RE: Revised SpeechRecognitionResult

Are there any objections to my statement below?  If not, I'd like to push this into the spec.

Thanks


From: Young, Milan
Sent: Wednesday, May 23, 2012 1:39 PM
To: 'Satish S'; Hans Wennborg
Cc: public-speech-api@w3.org
Subject: RE: Revised SpeechRecognitionResult

We can point to standards on both sides of the fence.  Perhaps it is a better use of time to consider our particular use case.

I'd argue that 90% of developers will not even think about the second item on the nbest list.  So why complicate their mental model let alone syntax with SpeechRecogntionAlternatives?

For the 10% that do understand an nbest list and its proper use, most will be familiar with VoiceXML which shares the same model.


From: Satish S [mailto:satish@google.com]<mailto:[mailto:satish@google.com]>
Sent: Wednesday, May 23, 2012 7:02 AM
To: Hans Wennborg
Cc: Young, Milan; public-speech-api@w3.org<mailto:public-speech-api@w3.org>
Subject: Re: Revised SpeechRecognitionResult

I'd prefer not having such shortcuts in the API. As a parallel, see the W3C File API's FileList interface
http://www.w3.org/TR/FileAPI/#dfn-filelist

To read the size of a file you'd have to do:
    var size = document.forms['uploadData']['fileChooser'].files[0].size;
but that hasn't resulted in a shorter version like
    var size = document.forms['uploadData']['fileChooser'].size;

If developers are accessing "item[0].utterance" more than once in their code they'd usually do
  var item = event.result.item[0];
  .. = item.utterance

Cheers
Satish


On Wed, May 23, 2012 at 12:11 PM, Hans Wennborg <hwennborg@google.com<mailto:hwennborg@google.com>> wrote:
>
> On Tue, May 22, 2012 at 7:22 PM, Young, Milan <Milan.Young@nuance.com<mailto:Milan.Young@nuance.com>> wrote:
> > Hello Hans,
> >
> > It's not uncommon for recognition engines to return a guess at what the user said/meant even for a nomatch result.  So we shouldn't rule this out in the API.
>
> Right. The spec currently says "nomatch event: [...] The result field
> in the event may contain speech recognition results that are below the
> confidence threshold or may be null."
>
> So that covers both cases.
>
> > As far as communicating this with a null vs event, I have a slight preference for an event.  Two reasons:
>
> I'm not sure what you mean by "communication this with a null vs
> event". I was talking about returning null or throwing an exception.
> Is that what you mean?
>
> >  * Easier for implementers.  This is a true alias.
>
> I'm not sure what you mean by true alias.
>
> >  * We may want to allow empty interpretations or utterances, and thus a null would be ambiguous.
>
> Ah, yes. So throwing an exception seems like the better option.
>
> Thanks,
> Hans
>