Wednesday, March 4, 1998 at 04:40:41
Fascinating subject. I wonder if you have thought about how this natural human tendency affects the development of high quality speech recognition software?
Yes, I have thought about how FPs (filled pauses) affect speech recognition software development, but in fact I am very late in coming to the field. In the early days of my research (last summer) I started putting together a bibliography. Almost all of the pages I could find on the web which addressed FPs were research reports by those involved in speech recognition technology. Many of these 'computational linguists' are studying how to predict and weed out such intrusions on the primary message. It's very interesting work, but at present there does not seem to be any prominent solution. So while the present state of speech recognition technology admits the software needs a great deal of training, it is apparent that the user also needs a great deal of training--to speak only in a manner the software can recognize. If you're interested in studying this topic further follow the links on the links page at the FILLED PAUSE RESEARCH CENTER.
This site is
maintained by Ralph L. Rose