Select your localized edition:

Close ×

More Ways to Connect

Discover one of our 28 local entrepreneurial communities »

Be the first to know as we launch in new countries and markets around the globe.

Interested in bringing MIT Technology Review to your local market?

MIT Technology ReviewMIT Technology Review - logo


Unsupported browser: Your browser does not meet modern web standards. See how it scores »

{ action.text }

The Intel study analyzed 24 University of Washington graduate students’ conversations over the course of a school year. The idea was to capture a week’s worth of data each month for nine months so that the researchers could have day-to-day interactions and snapshots of conversations over time. In order to make sense of all the data, the researchers collected extra information from the participants, using surveys. But since the conversations were spontaneous, it wasn’t possible to do so for all the collected conversations. So Choudhury’s team gathered a smaller data set from five people whose words were recorded by the software during their conversation; data from this set was used as a benchmark to measure conversation features. The results will be presented at Interspeech 2007, a conference held later this month in Antwerp, Belgium.

Choudhury says that it would also be fairly straightforward to integrate the researchers’ software into a mobile device so that it could infer a person’s “presence,” a concept used in instant messaging to denote whether or not a person is available to talk. “There’s been a lot of work showing that interruptability or situational awareness in our devices can be very useful,” she says. And people are becoming more comfortable with the idea of broadcasting information about themselves to the world, as evidenced by the popularity of the social-networking platform Facebook and microblogging tools such as Twitter and Jaiku.

Still, Intel doesn’t have immediate plans to put the software in mobile devices, and there are still some technical hurdles to overcome in order to do so. Choudhury says that her software is significantly more precise than previous software for detecting and segmenting conversations, and it’s able to detect a speaker in a conversation with more than 90 percent precision. But the accuracy of other aspects of the system–such as inferring mood–is still being determined. And no one has tested how these features would be practically integrated into a mobile device. Indeed, an individual user might need to set up rules for her phone, allowing people to contact her only when she’s in certain situations. In addition, most phones couldn’t yet analyze all the data in real time. Today’s phones, Choudhury says, could reasonably determine the structure of a conversation–when a person is talking and for how long–but more detailed analysis would require more processing power than is currently available in phones.

0 comments about this story. Start the discussion »

Credit: Beverly Harrison, Intel

Tagged: Computing, software, Intel, mobile phones, language

Reprints and Permissions | Send feedback to the editor

From the Archives


Introducing MIT Technology Review Insider.

Already a Magazine subscriber?

You're automatically an Insider. It's easy to activate or upgrade your account.

Activate Your Account

Become an Insider

It's the new way to subscribe. Get even more of the tech news, research, and discoveries you crave.

Sign Up

Learn More

Find out why MIT Technology Review Insider is for you and explore your options.

Show Me