Contact: Eric Mankin
University of Southern California
Computers learn to understand kids
Agent chimp understands what kids say
Creating a system that lets children talk to computers instead of using mouse or keyboard controls won a "Best Paper" Award for the USC Viterbi School of Engineering's Shrikanth Narayanan and a collaborator.
Narayanan based the paper on research done in 2000-01 with his co-author, Alexandros Potamianos of the University of Crete.
The scientists had to solve several problems to make the system work.
First, children's voices are very different from the voices of adults. The sound patterns aren't like those of adults, and a group of children will say the same words in more varied ways than a group of adults.
Programs that let computers understand people talking are called Automatic Speech Recognition (ASR) programs. Regular ASR programs make four times as many mistakes when they listen to children's speech than when they listen to adults'.
But the scientists were able to make changes and make a new ASR system that can understand kids as well as the old one can understand adults.
But do kids like talking to computers? The next part of the study was a controlled "Wizard of Oz" setup in which children played a well-known educational game called "Where in the USA is Carmen Sandiego?"
Half of the children used the standard mouse and keyboard techniques. The other half played by talking to the computer and saying what they wanted the computer to do. A human listening in ("the Wizard) controlled the computer for the child.
When the children were asked afterward on how they liked playing using voice versus mouse, an overwhelming number loved it. The score 5 meant best, and ninety-three percent rated it interface 4 or 5.
The last part of the paper describes how the researchers built a simple game using their child ASR. It was a spelling game with a character called Agent Chimp.
It wasn't a complicated game, but when a group 8-14 year children played it, they liked it.
You can't buy Agent Chimp yet, but the scientists have recently received a new grant from the National Science Foundation (which had also funded the earlier work) that will help them improve this game and create others.###