This is the forum archive of Homey. For more information about Homey, visit the Official Homey website.

The Homey Community has been moved to https://community.athom.com.

This forum is now read-only for archive purposes.

Usable speech recognition

124»

Comments

  • Am I the only one who was disappointed by this week's Software Update?

    What's next on our roadmap?
    In no particular order, this what we'll be working on next:
    • Insights v2 + Insights Reports
    • Bluetooth Audio speaker support
    • Smartphone App improvements
    • Google Assistant integration
    • Dashboards
    • Apps, apps and more apps
    Enjoy your weekends!



    Homey's "Microphone Performance" or "Improving Speech to Text" (which was on the roadmap last year) seems to have vanished completely, and instead, we now get Alexa and Google Assistant integration (mind you: I think Alexa and Google Assistant integration is great, but not INSTEAD OF improving Homey's own voice recognition!)
    Whatever happened to the "We've also played with Google's Speech API. It looks promising to increase the Text to Speech accuracy..." ? For me, several of the examples mentioned on Athom's Speech page simply don't work, because 6 out of 10 times, Homey's voice recognition won't recognize such sentences correctly.

  • Bumblez said:
    Am I the only one who was disappointed by this week's Software Update?

    No ... 
  • @bumblez indeed, you are completly right. but we can't do anything tp push athom. so there is nothing more than hope that homey will have its own brilliant speech recognation.

    for now it works 0% for me. even the trigger "okay homey" (worked before 1.0.1) isn't recognized here at the moment.
    so i wolud go to amazon dot if athom would say: ok guys, we can't get it right, sorry. you have to use amazon or google for speech. but they say that they will polish the speech recognation of homey, but don't say when!!

      
  • phil_s said:
    @bumblez indeed, you are completly right. but we can't do anything tp push athom. so there is nothing more than hope that homey will have its own brilliant speech recognation.

    for now it works 0% for me. even the trigger "okay homey" (worked before 1.0.1) isn't recognized here at the moment.
    so i wolud go to amazon dot if athom would say: ok guys, we can't get it right, sorry. you have to use amazon or google for speech. but they say that they will polish the speech recognation of homey, but don't say when!!

      
    Same for me, honestly I would not even mind if Athom would go that route
    my only requirement would be the ability to DIY e.g.  the abiity to hack something together my(/our) selves
       
  • It would already help a lot if Athom were honest about what we can and cannot expect...
  • honeyhoney Member
    phil_s said:

    so i wolud go to amazon dot if athom would say: ok guys, we can't get it right, sorry. you have to use amazon or google for speech. 
    In that case they should provide discount from the price of Homey. Me personally I don`t like the Alexa, tell Homey homey solution. Alexa can work with other systems without telling to tell. Hey honey can you tell Alexa to tell homey to turn the lights orange.

    Bumblez said:
    For me, several of the examples mentioned on Athom's Speech page simply don't work, because 6 out of 10 times, Homey's voice recognition won't recognize such sentences correctly.
    That`s not that bad, mine is much worse. They have changed the parser recently. Since homey pretends more often that it understood the question but it does something else. 

    Oh about the Google API, they scrapped that because `it takes much time` to implement it. What a rubbish! I can`t really tolerate when they think that I am stupid.

    Speech related GitHub issues are ignored or just closed. So keep hoping.
  • casedacaseda Member
    edited March 2017
    They won't say that it's going to keep working bad, cause it won't. 

    They already improved voice handling in 1.2.x and are working constantly(!) at the server side to increase responsiveness and accuracy.

    Voice is hard to filter right from backgrounds noises, it is like listening to a whisper voice from 1 meter away when a car drives by in a tunnel (echo), I would like you to show me you can do that. 

    And if you can, create a filter for it, on your own (yes,  there is only 1 person working on voice, and it is not the only thing he is working on) within 1 year.
    Even if they went the Alexa/google api route, that won't increase filtering!

    Don't believe me? Try using voice with the mobile app.
    It is using the exact same speech to text parser as homey itself, but works better since 1, noise cancelation of the modern phones is good on its own, and 2, the microphone is close by.

    And athom isn't ignoring you, not too long ago there was a big response from athom about voice.. With this very same thing I'm saying now.

    The only reason for them to include Alexa and Google home, is because they want to get all devices connected in 1 place. 
    A lot of people wanted a second microphone in a different room, and is exactly what they are doing with adding support for that, which is a lot cheaper and easier then creating your own second device just for voice. 
  • right @caseda ! so now it is not really a showstopper for me. i can wait just a few month. but this year must be this problem fixed in most cases. i hope so ;-) 

    homey delivers much other playgrounds at the moment... 

