preview

Case Study On Neural Machine Translation

Decent Essays

Facebook partnered with Bing in 2011 for translation services, making it an early pioneer in the move toward neural machine translation.
In December 2015, Facebook dropped Bing and began to outsource its translation services while developing its own translation technologies.
Facebook's own neural machine translator went fully operational on August 3, 2017.
Overview

Hello and thank you for your question about Facebook neural machine translation. The short version is that Facebook has been moving toward integrating their own translation technology since 2011, and have recently (August 2017) gone fully operational with their own Neural Machine Translator. Below you will find a deep dive of our findings.

METHODOLOGY

Our goal for this …show more content…

Facebook was an early pioneer of online translation when in 2011 it started using automated systems to translate users’ posts and comments in the News Feed. Facebook used Bing initially because they didn't yet have their own technology.

In January 2015, Facebook open-sourced Torch (their deep learning library). Facebook explains TORCH as an open source development environment for numerics, machine learning, and computer vision. Many projects on machine learning and AI at FAIR (Facebook AI Research) use Torch.

Also in 2015, Facebook acquired Wit.ai, a startup using understanding of natural language in text/voice to power user interfaces.

In December 2015, Facebook dropped Bing and started using its own translation technology. The issue was that Bing was built to translate more properly written website text, not human slang. Facebook stated that the disconnect between the languages people speak and the content they want to connect to on the Internet is what made them want to create their own neural network-based machine translation (MT) system.

Also in Dec 2015, Facebook opensourced it's hardware designs for anyone to explore through the Open Compute Project. The server was called BigSur, which is described as servers with graphics processing units (GPUs), which have become the chip of choice for deep learning.

Later, in June 2016, Facebook stated that Statistical

Get Access