High school students develop better emotion detection algorithm


Speech recognition systems today have come a long way. It’s nowhere near the holy grail of natural language understanding, but good enough to apply in human-machine interactions such as the IVR. Now there’s even speech systems which will detect human emotions.

It’s actually very useful in the contact center application. For example, if the IVR can detect anger soon after a caller mutters some words, then the system can be designed to bypass the rest of the menus and automatically send the caller to a supervisor. Hopefully then the supervisor can immediately address the source of the anger. A much better experience for the caller — and for the tier 1 agents too, since nobody has to endure the time spent with this agitated caller. The escalation process is automatically initiated without agents transferring the caller around…

Two Oregon Episcopal School students decided that they can do better to come up with a new algorithm to detect emotion in speech. Well, they were right, and now Akash Krishnan, 16, and Matthew Fernandez, 17, will each be $50,000 richer after winning the grand prize award for their entry, “The Recognition of Emotion in Human Speech.”

Just how much better was their algorithm?

The team was inspired by the movie “I, Robot,” in which a robot can sense when its user is under stress. Using an emotional speech database with 18,215 files and five emotions (anger, positive, neutral, emphatic, rest), the team developed, trained and tested a classification engine to determine emotions from an input signal. The team achieved a 60 percent accuracy rate in identifying emotions versus 41 percent with a previous award-winning system. Their work has a broad range of applications from autism research to computer games and lie detection.

Dear Akash and Matthew: It’s never too early to create a LinkedIn profile. I, for one, cannot wait to see your work in real-world applications especially in the contact center. We all know there’s room for improvement in that area…

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s