The social giant Facebook is one of the not so many companies that extensively use AI to do some of the behind-the-scene works.
From fighting fake news to creating a new language on their own, Facebook's AIs have done quite a lot of things. Another attempt in using AI, is to use them to help create a better translator, enhancing user-experience.
Facebook's AI research team have turned translation services over to AI, completely. In a blog post, Facebook's Applied Machine Learning team said that they've been training the AI to better understand how humans say things like slangs, do typos and others, in order to provide a much more accurate translations.
The neural-network Facebook uses is called a convolutional neural-network (CNN). The more common approach had been to use a recurrent neural-network.
While the strategy to use CNN for translation has not been that good, Facebook's approach to the technology is because a CNN can process all the data at once, while recurrent neural-network can only process things in order.
Facebook's strategy in using CNN is obviously processor intensive. But with modern computers, the company don't see thing as a major drawback.
As a result, Facebook's translations became noticeably better than before. For example, it can translate word-for-word and tries to find a better approximation for intent. Even if a word doesn’t exist from one language to the next, the translator can create some words on its own to present a readable translation.
This is a bit of contrast that before when the translator could sometimes create nonsense sentences.
Facebook is the largest and the most populous social media network on the internet. Connecting more than 2 billion users, it's obvious that its users speak different languages.
But having them all to gather into one social sphere, language can be an issue. It can be a border for communication.
This is where Facebook wants to make communication effortless. So even when people who don't speak the same language share one thing to another, communication should flow as it should normally, beyond dialects or any boundaries.