Topics

[espeak-ng:master] reported: Cantonese have some numbers missing when reading #github


espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #778 Cantonese have some numbers missing when reading
By valdisvi:

Larger numbers were missing in pronunciation rules. I added them by looking here. Is this ok?


espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #778 Cantonese have some numbers missing when reading
By Icenowy:

@valdisvi Chinese has some weird rule to interpret long numbers.

First, we usually break at 10000, not 1000. there's no special word for 1kk (it's 百万(hundreds 10000s) ) (BTW this rule do not break the current sample in the issue, because it's <10000) Secondly, when a zero appear after a whole number, a (zero) needs to be added. And when a 10 appears after the zero, it needs to be read as 一十(one ten). (although when a 10 appears independently, it is read as just )

So the sample 2019 should be read as 两千零一十九(2 thousand zero one ten nine)

Before implementing these complex rules, why not just fallback to read all numbers one by one? This should be more acceptable.

BTW I saw rules for some try on complex number parsing in zh_dict, however cmn uses just one by one, the complex rules are not called.


espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #778 Cantonese have some numbers missing when reading
By valdisvi:

To deal with this, more detailed rules should be added as it is described here. If only pronunciation for single digits is left, for some reason digits in between are skipped.


espeak-ng@groups.io Integration <espeak-ng@...>
 

[espeak-ng:master] New Comment on Issue #778 Cantonese have some numbers missing when reading
By Icenowy:

@valdisvi I think problem is at 60e9e3ebcbb6c5d77f5ddb4bc1f3bee61abe1b6e , in which the codepath which should run for cmn is run for yue.