SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : Fonix:Voice Recognition Product (FONX) -- Ignore unavailable to you. Want to Upgrade?


To: Mark Cox who wrote (1458)12/9/1997 7:05:00 AM
From: Dr. Bob  Read Replies (2) | Respond to of 3347
 
Parx,

You have outlined the differences between HMM and fonix, as fonix sees them; whether these contrasts will be as stark as you have outlined them or not is something time will tell. I'm certainly increasingly believing that HMM can do some of the things on the fonix list, things that it wasn't supposed to be able to do.

For example:

"7. Cannot be placed on a chip (is this true)" - I believe the announcement from LHSPF/SGS-Thomson stated that Thomson would be imbedding LHSPF technology into its chips for use in cars, cellular phones, etc.

"1. User dependant (sic)" - Certainly there are user-independent capabilities out there now with vocabularies the size that fonix has demonstrated - I run into them on automated phone systems on a regular basis. They require no training; work the first time I call that business, and navigate me through choices until I come to a real person - if I ever need to. Delta Airlines is the one I've used most frequently.

"2. Vocab limited..." - I hardly think, from what we have seen to this point, this can be considered a limitation of HMM and a strength of fonix!!

No #3 listed.

"4. Accuracy limited by technology." - Once again, from what we have seen so far (I'm recalling the person who went to the demo and recited the sentence they were given backwards, and fonix couldn't handle it) - hardly a strength of fonix so far!!

5,6, and 8. Becoming less and less an issue every month.

"...a clumsy solution limited in scope." Well, I thought so, too, last year, listening to the fonix party line - but I think the advances over the past year have made me much more uncertain about that, and if this next year brings the advances I think it will (Windows 98 with voice recognition, Thomson's imbedded product, and more), this appraisal of HMM may seem very inaccurate

Bob



To: Mark Cox who wrote (1458)12/9/1997 12:57:00 PM
From: Jinping Shi  Read Replies (1) | Respond to of 3347
 
Parx:

I agree with you. From a statistical modeling point of view,
neural network approach is much more powerful than hidden Markov
model. Basically, whatever one can do with HMM, one can do it with
neural network approach.

Jinping, longer