|Re: Calculator natural language user interface |
Message #12 Posted by Donald Williams on 7 Jan 2012, 12:29 p.m.,
in response to message #11 by peacecalc
"The technical one: the spoken word has to be transformed into a signal which can be handled by the machine."
Nothing trivial about this problem, but I would argue that we are well along the path to a solution.
Example: I never use the on screen keyboard on my Garmin Nuvi GPS. It is much easier and simpler to use voice commands, than to attempt keyboard entry.
"The semantical one: is the signal for the machine that, what it should be: a certain command in a distinct situation."
Agreed. That is the difficult problem. The lexicon of a 4 banger calculator would be completely manageable. The lexicon of a 50G would be daunting.
Example: I think the Garmin GPS minimizes that problem by using voice prompts. Knowing the context created by the voice prompts it can then select the appropriate lexicon for the next command. I don't think you could implement a calculator that way and I don't know how to attempt to solve the dilemma.
"The anthropogenical one: In which ways I'm transforming myself in using such a machine? My speaking, my behavior and last not least my cognition of problems."
The device will transform the individual.
Example: I bought that Garmin GPS many years ago when it was "new" technology. I was not interested in voice recognition. I thought it was simply a marketing gimmick. It turned out that the worst feature of the device was the on screen keyboard. There is so much latency with key press response that it was frustrating to use, so I gave it to my wife expecting her to be disappointed also. In fact she commented often how pleased she was with the device. Her secret was that she never used the keyboard and now I realize that I never will either. The voice response is an order of magnitude faster than the keyboard even if it had a perfect keyboard.
OT: Since I introduced the topic with a comment about SIRI let me make another observation. The thing that impresses me the most about SIRI is not what SIRI understands, but what SIRI does when it does not understand. It is my first experience with a man machine interface that does not respond with "invalid input" and then just quit.
Edited: 7 Jan 2012, 1:39 p.m.