Facebook has been dabbling in the AI space for a while now, announcing things like intelligent chat bots for Messenger. The social networking giant has now announced a brand new venture dubbed DeepText which takes machine learning to a whole new level.
Facebook describes DeepText as a learning-based text understanding engine which can apparently understand the content of thousands of posts per second with near-human accuracy. The technology isn’t just restricted to English either, with the company claiming it can analyze over 20 different languages right down to each tongue’s slang terms.
DeepText is intended to be used as a way to better understand the context of the millions of messages and posts which Facebook has to deal with everyday. The idea might turn privacy hounds off since it basically means the site is reading everyone’s data and using it to better target them for ads.
Facebook is already putting DeepText through its paces, demonstrating its abilities through its intent detection and entity extraction skills. For instance, if a user says they need a ride in Messenger it’ll instantly suggest an Uber. However, if they say they just came out of a taxi, the technology is supposed to understand that the person is not looking to make travel arrangements.
The concept extends to Facebook posts as well. If a member writes about wanting to sell their bike for $200, DeepText is supposed to detect that the missive is about vending something. It’ll then take out relevant information like the item being put for sale and the price and automatically suggest ways the seller can utilize tools to make the entire process more convenient.
Facebook is also planning to take advantage of DeepText to battle spam in crowded public posts. The technology is still at its nascent stage though, so it’ll take some time before the site can sink its teeth into all your social interactions.