  • @caseda ;  I never said it was easy. But in all advertisements for Homey, voice recognition is the first feature they promote, so it should be somewhere at the top of their priority list. Apparently companies like Google, Amazon and Apple can get a lot better results with this (voice recognition in an environment with background noise) than Athom can. So the way I see it, Athom has 3 options:
    1. Give up, and inform the users that it's not going to happen.
    2. Fix it themselves (on short term, not 2 years from now!)
    3. Join one of those big companies, and use their methods inside Homey, to improve recognition.

    Whichever option they choose, I would appreciate it if they'd let us know (in one of their bi-weekly Software Updates) what their current status is, and how they plan to move forward.

  • Fun to see u compaire Athom with Apple and Google. In my opinion U have 3 options:
    1. Give up and buy something else.
    2. Put Homey in the closet and wait for Athom to make Homey better then the rest.
    3. Stop complaining and make the best out of it. There is so much fun with Homey and it will be better then the rest!

    This is all because compairing Athom with Apple and Google is way off. Like compairing the old (or the new) Nokia 3310 with the Iphone 7+. Appels met peren vergelijken we call it in Holland.
  • "Äpfel mit Birnen" (german)  ;)
  • phil_s said:
    "Äpfel mit Birnen" (german)  ;)
    Sorry! :)
  • ZperXZperX Member
    I don`t think he compared compared Athom to Google. He compared their voice to text services. And there is a saying that you don`t have to reinvent the wheel. And if you don`t have the resources than don`t leave your customers on no man`s land.

    I think if a product does not work as it is advertised the customers have right to complain.
  • @ZperX is correct. I am comparing the quality of their voice recognition services, as caseda said that it was too difficult to filter background noise. The fact that those companies can do it considerably better than Athom does now, means that it CAN be done.
    And @Rododamelshe, I guess you also missed what I wrote in option (3)... Athom clearly doesn't have the resources to create their own voice recognition services that are on par with those of those three big companies. So why not use the technology of one of those companies, instead of trying (and apparently failing) to reinvent the wheel with such limited resources?

    "If you can't beat them, join them!"
  • JooleeJoolee Member
    edited March 2017
    Fun to see u compaire Athom with Apple and Google. In my opinion U have 3 options:
    1. Give up and buy something else.
    2. Put Homey in the closet and wait for Athom to make Homey better then the rest.
    3. Stop complaining and make the best out of it. There is so much fun with Homey and it will be better then the rest!

    This is all because compairing Athom with Apple and Google is way off. Like compairing the old (or the new) Nokia 3310 with the Iphone 7+. Appels met peren vergelijken we call it in Holland.
    That's not fair. Athom doesn't give the option to 'Give up and buy something else' because they simply are not clear about their plans for voice recognition. Maybe they plan to put 5 people on voice recognition the next couple of months and will bring a mega-update in a few months. Maybe they plan to ignore the issue altogether and we're better off buying something else. Maybe they are already bringing regular updates to the voice recognition and we're simply not noticing because they don't know what they're doing. Maybe the external company they promised that's working on writing new firmware for the audio processing chip will finally deliver next week and all our problems will be history... We just. don't. know. That's the whole point, we just don't know because there is no communication about the issue.

    The same point can be made about option two. About option three. You're right, of course. But in that case, Athom should stop advertising Homey as being able to understand voice commands. And they're clearly not doing that...
  • casedacaseda Member
    edited March 2017
    Bumblez said:
    "If you can't beat them, join them!"
    There is a lot more then just using their services, especially as a startup like athom is.
    when athom began, there wasn't even a alexa (publicly) or google home.
    And apple's siri is and was still closed source, only by hacks is it usable.

    And they aren't reinvent the wheel, they are using a good(!) (not by arhom created) speech to text service, but still lack a little of the FILTERING of the audio that is send over there.
    like i said before, if the filtering is not good, then your parser doesn't matter what you use, even if it is alexa, google home or even siri...

    and sure trying to Compare (yes you are comparing)
    Athom (with only 18 employees, and only 1 working on the voice with limited resources)
    to something like
    Google or Amazon (that can have a hundred people busy with voice, and have thousands of dollars to spare).
    and even they don't have Dutch voice recognition...

    amazon is already saying since the beginning that they will support most languages, but it's here now for 2,5 years and only 3 languages are supported... where there are 2 based on the English type.
    go start complaining at them that they STIL don't have a dutch voice recognition after this much time.
  • The 1st live speach to text demo I saw was 20 years ago from Lernhout and Hauspie. I know it can be much better today, but than you have to pay an other price. I was also disappointed in the first place about the voice recognition, but have seen much improvement and I'm sure it will be a lot better in the future. In the meantime I'm enjoying the rest of Homey's possibilities.  B)
Sign In or Register to comment.