Bob, don't disappear. There are plenty of things to discuss and I have an alternative viewpoint on some of the responses you make.
I know it may be simplistic but I have in mind that Microsoft may have wanted to simply buy control of fonix but there was no way management would let it. Also, considering some of the secrecy involved there may have been voices within Microsoft, those euphoric about the advances in HMM for example, who may have advised the negotiators not to bother.
Company's placing bets on LHSPF possibly borrow from the same issue. I have my doubts, but its possible that Microsoft and SGS Thompson still hav'nt clicked about what fonx tech is supposed to represent. In fact many of the participants at Speechtek in NYC that I spoke to knew only of fonix' existence but not about its methods so that ignorance was strikingly common and still is.
Here's a little comparison between the two technology's from my reading these pages. If you like a chocolate stick in your ice cream and your friend shows you one with two sticks, next time you get one, you ask for two sticks also.
HMM
1 User dependant [despite shorter training times]
2 Vocab limited [due to the memory used for patterns and time taken to access and select the correct pattern. This explains why 3 vocabs generally cover domains such as law, medicine, etc]
4 Accuracy limited by technology
5 Needs lots of RAM and Hard disk space
6 Needs lots of Processor power [to speedily process the matching]
7 Cannot be placed on a chip [is this true?]
8 Due to large amounts of processor, RAM and hard disk, the technology finds use in limited areas such as text generation
Despite very big advances, this remains a clumsy solution limited in scope.
Fonx
1 User independant [as clearly demo'd despite simplistic demo]
2 Vocab unlimited [not demonstrated since emphasis was on the technology]
3 Accuracy can be greater than HMM since there is a correcting/context algorythm
4,5,6 Less memory required on Hard disk due to no stored word patterns
I'm sure RAM use may be large also but the processor is less used since the technology can be its own processing chip.
Takes up less space and can be used in myriad applications on a chip with variable amounts of memory
8 Despite delay's in development this technology promises to be most flexible and compact.
If the fonx tech has these advantages, then it may well end up being the engine of choice for all of the competition who want to remain in competition. to paraphrase the BASF ad on TV "Fonx does'nt make the speech recognition system you use, it makes the speech recognition system you use better".
Since LHSPF and Dragon only concentrate on speech to text and vice versa, these relatively small markets may be the only ones they service and even then they will be hard pressed not to take up the ice cream with two chocolate sticks.
You are right about one general conclusion though. Investors are going to have to hope that this is truly outstanding technology [highly likely] AND that the management team can execute flawlessly. I dont think they have to execute flawlessly since they have'nt and are still on course to getting multiple deals on the table. I therefore think the odds are far more in favor of fonx than you have suggested.
Parx |
