3 ms·

> It will be very helpful if someone come up with automatic tonal detection systems for language learners to automatically check the correctness of their pronun

> It will be very helpful if someone come up with automatic tonal detection systems for language learners to automatically check the correctness of their pronunciations […] The number embedded in the Jyutping pronunciation indicates the tone. However, the automatic tonal detection is not as straightforward as it might seem due the tone sandhi (the tone change of the word/syllable depending on surrounding morphemes), and not all tonal languages have the tone sandhi. Cantonese, Teochew and Hokkien languages can have a pretty complex tone sandhi, though. For Cantonese and Jyutping specifically, the tone sandhi is marked as, e.g., «faan4*2» which means that the standard pronunciation is «faan4» (the low falling tone) can change into «faan2» (the high rising tone) depending on the following morpheme. The tone sandhi does not have fixed rules and depends on each morpheme. Depending on the tonal language, the preceding morpheme can initiate a tone change, or it can be the following one. Certain tonal languages can have a complex tone sandhi that affects the tonal change not just in a single morpheme but it the rest of the morphemes that comprise a full word. We would really be talking about a full contextual, semantics aware automated translator that applies the approriate tone sandhi on the contextual basis to give a Western reader the correct tone rendition in the transliteration.*

Umofomia 3y ago
Note that while Cantonese has the phenomenon of changed tones (變音 - https://en.wikipedia.org/wiki/Changed_tone https://en.wikipedia.org/wiki/Changed_tone), it is not actually considered tone sandhi (https://en.wikipedia.org/wiki/Tone_sandhi#What_tone_sandhi_is_and_what_it_is_not https://en.wikipedia.org/wiki/Tone_sandhi#What_tone_sandhi_i...). Tone sandhi is phonologically motivated, i.e., the tone changes arise from the pronunciation of the surrounding words, and are thus largely predictable. Cantonese changed tones, however, are generally lexically motivated, i.e., the changed tone is part of the realization of the word itself. Cantonese changed tones are thus more akin to Mandarin's erhua phenomenon (https://en.wikipedia.org/wiki/Erhua https://en.wikipedia.org/wiki/Erhua), which is also seemingly as random with regard to the words to which it applies. > For Cantonese and Jyutping specifically, the tone sandhi is marked as, e.g., «faan4*2» Note that the Jyutping standard actually doesn't specify how tone changes are marked. I believe the * convention originated at https://www.cantonese.sheik.co.uk/ https://www.cantonese.sheik.co.uk/ to facilitate Cantonese learning. Wiktionary has adopted a similar convention but using a hyphen (-) instead (https://en.wiktionary.org/wiki/Wiktionary:About_Chinese/Cantonese#Romanisation_considerations https://en.wiktionary.org/wiki/Wiktionary:About_Chinese/Cant...).
- inkyoto 3y ago
  It is true that the tone change in Cantonese is largely predictable due to it having the lexical nature. Since the parent raised a question about a generic solution to the tone marking problem, I thought I would use the tone sandhi as an example of a challenge that compounds the matter. On the other hand, languages such as Hmong exhibit the complex tone sandhi. Tone sandhi is not as pronounced in Cantonese as it is in some other languages. > Note that the Jyutping standard actually doesn't specify how tone changes are marked. No, it does not, and whether an asterisk or a dash is used, it is largely immaterial. What is important, though, is that not every Cantonese dictionary marks off a lexical tone change which might misguide the learner or the seasonal reader.

Umofomia 3y ago

Note that while Cantonese has the phenomenon of changed tones (變音 - https://en.wikipedia.org/wiki/Changed_tone https://en.wikipedia.org/wiki/Changed_tone), it is not actually considered tone sandhi (https://en.wikipedia.org/wiki/Tone_sandhi#What_tone_sandhi_is_and_what_it_is_not https://en.wikipedia.org/wiki/Tone_sandhi#What_tone_sandhi_i...). Tone sandhi is phonologically motivated, i.e., the tone changes arise from the pronunciation of the surrounding words, and are thus largely predictable. Cantonese changed tones, however, are generally lexically motivated, i.e., the changed tone is part of the realization of the word itself. Cantonese changed tones are thus more akin to Mandarin's erhua phenomenon (https://en.wikipedia.org/wiki/Erhua https://en.wikipedia.org/wiki/Erhua), which is also seemingly as random with regard to the words to which it applies. > For Cantonese and Jyutping specifically, the tone sandhi is marked as, e.g., «faan4*2» Note that the Jyutping standard actually doesn't specify how tone changes are marked. I believe the * convention originated at https://www.cantonese.sheik.co.uk/ https://www.cantonese.sheik.co.uk/ to facilitate Cantonese learning. Wiktionary has adopted a similar convention but using a hyphen (-) instead (https://en.wiktionary.org/wiki/Wiktionary:About_Chinese/Cantonese#Romanisation_considerations https://en.wiktionary.org/wiki/Wiktionary:About_Chinese/Cant...).

inkyoto 3y ago
It is true that the tone change in Cantonese is largely predictable due to it having the lexical nature. Since the parent raised a question about a generic solution to the tone marking problem, I thought I would use the tone sandhi as an example of a challenge that compounds the matter. On the other hand, languages such as Hmong exhibit the complex tone sandhi. Tone sandhi is not as pronounced in Cantonese as it is in some other languages. > Note that the Jyutping standard actually doesn't specify how tone changes are marked. No, it does not, and whether an asterisk or a dash is used, it is largely immaterial. What is important, though, is that not every Cantonese dictionary marks off a lexical tone change which might misguide the learner or the seasonal reader.