words, giving it more context to base its solutions on. It has also extended this capacity to cover languages that it hasn’t been trained to translate between directly:

“For example, if the neural network has been taught to translate between English and Japanese, and English and Korean, it can also translate between Japanese and Korean without first going through English. This capability may enable Google to quickly scale the system to translate between a large number of languages.”

Researchers think this happens through a process called zero-shot translation: the machine translator finds a common ground whereby sentences with the same meaning are represented in similar ways regardless of language. In other words, it creates an intermediate language that is not readable or usable by humans, but that facilitates the translation process.

What do you think about these developments in machine translation? Is it a tool that we can use to our advantage, or will it eventually kill the art of translation and put translators out of business? Should we be concerned that a computer has made up its own language that we can’t understand?

Leave a Reply