[ih] Historical Tracing from Concept to Reality over 5 decades?

[John Gilmore]

On 7/6/20 12:25 PM, Steve Crocker wrote:
> >                                                       From mid
> > 1971 to mid 194 I oversaw the first few years of the Speech
> > Understanding Program.  The goal was understanding connected speech
> > input in a constrained task environment.  1,000 word vocabulary,
> > standard American male broadcast speech in a clean environment.  No
> > requirement to operate in real time.  There are traceable tracks from
> > that work to today's Alexa, Siri, et al.

I had a friend who worked on IBM's speech recognition effort on 360/370
mainframes back in the 1970s.  She was based at IBM Research in New York
State.

For years, an easy test to see if a modern speech recognizer was derived
from their code or data structures was to see if it could correctly
recognize the word "Poughkeepsie"!

	John