Session6: Tones 1. Helen Meng from Chinese University of Hong Kong gave a talk on Mandarin tone. * various kinds of tone for a syllable - tone 1 - 6: long duration (non-entering) - tone 7 - 9: short duratin (entering) * (syllable + tone) notation is essential for Chinese synthesis - pinyin (for Mandarin) - jyutping (for Cantonese) * proposal of <tone> element - tone changes correspond changes of meanings/context/speaking style * several tone sandhi rules depending on dialects - rule for Mandarin: tone3 + tone3 => tone2 + tone3 - rule for Cantonese: tone4 + tone4 => tone4 + tone{2,1,4} * how to denote the combination of characters and tones ? - <phoneme alphabet="t" ph="2 4">不要</pnonenme> - <phoneme alphabet="pt" ph="/bu 2/yao 4/">不要</pnonenme> * if word boundary differs, meaning and tone can differ... - 打死/老鼠 (kill a mouse) - 打/死老鼠 (hit a dead mouse) 2. Jianhua Tao from Chinese Academy of Sciences gave a talk on Tone Sandhi * various dialects in Chinese - 9 or 10 variety - almost local language - various tone sandhi rules * tone and tone sandhi - 25 tones for a character (24 are defined in ISO 10646) - (25 + 25) * 2 = 100 sandhi possible * How to denote tone sandhi? - Can tone sandhi problem be considered as co-articulation problem? - Add a tone element with between 9 and 24 (number is not important now) values? - Is this sufficient for specifying most Chinese dialects? - tone sandhi is example of tonal errors? * 2 Levels of language specification should be considered - Phonemic alphabet: phoneme is sound differenciate lang vs. - Phonetic alphabet: differenciate same phoneme (e.g. using tone) * "xml:lang" specificatin - how about use "xml: lang" in combination with "speak-as" attribute? language (character) specification xml:lang="zh-Hans" how-to speak (accent, etc.) specification speak-as="zh-CMN-CN" speak-as="zh-CMN-TW" speak-as="zh-CMN-SG" speak-as="zh-CSZ" speak-as="zh-CAN" c.f.) RFC3066 (new) xml:lang="zh-CMN-CN xml:lang="zh-CMN-TW xml:lang="zh-CMN-SG xml:lang="zh-CSZ - lang element extention for how to specify link of syllable? * relation between dialects and syllable inventory (=phonemeic inventory) - pinyin is tone and syllable (and phonemic) inventory - character -> syllable mapping - phonemic distincution - phonetic markup: phonemic + something * How about Korean? - tone element proposal by Korea? - What does Korean use as phonemic - phonemic rep vs. phonetic rep * what character set? - not only IPA but also supportint specific phonetic system for many speech vendor's own languages? * creation phonemic and syllabic inventory to distinct - phoneme level distinction which is phonemically important - maybe syllable & tonal [SUMMARY] We agreed to recommend: ----------------------- * necessary for SSML to support and encourage the definition of dialect-specific phonemic inventories. These may be syllabic but must incorporate all phonemically relevant distincitions (such as lexical tones, syllable duration, etc.) Still under discussion: ----------------------- * need for explicit tone marking element (beyond the use of phonemic inventory) * need for explicit marking of tone sandhi (beyond the use of IPA) - slight leaning towards IPA being sufficient because this is allophonic variation * need for explicit marking of phonemically relevant syllable duration