Facebook is scanning everything you write
The social network has therefore built a new artificial intelligence (AI) program that is tasked with scanning ALL the text on the site . With over a billion users regularly posting content, Facebook has a lot of information to sift through.
The idea is that it will help catch spam and other unwanted content while, at the same time, generating the right results for search queries and requests.
Called ‘DeepText’, the AI can understand text in various different languages with ‘near-human accuracy’.
“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 .
The software is currently being tested with Facebook Messenger and has emerged from the company’s move into chatbot-based messaging .
It will soon move out across the entire network.
Read more: Facebook is testing ANOTHER big change to your news feed – but will you like it?
“Text understanding includes multiple tasks, such as general classification to determine what a post is about — basketball, for example — and recognition of entities, like the names of players, stats from a game, and other meaningful information,” the company said.
“But to get closer to how humans understand text, we need to teach the computer to understand things like slang and word-sense disambiguation. As an example, if someone says, “I like blackberry,” does that mean the fruit or the device?”
Eventually, Facebook says it will be able to use the software to improve your experiences on the site.
It would, for example, detect if you were posting about selling an old possession and offer built-in tools that might make the transaction easier to complete.
Or it could help remove spam posters automatically if you run a particular page or group.
It’s just the latest in a series of developments at the ever-growing social network to better understand how and why people use it.
Facebook is also believed to be experimenting with the idea of letting users customise their own news feeds .
Poll loading …