Alexa will turn a device on but doesn't understand to turn it off.

[famewolf]

Sometimes the system just acts weird.  It can properly handle "Alexa, turn the rice cooker off" but does nothing but give a failure chime if told "Alexa, turn the rice cooker on"....the only solution I could find was renaming the device "ricepot"  which I consider a kludgy workaround.

[ScottinPollock]

Try changing things up a bit. "Turn on my rice cooker", or " rice cooker, on".

I find the some device names better than others, but can usually find a reliable phrase for the not so good ones.

[DParker]

Ditto what Scott said.  I don't run into that sort of recognition failure often, but I have found that phrasing commands the way he suggests is much more reliable than doing it the other way around.

[famewolf]

I appreciate the advice but I did try various methods.... turn on rice cooker....turn rice cooker on.  etc...I either got the chime or alexa said "Sorry, I didn't understand the question"....I submitted it to echo support who said they confirmed the voice confirmation was turned off for that phrase but said it should still work...well it doesn't...and if just changing the name to "ricepot" resolves the issue then the problem isn't on my end.  Although alexa does do a lot of tasks well it dropped the ball on this one.

[renegade600]

how about just turn on rice?  there are a few phrases that will not work for me either.  I think it has to do with the individual voice tone, how you pronounce things, and any other sound that could interfere.   

[DParker]

It's funny how quickly our expectations of speech recognition as a consumer technology have become so high.  Just a few years ago the idea of controlling your home, retrieving all manner of useful information, initiating entertainment, etc simply by speaking to a machine...and having it speak back (in a pleasant, natural-sounding voice, no less) was still mostly perceived as the stuff of science fiction.  That's because analyzing human vocalizations in order to recognize words is a complex and computationally intensive process. Until fairly recently the state of the art of the technology consisted of the ability to recognize individual spoken words for the fairly simple purposes like converting them to text for dictation, or combining one or two of them into rudimentary software control commands (ie, "Save file", "Launch Excel", etc.)  And that required a fair amount of training the software to reliably recognize a single person's voice, pronunciation and inflection.  And even then relatively minor variations in the way you said the same words could completely befuddle the software.

Now we have cell phones, automotive head units and household appliances that come from the factory able to re accurately and reliably recognize words spoken by just about anyone who is capable of relatively clear enunciation.  And not only that, they layer on top of that the ability to makes sense of combinations of words, in the form of phrases and even whole sentences spoken in a natural, casual and sometimes even grammatically incorrect manner...and THEN do something useful (or at least entertaining) with it.

I like to remember this from time-to-time...not only because it's nice to recognize how far this stuff has come, but also because it keeps me from becoming frustrated on those occasions when things don't immediately...or maybe ever...work like I think they should.