chaug wrote:
> This morning I had major issues with getting the skill to shuffle play
> my playlist. There were three songs that produced just silence, two of
> which I noted down and later tried to play using stream.mp3 as described
> above. Both of them played in the web browser.
>
> I also checked the alexa voice history to see if that would explain some
> of the other failures. Turns out that in one case, she understood "tell
> me the server" instead of tell media server. Which is obviously pretty
> poor AI performance, given that she has a skill called "media server".
> In fact, I would claim that there is no AI involved here. It's just
> stupid to immediately fall back on the default answer "Here's something
> I found on the web..." I don't get why you wouldn't check against
> similar sounding skill names before coming up with a nonsense guess of
> what the user might have meant by "tell me the server".
>
> In another situation alexa just heard the wake word but not the
> following wake word. The only difference to the second time I said it
> was that the pause between "Alexa" and "next" was a tiny fraction of a
> second shorter. She simply doesn't seem to listen long enough after the
> key word. Half or even a quarter of a second more would probably solve a
> lot of issues. I noticed earlier that "Alexa, skip" works most of the
> time while "Alexa, next" doesn't. My explanation for this is now that
> the "s" in skip makes enough of a hissing sound to keep Alexa awake
> while the "n" in next doesn't. That's how tiny the difference is.
>
> The only thing I can do as a user here is try and ignore the Alexa ring
> light and just talk. What I mean is that I noticed a tendency in my
> family to say "Alexa", then wait for here to indicate that she's
> listening, and then give the command. As I noticed this kinda natural
> behaviour in others, I realised that I was also doing it, though to a
> lesser extent. I don't really like to train myself, though, to not wait
> for the light to go on. Firstly, because it means that I will have to
> repeat commands when it turns out that Alexa didn't wake up, but also
> because Alexa is already teaching us habits that are not good for human
> interaction (like not saying thank you when someone did something for
> you or ordering her to stop talking) so that I think it would be good to
> maintain the attitude of waiting for a response before starting the
> conversation.
>
> Nut apart from those concerns, I won't quite understand why the slight
> pause between the wake word and the command works without problems
> outside of the media server skill. It's only when MediaServer is active
> that this problem occurs. Do you as skill developer have control over
> the waiting period or is Amazon defining shorter in-skill periods?
>
>
>
> I will try simon says asap but I don't think that is the issue. The Echo
> dot that I've tested this most on is brand new and the same problem has
> occurred on at least one other echo dot.
>
>
>
> If "shuffled" is added at the end, isn't there a risk that it gets
> interpreted as part of the playlist name? Then again, always removing
> the last "shuffled" from the playlist name would probably work. So even
> if you had a playlist ending with "shuffled" saying playlist something
> shuffled shuffled would work. So, yea, that would be great have!
>
> Alternatively: how about "shuffleplay" and "shufflestream" as commands?
> i.e. "Alexa, tell mediaserver to shufflestream favourite 8" instead of
> "Alexa, tell mediaserver to stream favourite 8 shuffled"
Wow, "tell me the server" would be a first ! Amazon calls Alexa an AI,
but just so you know: in the voice model you have to have separate
entries for "what's playing" and "what is playing" because the AI cannot
even join those dots. :confused: I kid you not.
When you install a skill, a random Amazon server gets assigned to you.
It would appear that different servers may have different patch levels
or even be running different versions of the Alexa engine. Your
assigned server might have a tweaked version of something (outside my
control) that behaves oddly. We will never know. It's certainly not in
the hands of a skill developer to pace anything. Alexa hears, thinks for
a bit, and decides what to pass on to skills. I've often seen 'old'
console.log() statements in the skill logs that dated from a version of
my code that was changed weeks ago and should not be still 'live'.
Meaning some users were not running my latest code even though I updated
all URNs. Hmmm.
I can always strip the word 'shuffled' from the {Playlist} slot value,
so that's not an issue. Adding shuffleplay and shufflestream commands
would blow up the voice model and I'm already close to the limit of what
a skill is allowed to have. Personally, I do find a tiny pause between
Alexa and the command fares better (you can hear that in my videos) but
YMMV.
------------------------------------------------------------------------
philchillbill's Profile: http://forums.slimdevices.com/member.php?userid=68920
View this thread: http://forums.slimdevices.com/showthread.php?t=111016
_______________________________________________
plugins mailing list
plugins@lists.slimdevices.com
http://lists.slimdevices.com/mailman/listinfo/plugins
