GNOME Bugzilla – Bug 357507
SayAll does not use the "uppercase" voice settings for uppercase text
Last modified: 2006-09-25 18:35:20 UTC
Please describe the problem: When navigating through text in a document, Orca indicates the presence of capital letters by speaking them in a different voice (the "uppercase" voice). It fails to do this when SayAll is used. Steps to reproduce: 1. Create a document in which some text is written in all caps 2. Use the arrow keys to read the text 3. Use SayAll to read the text Actual results: When you use the arrow keys to read the text, anything that is written in all caps is spoken using the "uppercase" voice settings. When you use SayAll to read the text, the entire document is read using the "default" voice settings. Expected results: The "uppercase" voice settings would be used for the text written in all-caps, even when SayAll was being used. Does this happen every time? Yes. Other information:
Created attachment 73383 [details] [review] Patch to hopefully fix the problem.
The patch looks good to me. Just a note on the limitations - this will only use the uppercase voice if the entire utterance is upper case. Thus, it will not mix voices for a utterance such as "This is UPPER case." This is OK, but we need to at least understand and remember this limitation. The fundamental source of this limitation is that we are dealing with voice parameters on a per-utterance basis (i.e., we use a single voice for an entire utterance). Thus, if we wanted to change the pitch for only the word "UPPER" in the utterance, "This is UPPER case", we'd need to break it into three utterances ("this is" "UPPER" and "case"). This would break the overall f0 contour and other prosody for the utterance. Modern speech synthesis systems tend to support SSML, which allows us to modify voice parameters intra-utterance. Thus, we could potentially change "this is UPPER case" to have SSML markup to allow us to keep the prosody for the utterance, yet give hints to the engine for when/where/how to change the pitch. We're not that sophisticated however, and neither is gnome-speech.
Thanks Will. Understood. This patch provides a consistent behaviour with the other navigate-by-line speech utterances we've got now, to I've checked it into CVS HEAD and am closing as FIXED.