IBM's speech recognition system now boasts a 5.5 percent error rate.
This milestone is an industry record for conversational speech recognition, according to IBM, representing a 20 percent improvement rate from the 6.9 percent error rate IBM reported six months ago. This error rate is based on human parity, or the error rate per word of two humans speaking. The benchmark for human parity is 5.1 percent error, according to research by IBM and Appen.
"These speech developments build on decades of research, and achieving speech recognition comparable to that of humans is a complex task,” said Michael Karasick, vice president of cognitive computing at IBM.
"This progress will have important implications for how man and machine collaborate in the future, making the interactions more natural and productive. We believe it is only a matter of time before we achieve parity on speech recognition with humans."