Siri, are you afraid of Facebook's new speech recognition tech?
Zuck & Co gain mobile speech APIs with Wit.ai gobble
Facebook is buying speech recognition startup Wit.ai, in an apparent bid to counter rival chatterbox tech from the likes of Apple, Google, and Microsoft.
"Facebook's mission is to connect everyone and build amazing experiences for the over 1.3 billion people on the platform – technology that understands natural language is a big part of that, and we think we can help," Wit.ai said in a blog post on Monday.
Wit.ai's offering, in a nutshell, is speech recognition as a service. The company was cofounded in 2013 by Alex Lebrun, a former vice president at voice recognition heavyweights Nuance, and Willy Blandin, a specialist in machine learning and natural language processing.
Developers use the firm's online console to define examples of the type of voice commands they expect users of their applications to say, along with the associated "intents" – signals that are tripped when the Wit.ai speech recognition engine recognizes the appropriate phrases.
Developers then code their apps to capture audio of the user's voice and upload it to Wit.ai's servers, which process it and return structured data based on what the recognition engine thinks it heard. When the application gets a message indicating a certain intent was requested, it can take the appropriate action.
The engine can even learn as users send in more input from a given application, offering new possible phrasings for given intents. If the engine guesses wrong, developers can go in and retrain it or add new intents as necessary.
It's not hard to guess why this tech would be attractive to Facebook, which so far has trailed behind its rivals in the speech recognition department. Apple's Siri personal assistant has been a major selling point of recent iThings, and Microsoft has been trying to make similar waves with its own Cortana. Google, meanwhile, has been busy baking speech tech into everything it can think of.
Speech recognition is especially important for the mobile market, where talking to your phone is often easier than typing in long messages using fiddly onscreen keyboards. And given that most of Facebook's daily and monthly active users are now accessing the service via their mobile devices – and 59 per cent of the social network's ad revenue now comes from mobile – speech recognition is an area that it desperately needs to address.
The terms of the deal were not disclosed, but it could be a big payday for Wit.ai's early investors. The Palo Alto, California–based firm raised its initial $3m of funding as recently October 2014, in a seed round led by Silicon Valley venture capitalists Andreessen Horowitz.
How Wit.ai's current executive team will merge into Facebook was also not explained.
One important change was mentioned, however. Currently, the Wit.ai service is free for "open data projects," where developers allow others to reuse their intents and expressions. Commercial users have to pay, however, with pricing as high as $1,499 per month. On Monday, the Wit.ai team said that under Facebook, "The platform will remain open and become entirely free for everyone." ®