Facebook has said that it is developing a new artificial intelligence (AI) system called ‘DeepText’ that is able to analyze what people are writing about in posts.
According to Facebook, the system can analyze several thousand posts per second and can offer near-human accuracy across 20 languages. It aims to understand the sentiments behind posts, and generate responses to certain search queries.
“DeepText leverages several deep neural network architectures, including convolutional and recurrent neural nets, and can perform word-level and character-level based learning,” Facebook said in a blog post.
“We use FbLearner Flow and Torch for model training. Trained models are served with a click of a button through the FBLearner Predictor platform, which provides a scalable and reliable model distribution infrastructure. Facebook engineers can easily build new DeepText models through the self-serve architecture that DeepText provides.”
Facebook says the system can “automatically pick out hate speech from status updates that merely contain controversial terms” and “could prevent a human from ever suffering the harassment and having to report it manually.”
According to the social networking giant, computers need to be taught about how to understand things like slangs and word-sense disambiguation “to get closer to how humans understand text.”
“The community on Facebook is truly global, so it’s important for DeepText to understand as many languages as possible…..Using deep learning, we can reduce the reliance on language-dependent knowledge, as the system can learn from text with no or little preprocessing. This helps us span multiple languages quickly, with minimal engineering effort,” the blog post reveals.
Facebook says it has already been testing its deep learning-based text understanding engine on Facebook Messenger.
If the system identifies a sentence indicating that the user needs a ride, it will suggest using Messenger transportation integrations with services like Uber and Lyft. Similarly, if someone writes about having something to sell, the Deep Text-based bots would try to collect details of what is being sold and then suggest using Facebook’s sales tools to let ad reach a wide audience.
Facebook plans to use this ability to recommend better services to people and improve people’s experiences. The system will also help people catch spam and other unwanted messages.
The Success Story of Facebook
The Facebook website was launched in 2004 by Mark Zuckerberg, along with fellow Harvard College students Andrew McCollum, Eduardo Saverin, Chris Hughes, and Dustin Moskovitz. Initially, the websites membership was limited to Harvard students, but it was later expanded to include higher education institutions in the Boston area, the Ivy League schools, and Stanford University. Since 2006, anyone aged 13 or above can register on the website.
By late 2007, there were 100,000 business pages on Facebook. These pages were created by companies to promote themselves and attract customers. On October 24, 2007, Microsoft revealed to have purchased a 1.6% share of Facebook for $240 million. Once year later, Facebook announced its plan to set up its international headquarters in Dublin, Ireland.
In July 2010, the company announced to have touched the milestone of 500 million users, making it the world’s largest online social network at the time. At that time, almost half of the Facebook users were using the site daily, for an average of 34 minutes (according to the company’s data).
According to SecondMarket Inc., Facebook’s value in November 2010 was $41 billion, making it the third largest American web company after Google and Amazon.
The initial public offering of Facebook was held in February 2012. Three months later, the company started selling stock to the public, and reached an original peak market capitalization of $104 billion. As of March 31, 2016, the social network had over 1.65 billion monthly active users, and it is now also the world’s most popular social networking site based on the number of active user accounts.
You must be logged in to post a comment.