2 ms·

The whole - 'what do you believe that other people don't' The public can 100% manage and love terminal interfaces... as long as it looks like a text message.

Pxtl 5y ago
I've come full circle on that using voice assistants. Y'know - when you're a kid "terminal interfaces are bad, guis are great". When you get older and learn "Terminal interfaces are good for compsability and piping, they're great." Then you start developing software for users and learn "terminal interfaces have poor discoverability, GUIs present them their options in clear, informative ways". Then the promise of natural-language processing appears and it's "terminal interfaces are great, you don't have to be perfectly precise any more". Now, with my google assistant? Text interfaces are awful, because I have no idea what combination of magic words will confuse it. What will be interpreted as a parameter to which command. Even simple commands get broken if I accidentally add phrasing where it interprets it as a different command... usually just giving up with "I don't know how to do that" and then Google gets a nice recording of elaborate cusswords.
- bluGill 5y ago
  You probably can't do this with google (rhasspy on the other hand...), but what is really needed is every-time someone tries something that doesn't work they go program the response that would have been correct. I think speech recognition is good enough for this, but what is left is making a complex tree. See the 20 questions game where because so many people have put in new answers it now can guess almost everything after just a few questions. This is not easy and it takes a lot of effort to figure out where things didn't work and fix it so it will for the next person. At least some of those things you fix will never be asked again, but you need to do it. Part of the above is speech recognition needs to expose a confidence and if it thinks it knows ask useful questions about what you mean. (they already do in some cases)

Pxtl 5y ago

I've come full circle on that using voice assistants. Y'know - when you're a kid "terminal interfaces are bad, guis are great". When you get older and learn "Terminal interfaces are good for compsability and piping, they're great." Then you start developing software for users and learn "terminal interfaces have poor discoverability, GUIs present them their options in clear, informative ways". Then the promise of natural-language processing appears and it's "terminal interfaces are great, you don't have to be perfectly precise any more". Now, with my google assistant? Text interfaces are awful, because I have no idea what combination of magic words will confuse it. What will be interpreted as a parameter to which command. Even simple commands get broken if I accidentally add phrasing where it interprets it as a different command... usually just giving up with "I don't know how to do that" and then Google gets a nice recording of elaborate cusswords.

bluGill 5y ago
You probably can't do this with google (rhasspy on the other hand...), but what is really needed is every-time someone tries something that doesn't work they go program the response that would have been correct. I think speech recognition is good enough for this, but what is left is making a complex tree. See the 20 questions game where because so many people have put in new answers it now can guess almost everything after just a few questions. This is not easy and it takes a lot of effort to figure out where things didn't work and fix it so it will for the next person. At least some of those things you fix will never be asked again, but you need to do it. Part of the above is speech recognition needs to expose a confidence and if it thinks it knows ask useful questions about what you mean. (they already do in some cases)