"Voices Carry: One More Paradigm Shift, and You'll be Able to Tell Your Computer Everything," by Matthew Steinglass. Lingua Franca, July/August 2000.
In this article, the author discusses the history and future of voice-recognition techniques. Most voice-recognition software today uses what are called hidden Markov models, the dominant technique in voice-recognition since the 1970s. While noting that computers are still decidedly inferior to humans in deciphering oral communication, Steinglass states that some experts think that we are due for a breakthrough that will make voice the prevalent means of communication with computers within the next five years. Recently, neural nets have been used and may provide the hoped-for breakthrough.
--- Michael Breen