W3C home > Mailing lists > Public > www-style@w3.org > September 2012

Re: css3-speech, additional values for speak-as: duration, date, time, etc

From: Daniel Weck <daniel.weck@gmail.com>
Date: Wed, 5 Sep 2012 09:41:37 +0100
Cc: www-style@w3.org
Message-Id: <98893F14-0AC1-4F91-956F-A7CDA2BCAC42@gmail.com>
To: James Craig <jcraig@apple.com>
Hi James, thank you for your input!

As you probably already know, there is a W3C Note [1] for the SSML "say-as" [2] values, which contains definitions for date / time constructs. Although the 'speak-así property [3] of the CSS Speech Module Level 3 *conceptually* borrows from this SSML feature (the technical implementation is actually quite different), one design goal in this standardisation round was to keep the CSS Speech functionality as simple as possible (to encourage implementation uptake), whilst providing sufficient levels of "meaningfulness" / "usefulness". This rationale lead to the current minimalist feature set.

I can see how duration / date / time constructs would be of interest to producers of speech-enabled content, and I would definitely want to see some work being done in the next round (Level 4?) of the CSS Speech Module. Telephone numbers is yet another syntax to investigate, and perhaps other ones as well. This requires careful examination of the state of the art of TTS engines / APIs, and we need to check the status of other synthetic speech standards delivered by the Voice Browser WG, to avoid redundant / incomplete / conflictual interpretations.

At this stage of the Recommendation track for Level 3 of the CSS Speech Module (aka CSS3-Speech), it seems a little too late to embark upon such revision of the 'speak-así property. Other CSS WG members may have a different viewpoint though, if so, please chime in! :)

Once again, thanks for your feedback.
Kind regards, Daniel




On 30 Aug 2012, at 01:43, James Craig wrote:

> Requesting some additional values for speak-as on behalf of the Apple Accessibility Team:
> 	duration (default 'duration' treats' ambiguous ##:## treated as mm:ss)
> 	values:	duration | duration-mmss | duration-hhmm | duration-hhmmss
> 	time (##:## assumed to be hh:mm, and ##:##:## assumed to be hh:mm:ss) 
> 	value:	time
> 	date (ambiguous ##-## treated as either date-mmdd or date-ddmm, possibly dependent upon locale)
> 	values: 	date | date-mmdd | date-mmdd | date-mmddyy | date-ddmmyy | date-yyyymmdd | more?
> The purpose is to differentiate ambiguous time and duration formats like "4:18" which might be spoken as any of the following:
> 	"four eighteen" (time)
> 	"four minutes, eighteen seconds" (duration | duration-mmss)
> 	"four hours, eighteen minutes (duration-hhmm)
> Ambiguous date formats such as "1-12" could be spoken as:
> 	"january twelfth" in EN-US (date-mmdd)
> 	"december first" in EN-UK (date-ddmm)
> 	"doce de deciembre" in ES-ES (date-ddmm)
> Leading zeroes/zeds should not be necessary in dates if delimited by standard date format punctuation (periods, dashes, etc.) or spaces.
> Thanks for your consideration.
> James Craig
Received on Wednesday, 5 September 2012 08:42:43 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 16:28:48 UTC