A new system under development at MIT could reduce the cost and time needed to develop a speech-recognition system by analyzing the correspondence between images and spoken descriptions of those images.

>> Source: Computer Power User

New System Could Lead To Fully Automated Speech Recognition