Social media giant Facebook on Thursday announced the development of a new artificial intelligence (AI) system called “DeepText” that can analyse what people are writing about in posts with near-human accuracy across 20 languages, reports IANS. Everything that you type and post could, in theory, be scanned by algorithms to help other users see your content and help advertisers target their messages to you.
Advertisement
Facebook says that they find at least 25 percent of their engineers now regularly making use of their internal AI platform to build features and for doing business. 400,000 new stories and 125,000 comments on public posts are shared every minute on Facebook. Since text plays a key role and is the most dominant form of communication on Facebook, “understanding various ways text is used” on the network is essential to how people react to the company’s services.
The same goes for finding and surfacing the most relevant or high-quality comments (in multiple languages) among exchanges with celebrities and public figures.
But by the time the content is seen and flagged as unpleasant and removed by Facebook, it may already have been seen by the people concerned and caused the damage.
The system can also be used to combat hate-speech while being powered by FBLearner Flow, and it’s also being actively used in the Facebook Bot platform where the NLP services come into play to deliver a seamless experience which is not limited to a particular language and can be scaled up to add more languages.
It said the AI system was developed to help people get more out of the site and to help catch spam and other unwanted messages.
Although introduced today, DeepText is already being tested across some Facebook properties, such as Messenger.
Advertisement
The company is also investigating new deep neural network architectures, it said, such as bidirectional recurrent neural nets (BRNNs). “As an example, if someone says, “I like blackberry”, does that mean the fruit or the device?” As for now, interpreting simple sentences is a piece of cake for smart systems; however, there is still time before DeepText can make out more the complex sentences such as “I need a pickup after 20”. One of the goals of DeepText is to understand more languages more quickly.
