Re: [orca-list] Does Orca use SSML tags other than speak and mark?

[Arkadiusz Kozioł <zywek-mailing nvps pl>]

Rastislav, can you try to fix the problem, which I described a long time 
ago? When orca reading a text and go to the next line, it's stopping and 
there is an intonation, which can be understandable as the end of 
sentence. Joanie sait me, that it is the speech dispatcher problem.

W dniu 18.07.2020 o 00:59, Rastislav Kiss via orca-list pisze:
> Hello everyone,
> this question is perhaps primarily aimed for Orca developers, as it's
> related to the screenreader's internal working.
> I'd like to ask, does Orca use SSMl tags other than speak and mark when
> sending messages to speech dispatcher?
> I've tracked sd's pipe to find out what format does Orca use, when
> sending speech requests. I have noticed one <speak> tag in every
> message, as well as <mark> tags for setting markings on the spoken
> text.
> Except for these two, I didn't see any tags, even on places where they
> could occur, for example when spelling or reading links.
> Are there some I could possibly have missed?
>
> The reason why I'm asking is my new program, which I've been developing
> during the last week or two. As hollidays are finally here, I wanted to
> improve my language skills a bit, by learning Mandarin Chinese.
> However, espeak doesn't read chinese characters in neither slovak nor
> English version, and Chinese version can't read even english text
> properly due to pinyin, slovak is beyond any imagination. I don't know
> about anything like Nvda's switchsynth for Orca (is there some
> centralized depository of Orca addons btw?), so I've decided to solve
> the problem myself.
> I've made my own speech-dispatcher module, which doesn't contain speech
> synthesis on its own, but which is able to load other sd modules and
> after receiving text to speak, split it to parts written in various
> alphabets, including latin and chinese and send the respective parts in
> right time to right engine according to configuration.
> Latin, chinese and cyrillic alphabets are supported right now, but
> others can be added very easily if needed.
>
> Anyway, the program works very nice, but one thing I really didn't want
> to bother with was full SSML implementation, especially if most of it
> won't be ever used and would just slow-down the engine.
> It seems to be ready for release, I have implemented speak and mark
> tags to be handled properly, however if other tags occurred in the
> speak request, such as <emphasis>, in unlucky-enough situation, this
> could cause splitting the opening and ending of the tag to various
> engines, and would result in infinite loop.
> That is something I'd like to prevent at all costs, as working speech
> is crucial for us. Plus, as I saw today, an infinite loop effectively
> takes down not just the speech, but also the whole screenreader, so...
>
> Are there any other SSML tags used in Orca than <speak> and <mark>? Is
> this situation likely to change in future?
> SSML seems to be capable of dealing with various things such as
> spelling or emphasising text, but Orca seems to rather be solving them
> on its own. That's the point of the second question, whether this is
> wanted or just temporary behavior until more complex SSML rendering is
> created.
>
> Thank you in advance!
>
> Best regards
>
> Rastislav
>
> _______________________________________________
> orca-list mailing list
> orca-list gnome org
> https://mail.gnome.org/mailman/listinfo/orca-list
> Orca wiki: https://wiki.gnome.org/Projects/Orca
> Orca documentation: https://help.gnome.org/users/orca/stable/
> GNOME Universal Access guide: https://help.gnome.org/users/gnome-help/stable/a11y.html