Re: [Espeak-general] espeak-zh multiple tone 3s

[Silas S. B. <ss...@ca...>]

Got some more data here:

(1) wǒ xiǎng qǐng nín
Natives 1 and 2: wó xiáng qǐng nín
Native 3, and SinoVoice demo on
http://www.sinovoice.com.cn/english/tiyan.html :
wó xiǎng qǐng nín

(2) Měiyǔ bǔxíbān
Natives 1 and 3 and SinoVoice: Méiyǔ
bǔxíbān
Native 2: Méiyú bǔxíbān

(3) kěyǐ tǎolùn
Natives 1 and 3 and SinoVoice: kéyǐ tǎolùn

(4) ..gěi nǐ bǎochí
Natives 1 and 3 and SinoVoice: ..géi nǐ
bǎochí

(5) jiàodǎo nǐ shǐ nǐ dé yìchu
Natives 1 and 3 and SinoVoice (and previous
group): jiàodǎo nǐ shí nǐ dé yìchu

(6) zhǐyǒu shǎoshùrén
Natives 1 and 3 and SinoVoice: zhíyǒu
shǎoshùrén

(7) kěyǒu-kěwú
Native 1 and SinoVoice: kéyóu-kěwú
Native 3: kéyǒu-kěwú

(8) zhìshǎo yǒu liǎng ge
Natives 1 and 3 and SinoVoice: zhìshǎo yóu
liǎng ge

(9) lìngrén nányǐ lǐjiě de
Native 1 and SinoVoice: lìngrén nányǐ
líjiě de
Native 3: lìngrén nányí líjiě de

(10) kěyǐ zěnyàng
Native 1: kéyí zěnyàng
Native 3 and SinoVoice: kéyǐ zěnyàng

(11) shénme fāngfǎ kěyǐ gǎishàn
Natives 1 and 3 (and SinoVoice although it
seemed to fault on the "me"): shénme fāngfǎ
kéyǐ gǎishàn
Native 2 and previous group: shénme fāngfǎ
kéyí gǎishàn

(12) kěyǐ gǎishàn
Native 1: kéyí gǎishàn
Native 3 and SinoVoice: kéyǐ gǎishàn

On the surface, it seems that most of the above
can be explained by 3 rules:

Rule 1.  The pattern "33 3..." is changed to
"23 3...", and this is fixed in the output
before the rest of the input is evaluated.
This covers examples 2,3,6,10,11,12.

Rule 2.  In the pattern "(not-3) 3 3 3 2",
irrespective of word boundaries, the 3332 is
changed to 2332.  This covers examples 1,2,4,7.

Rule 3.  Apply the beginners' rule to the rest. 
This covers examples 8 and 9.

BUT those rules do not catch example 5 properly.

We still haven't got to the bottom of this.

Silas


-- 
Silas S Brown http://people.pwf.cam.ac.uk/ssb22

"What is now proved was once only imagined" - William Blake