"Graffiti" for voice recognition?

[plragde]

Scientists at HP propose speaking in an artificial language to computers, since regular language is too difficult to recognize reliably. (I got this from an item in the Toronto Globe and Mail, which in turn took its information from the magazine New Scientist. It isn't NX or even Clie-specific, but I thought it might amuse you. --PR)

http://www.hpl.hp.com/techreports/2001/HPL-2001-182.html

[ballistic]

Computers should learn their master's speech/language (and handwriting for that matter), not the other way around. Are we becoming slaves to technology, or should we make technology work for us?

If cell phones can voice dial, why can't my PDA (NX70V) recognize a limited dictionary of voice commands/voice navigation?

Sounds like the engineers are once again taking the easy way out.

[cbulock]

Originally posted by ballistic
Computers should learn their master's speech/language (and handwriting for that matter), not the other way around. Are we becoming slaves to technology, or should we make technology work for us?

If cell phones can voice dial, why can't my PDA (NX70V) recognize a limited dictionary of voice commands/voice navigation?

Sounds like the engineers are once again taking the easy way out.

I think you have a point, but right now we are still progressing, and I think something like this is a good way to get started toward full voice recogniction. Much like Grafitti was designed to make it easier for the computer to understand handwriting, the same should be done for voice. Now we have faster processors, so things like natural handwriting recognition is now possilble, a few years from now, processors will be fast enough for voice recognition. But it will require a lot of power and the current 200-400 MHz ARM processors probably couldn't handle it.

[ballistic]

I'd rather have limited voice recognition functionality (smaller dictionary of terms that WILL be recognized) than to learn some bizarre 'pidgin' computer language.

How about using the voice memo feature of the NX to import dictated notes/emails etc into a desktop app such as Dragon Naturally Speaking? That way, you'd be using your desktop's superior processing power to do the voice recognition.

I'd like to see this type of productive desktop<<>>NX integration.

[ballistic]

Originally posted by cbulock


I think you have a point, but right now we are still progressing, and I think something like this is a good way to get started toward full voice recogniction. Much like Grafitti was designed to make it easier for the computer to understand handwriting, the same should be done for voice. Now we have faster processors, so things like natural handwriting recognition is now possilble, a few years from now, processors will be fast enough for voice recognition. But it will require a lot of power and the current 200-400 MHz ARM processors probably couldn't handle it.

BTW, as far as natural handwriting recognition, the Newton OS 2.1 had it 5 years ago. I recently got a MP2100 and I was amazed at how good the HWR is. I also realized how unnatural Graffiti is, even though it is 'second nature' to me from using it for the past 5+ years. After a short time using my new/old Newton, I am now craving that type of HWR for my NX.

The Newton does it with a 161MHz processor and 4MB of 'heap' memory. The NX's processing power easily surpasses the Newt's, so natural HWR on the Palm OS is now overdue.

[cbulock]

Originally posted by ballistic


BTW, as far as natural handwriting recognition, the Newton OS 2.1 had it 5 years ago. I recently got a MP2100 and I was amazed at how good the HWR is. I also realized how unnatural Graffiti is, even though it is 'second nature' to me from using it for the past 5+ years. After a short time using my new/old Newton, I am now craving that type of HWR for my NX.

The Newton does it with a 161MHz processor and 4MB of 'heap' memory. The NX's processing power easily surpasses the Newt's, so natural HWR on the Palm OS is now overdue.

Yea, but it seems the Newtons had quite a reputation for their handwriting recognition. I hear the MP2100 has great recognition though. But I like that once you learn Grafitti, you can achive almost perfect recognition, and I think it would be cool to have the same thing for voice recognition.

[ballistic]

I'm sure some users like yourself wouldn't mind learning a new language that only their PDA recognizes. Not for me, though.

As far as learning Graffiti, it looks like the original Graffiti is going to be dropped from any future Palm OS, so you'll have to keep your older OS version and older devices in order to use this proprietary PDA input method. Now you'll have to learn a new style of writing in Graffiti 2.

Do you see my point? Slave to technology. PDAs should be intuitive to use, and should not require you to learn their preferred input method.

[Molerat]

Bravo ballistic.

Even if you want to avoid the "slave to technology" question, market trends shove it through the backdoor. If Palm did adopt a pseudo-language scheme for voice recognition, and the idea caught traction, we would be faced with the question. After all, it is not the *best* technology that wins, rather the one with the most momentum.

Computers still have floppy drives, we still use QWERTY keyboards, and the standard GUI model has not changed (much) in twenty years.

If a "Grafitti language" is adopted, it may not be so easy to "get rid of" down the line. Whether by choice or design, we run the risk of becoming slaves to a particular technology.

I believe in a phased approach. Design speech recognition with scalability, and flexibility in mind. Design for the future, not the profit margin.