Expect Labs

Expect Labs Taps Nuance’s Voice Recognition Tech In A Bid For Better Conversation Analysis

Next Story

Koozoo Raises $2.5M Seed Round Led By NEA And Tugboat To Cover The World With 24-Hour Mobile Video

San Francisco-based Expect Labs has been keeping awfully quiet since it announced that it raised $2.4 million in funding from Google Ventures and Greylock Partners back in October, but that silent streak ends today.

The team has just shed a little more light on how its vaunted Anticipatory Computing Engine will work — Expect Labs has tapped Nuance and its impressive voice recognition technology to help them in their mission to analyze your conversations and surface relevant information in real-time.

While CEO/co-founder Tim Tuttle and his colleagues hope that the ACE will eventually find its way into a number of apps and services (thanks to a Developer Platform that was hinted at when news of its funding round broke), what really helped to spark people’s imagination was the impressive MindMeld iPad app demoed at TechCrunch Disrupt SF 2012. Though you’d be forgiven for thinking it to be a simple voice/video chat service at first glance, that’s only half of the equation here — MindMeld provides a running stream of information relevant to the conversation at hand by monitoring a user’s physical location and online social activity, as well as quietly listening to the goings-on.

If that sounds tough, well, you don’t know that half of it. The work begins well before Nuance’s voice recognition tech enters the picture — Expect Labs’s system works by continually listening to a conversation in progress and analyzing the raw sound data and segmenting it into little chunks. Those sound chunks that the system believes contains relevant information are then passed along to Nuance, which converts the input into text to be used as search queries.

It’s worth noting that the MindMeld iPad app still hasn’t been released yet, though I’m told the Nuance partnership had little bearing on that. According to Tuttle, the original plan was to release a lean version of the MindMeld app shortly after its TechCrunch Disrupt debut and to rapidly iterate after that. That all changed after a deluge of positive press forced the team to consider “new feedback” from outside sources and rethink that initial plan.

“We wanted to make the product a little bit better to make sure that it disappointed people less,” Tuttle told me.

Fair enough, Tim, but please hurry — I (and more than a few others, naturally) want to start playing with this as soon as humanly possible.