Just happened upon this Old post from 12-8-98...Philip had a pretty good vision on the future of this technology.
From: pgl53
Date: Tue Dec 8, 1998 1:44 pm
The point here is that the software for speech
recognition has failed theis past year, part due to poor
performance of these programs, and part due to
overestimation of the market ( PC ) for them in the first place.
The conclusion is correct, that the future will lie
in embedded products, whether it'll be lhspf, or
fonix, or TI or LU etc is not clear today. Fonix
altered their business plan a year or so ago to supplant
the idea of a shrink wrap software product, in this
they were correct. This strategic decision led
them to the tactical issue of immediate refinancing of
their company to pursue their 'suite of
human/computer interaction products.' This has been costly, led
to dilution and erosion of share holder value;
however, to have proceeded otherwise would have been
to guarantee failure. Additionally, it it
now apparent that the HMM approach will not make it.
The future will require some sort of embedded and
compressed approach, probably employing extreme compression
which fonix has. The role for NN technology will expand
greatly in the next year.Where will fonix end up in
this, that will depend upon their marketing and the
effects of the financing. However, to dismiss speech
technology as a commodity misses the picture entirely. It
will be a commodity as with apple/talk which isn't
used and with IBM and Phillips technology and
lhspf all of which are failing in the market place due
to their ineffectiveness and poor quality. Whatever
company gets it to really work, will have added
value, not commodity value. |
