- NEW!
by branciforte3241 on September 10th, 2009
Alternately repeat the words "dist" and "disp" out loud, noticing how your lips move differently. They do not sound the same.
If they sound different to a human then they sound different to a computer. There's no difference in the sound waves received by either one.
How good is Cuppen's algorithm compared to other eigenvalue solvers?
by computational on February 4th, 2011
| 1 person likes this
what are great value paper towels made of ?
by justinbieberloverforever on March 16th, 2011
| 1 person likes this
Could so-called "internet worms" and "email viruses" be virtually all *Microsoft* poor, insecure and sloppy program designs?
by buttman on December 27th, 2010
| 1 person likes this
Any Uni in US offer Bachelor of sci major in both Computer Science and Software engineering?
by Shauni_K on January 19th, 2011
| 1 person likes this
Are both UDP and IP unreliable to the same degree?
by pearls92 on April 18th, 2011
| 1 person likes this
You're reading Is it true that computer voice-recognition programs will NEVER be perfected, because "Give me a new display" and "Give me a nudist play," sound EXACTLY THE SAME, phonetically, to a computer?
Your favorite Q&A site is on Facebook! Become a fan to connect with the ever-growing community of Answerbaggers, share your thoughts and find out what Answerbag's up to next!
Comments
Computer experts at both MIT & Stanford have already demonstrated that, unless a person uses specific diction (and the average computer user will not), voice-recognition programs will be unable to distinguish between the two phrases noted in my question, and that's just one of the problems plaguing such programs.
.
Another program that's in development, is one where the printed words appear on the monitor, as they are being spoken. The problem with that program, is that voice recognition protocols can't distinguish among words like "they're," "their," "there" or, "two," "to" and "too."
.
Talking about this reminded me of the "gorilla" exchange in the movie, "Captain Ron," where Kurt Russell's character tells Martin Short's to "Watch out for the guerillas!" Short replies, "Don't be silly. They're aren't any gorillas on this island!"
.
+6 for your insightful (but "flawed," respectfully) answer.
by Anonymous on September 11th, 2009
Not flawed. If a human can tell the difference it is because they sound different. If they sound different, then a computer will eventually be able to tell the difference. It's just that the current wave of voice recognition algorithms do not segment and feature-isolate the data in a way that highlight the differences you cite. I think you misunderstood what the MIT/Stanford people were referring to. I don't know what they are saying; I went to Carnegie Mellon, home of the Sphinx project.
But if you are convinced enough of their correctness that you would call my analysis "flawed", then why ask the question in the first place.
You should have asked the question using the example of "their"/"there". Now there's an example of a true homophone. The sound waves generated are the same. You distinguish the two would require the computer to do a level of processing many orders of magnitude more sophisticated. The computer would need to do a context sensitive analysis.
by branciforte3241 on September 11th, 2009
You didn't comment on the Guerilla/Gorilla "exchange" I mentioned.
.
My original comment stands, because I was not using my own words, I was quoting a "Silicon Valley" professional, who cited the research at the universities I mentioned, within the parameters of what was known at the time the statement was first uttered.
by Anonymous on September 11th, 2009