Comic Sans was An alternative history... 4302

Brian,

My point was that I actually HAVE the Nuance SR engine accepting pretty much unrestricted speech and utilizing "semantic granularity" to increase the accuracy sufficiently that it can then be fed directly into an AI program WITHOUT review or editing! Unfortunately, the folks at Nuance are SO convinced that what I am actually doing is absolutely impossible that without looking they have embarked in a direction that is INCOMPATIBLE with their EXISTING software, so to implement their proposed standard, they must go out of their way to disable capability that THEY NOW ALREADY HAVE! Hence, my point that they need to thin their ranks.

Further, their proposed standard can NOT reasonably handle interruptions, e.g. where the program you are using on the Internet momentarily pauses, so you say something during that pause. My alternative approach interrupts your interruption and says "One moment please." or whatever you wrote your program to say during an interruption. There are many other timing-dependent situations that must be handled to make unstructured conversational speech work well, NONE of which has been dealt with by the present Nuance people at W3.

The MAJOR underlying problem with the proposed standard is that ONLY things from the Internet can be spoken, so the entire system goes mute on all Internet delays. Further, the Internet is simply too slow to coordinate real-time activities like conversation. To operate a conversation, you need LOCAL control over timing issues, and their standard carefully EXCLUDES such local control.

backwards compatible with the existing IEEE-754 standard, and with the right word of configuration bits, it will emulate the IEEE-854 proposed decimal standard! Further, you can even emulate '854 with or without its obvious defects (e.g. NAN shortcomings)! In short, IBM has been working on solving just one problem, when there are literally dozens of outstanding problems with the existing FP to be solved.

Further, my proposal will run on EXISTING PC HARDWARE using EXISTING compilers and very minor modifications to existing programs, whereas IEEE-854 only shows how to build new hardware that will require more extensive changes to existing software.

There has been a deafening silence about decimal FP on emulate IEEE-854!

Yes, it IS amazing how a special interest will push a weak standard that only addresses a particular concern rather than a strong standard to solve ALL of the outstanding concerns INCLUDING theirs.

This is much the way that computers came to be SO screwed up as I describe in my article - one wrong turn at a time.

Steve Richfie1d