For the most part, Google Translate is servicable when it comes to translating words. When it comes to sentences though, Translate can be…shall we say, inadequate for the job. Wanting to fix that, Google has said that they intend to improve upon Translate by utilising deep learning.
Deep learning involves training neural networks on lots of data and then having them make predictions about new data. Thus far, Google Maps, Photos and Gmail have all been enhanced with this technology. In fact, Google’s DeepMind AI and AlphaGo relies on this technology to defeat Lee Se-dol in Go. While Translate already uses deep learning for its visual translations, Google is aiming to fully implement deep learning into Google Translate.
In an interview at the Structure Data conference in San Francisco, head of the Google Brain team Jeff Dean said that his team has been working with Google’s translation team to scale out experiments with translation based on deep learning. Specifically, the work is based on the technology depicted in a 2014 paper entitled “Sequence to Sequence Learning with Neural Networks.”
The end goal for Dean is to make translations more accurate. Dean says that Google is aiming to make Translate to rely solely on the type of neural networks documented in the paper.
Source: VentureBeat
