Real time speech translation, speech to text conversion, story summarization: all of these were “just around the corner” when I was a CMU grad student. I remember reading Dr. Raj Reddy’s proposal for a speech understanding system, what later became the Hearsay I project I believe. This was all going to happen in five years or so. By the time they developed Hearsay II his research group had a DEC PDP-10 (the cmub) pretty much to themselves. All the rest of us had to make do on the cmua.
That was in 1973.
So here we are, thirty years on, and it appears the real thing may really, finally be “five years in the future”. Some of the key elements are actually working under field conditions. It has always been inevitable we’d crack the speech understanding problem… eventually. It just took a couple years longer than we thought. [There were even more optimistic thoughts in the early 50’s, but that was computing before my time. Vacuum tube days. I think Fred Flintstone worked on the project.]
So here are a few very readable documents on the current state of military applications. The Phrasealator has been tested in Afghanistan by the guy who built it. The following are pdf documents. Right click and download.
- “An SBIR Success Story”, James Bass (script)
- “An SBIR Success Story”, James Bass (slides)
- “Human Language Technology TIDES/EARS/Babylon”, Charles Wayne (script)
- “Human Language Technology TIDES/EARS/Babylon”, Charles Wayne (slides) (large-ish)
PS: Note the military dune buggy at Kandahar Airport in James Bass’ slides. I want one!
