Imagine for a minute that you could visualize all the media in the world and everything being said about that media -- in real time. You would get a picture of engagement with media and its impact, which words and images caused someone to do something -- in this case, talk about it online.

A growing number of consumer brands and media companies are exploring this bleeding edge of data collection and analysis with a startup led by MIT professor Deb Roy. Mr. Roy directs the Cognitive Machines group at MIT Media Lab, where he studies how children learn language. He presented his research in a widely viewed TED talk earlier this year, in which he showed how he recorded every moment of his family's life to find out exactly how his infant child learned to speak.

But his research partner, Michael Fleischman, suggested there could be a business application: What if you recorded the world's TV -- or at least 43-odd channels of it -- and matched the programs and the ads to things that are being said about them online? That idea became Bluefin Labs, a startup backed by a host of venture firms including Redpoint Ventures, onetime Huffington Post backer Ken Lerer, and a generous grant from the National Science Foundation.

What Bluefin is doing is pulling in live TV on satellite dishes, fingerprinting it to identify the ads and shows, and matching that with social-media chatter. That chatter need not specifically mention the show or the ad itself. The more the machines at Bluefin Labs listen, the better they refine their semantic technology to know what people are talking about.

Bluefin Labs' Tom Thai, VP-marketing and business development, will be speaking at the Creativity and Technology conference in New York on June 9.

Visit the CaT website to buy tickets.

"People have persistent, public screen names and we listen to them over time," Mr. Roy said. "It's like following a group of your friends, except that we're scaling it up to do it across everyone."

After nearly a year of crunching the numbers at the rate of 2 million minutes of TV and 3 billion online comments a month, Mr. Roy has already drawn some pretty striking conclusions. It turns out there are about 20 million people in the U.S. -- identified by their public profile names -- who routinely communicate about TV in social media; 4.2 million did so in April. Some have outsize influence over the conversation -- and presumably the opinions of the millions upon millions more who are listening.

The media and marketing industry is just getting its head around how to use this information. Brands like Best Buy, Mars and Humana, as well as agencies like Wieden & Kennedy, Razorfish and Dentsu are using the data to understand which shows and timeslots, as well as which creative, generates the most response. TV networks like Fox Sports are using it to guide programming decisions.

At least one very large packaged-goods marketer -- not named here -- is using the data to guide its buying at this year's network upfronts.

What you get is a measure beyond Nielsen ratings , surveys and focus groups to find out which programs and ads triggered a response in the form of online communication, as well as when and how that communication occurred.

If, say, the goal of campaign was to start online conversation, the correlation is direct. Conversation has an amplification effect, obviously, but it also signals the impact of the media on the consumer.

"A piece of media will make an impression on a cognitive level; you may or may not encode it in memory," Mr. Roy said. "In rare cases it is an impression strong enough to vocalize. More and more, the vocalization is being set free in social forums, and that 's what we're tapping into."