Individuals who get frustrated trying to navigate through the instructions from recorded phone answering systems may soon be getting some help thanks to software developers at the University of Southern California.
The researchers, from the USC Viterbi School of Engineering’s Integrated Media Systems Center, are developing a new software program that will recognise frustration in a user’s voice – then take action to help, such as providing a soothing computer response or immediately transferring the call to a human operator.
The system distinguishes irritated speech from normal speech with 85% accuracy, according to Shri Narayanan, associate professor of electrical engineering, computer science and linguistics.
Narayanan said that companies could his system to ensure that callers do not hang up out of frustration. A commercially viable system could be available in about two years, he said, as researchers continue to improve accuracy and add response capabilities.
The system identifies frustration from such features as pitch, energy, duration of speech sounds, word content and contextual information.
The team used some 1,400 real calls recorded by a call centre to ‘teach’ the system to recognise frustration.
‘The underlying voice recognition technology we are developing for this application will be useful in many other areas, including automated training, education and games,’ Narayanan said.