The world is big and the world is diverse. In terms of socializing on the web, Facebook should know this better than anyone else. In fact, out of its over 1.5 billion users, less than half that number speaks English. So to speak, the rest consists of dozens of other languages.
For a social media that is meant for everyone, this case could evolve to make Facebook a place for "asocial". Facebook wants to change that.
To help users in reaching more global audience, the social giant is rolling out a tool that allows users to distribute posts in many different languages. Call this feature the Google Translate of social media, and Facebook wants its users to be able to translate posts to other available 44 languages.
This opt-in service is powered by Microsoft Bing and works on individual posts on Facebook Pages, including comments. The method is a semi-effective way, and Facebook is still testing the "multilingual composer". Though the initial tests were limited, the aim is to reach far off to the point that everyone on Earth can readily talk to everyone else.
"That's why I came to Facebook," said Necip Fazil Ayan, who oversees the company's translation efforts. "That's my personal agenda.'
Facebook's translation tool can give a whole meaning to social media. More than 5,000 businesses and celebrities have been using Facebook to reach their audience with more than 10,000 posts in multiple languages, all to be viewed 70 million times a day by their fans and followers. And when talking about numbers, more than a third of the time, they're viewed by foreign language speakers.
Ayan gave an example of an international footballers like Ronaldinho. The Brazilian star uses composer to post not only in Portuguese, but also Spanish and English.
"I only see the English," said Ayan. Now, millions of others can post in the same way.
Facebook's language composer is designed by Ayan and his team. It's specifically created for people with multilingual audience. Besides that, the translation can even provide its own translation. But its goal is to automate translation entire time, to everyone on Facebook.
The project is more like Google's Translate, but more focused on an area where Facebook has total control. While machine translation is still far from perfect, they're indeed improving. With Facebook being able to translate a language to other 44 available languages, the method of translation still uses traditional algorithmic models that rely on language statistics. But in some places, the company is leaning on its deep neural networks - networks of hardware and software inspired by neurons in the human brain.
So although far from perfect, Ayan said that there are noticeable improvement - especially in recent years.
For example, deep neural networks have been proven to enormously adept at learning certain tasks, such as recognizing photos and identifying spoken words, by just analyzing a huge amount of digital data.
This method also improves machine translation and natural language understanding. Using neural networking and Artificial Intelligence (AI), machines can get a whole meaning of words and sentences it translates.
Besides Google Translate, there are others that use machine learning to translate words they see. Microsoft's Skype is one example of a service that depends on neural networks. According to Joseph Sirosh, the person who oversees Microsoft's cloud computing, the technology in Skype Translator is improving by much, to an extent that it's not moving into Microsoft translation as well.
Though neural networking and AI for translation is still far from perfect. But researchers agree that we're on the right path towards that goal.
But in order to make that future happens, we need a lot more data. A lot here means many many more data. Neural networks and AI thrive and live on data - the more they can feed on, the more they're expected to become smarter.
Facebook's multilingual composer plays a role here because people can edit translation and add their own. This small feature means a big thing for Facebook since it can generate a lot more data. Its AI can learn a lot from users' suggestion, and any correction is just another input for Facebook's algorithms to understand.
If enough other users vote positively on the accuracy of that human translation correction, it will replace the one from Bing each time the Translate button is clicked. The human translations can be managed by page administrators using the "manage translations" link below posts on Pages they manage.
With all the languages outside the 45 where the company already does translation. "From Catalan to Turkish? We don’t have enough data," said Ayan. "So this too will help with our overall mission: breaking language barriers."
