Session 11: Other issues ---------- 16:15-17:15 Open discussion: * Background audio Several here have said it was mandatory. Jim asks about utility of SMIL. Paolo asks how you synchronize with SSML. Says we need some kind of events for synchronization. iFlytech says they just want a simple playback (just start and stop). Also needed for face synchronization. * Ruby Question is why we need Ruby. Kana sequence and phoneme sequence are different. It helps TTS engine reduce misreadings of Japanese. Not clear why this can't be done via or . Kazayuki will work with our Japanese participants to understand what the issue is here. * Mora Generally agreed that SSML needs to include flexibility in how timing is specified -- allow for mora, syllable, phoneme, and any others we find we need. * Translate Issue is how to indicate that something like "Windows2000" should be spoken out as the Chinese words for Windows and 2000. It turns out that the real topic is to be able to indicate to SSML processor how it should treat text it encounters that is not in its lexicon -- does the processor use a rule engine to figure out how to pronounce, does it assume it's a particular other language, etc. * Domain element This is for performance improvement -- text analysis & normalization. Worth considering. Domain attribute on voice is for choosing the voice that is best for this domain. * Syllable markup Two issues a) do you need it in order to annotate syllables with additional information -- question to consider is should SSML annotate at that level? b) may need in order to delineate syllables in Thai -- Richard doesn't believe this is necessary but we should find out. * Multimodal - Synchronization SSML needs to consider how it can be synchronized with other specifications, including the multimodal work. Richard separately suggests that the SSML group consider modifying the phoneme element to permit prosodic annotation over the contents of what is currently the "ph" attribute.