Home Technology Facebook AI aims at learning what users post

Facebook AI aims at learning what users post

Facebook said it is building a new AI-based search engine called Deep Text to understand better everything users post on its social networking platform.

That makes for a massive venture considering the sheer scale of its social channels. About 400,000 new posts and 125,000 comments on public posts appear every minute, which might make the task of Deep Text seem a bit too daunting. As a result, the intelligent engine is already working towards it and able to deal with thousands of posts and comments each second, that too in 20 languages and with near human-like ability.

Of course, the central idea behind the entire venture is to develop a much better understanding of human behavior and act accordingly. For instance, if Deep Text knows what the user’s post is all about, it will be in a much better position to serve it to those who are likely to appreciate it the most.

And for that to happen, Deep Text is endowed with enough intelligence to understand both the meaning and the sentiment of what people share on Facebook. For instance, it must know what a statement like: “I like BlackBerry,” actually implies. Whether that applies to the fruit or the smartphone.

Similarly, if Deep Text can decipher a user’s post is about selling any commodity, it will automatically reformat the post to not only make it appeal more to prospective buyers but also strive to reach out to those who are on the lookout for such items. Also, of course, details like price and item details will also be added to allow buyers have all they need to make a decision.

Needless to say, Deep Text will have a bigger role to play in Messenger as a better understanding of user text. That, in turn, is also the manner Facebook intends to monetize the messaging platform ultimately without resorting to serving ads directly.

The way it will work is this – user messages like: “I need a ride” will invoke Deep Text to ask the user if he/she would prefer a ride to be arranged with Uber or such services that sign up with Messenger. Also, Deep Text will have to know if the user indeed wants a ride or is just mentioning ‘ride’ in some other context. In fact, Deep Text’s intelligent attributes spawn wider than that as it just can’t depend on a single term to guess what the user is actually implying.

See Also: Facebook to track and force feed people ads

As such, Deep Text will have to make out the difference between such texts like:

  • “I just got out of the taxi,”
  • “The particular journey is best made by a car,”
  • “I have been taken for a ride,” or simply
  • “I need a ride back home.”

Deep Text will have to know the exact scenario when the user is looking for hiring a cab and so on.
Another inherent benefit with Deep Text is that it will allow Facebook to monitor user’s activity better and get rid of hate texts or other abusing materials. It will be a more efficient means to accomplish the task as doing the same with human intervention would be more time taking.

Overall, it’s an exciting new development Facebook is attempting. However, there are also those who are already voicing privacy concerns with Deep Text. It will be interesting to see how things pan out in the end.