espeak for thai language problem

[yzl123456]

I find some problems when using espeak-ng for thai language is that the output phoneme order is not correct, due to the writing order of thai language is different from the speak order of thai language, anyone find similar problems when using this for thai language? Thanks for anyone's response!

[sr12zar]

Have a look at espeak-ng-thai, it describes a 3 phase process to deal with thai text

• identify and separate words from string of text -- Thai does not separate each in writing
• 'normalize' words with complex vowels into simple form 'head - vowel - tail'
• add tone to words --tone is very important in Thai

This is done with SWATH, g2p.awk and bash scripts. I will try to answer your questions.

[yzl123456]

@sr12zar thanks，I see the repo, whether there is some website I can see your project's performance, like input some thai language, output corresponding phoneme, so that I can check if it's helpful me, thanks for your response!

[yzl123456]

@sr12zar Based on espeak-ng, I try some thai cases:

1. ไม่มีค่ะ
2. รายชื่อสมาชิก
3. ลิงค์ที่ใช้บ่อย
4. มีอะไรใหม่?
5. คนไทย ไม่มีเลข 13 หลัก

and espeak-ng output is like:

1. ˈai2m mskh s
2. ɹˌa2sja2chˈueː ʔˌa5sa2mˈa2schiɜk
3. lˈi2nɡa2kh ths ˈai2ch b ʔˈa2j
4. msʔˌa2sai2ɹˈai5ha2m
5. kha2nˈai2tha2j ˈai2m msˈe2la5kh nu5ˈeɜnɡsam hˈaɜlmk

and I give these results to someone who knows thai language, but he said it's terrible because it's output order is not the speak order. And I really want to get your "espeak-ng-thai" project's output for these sentences, thanks a lot!

[sr12zar]

No there is no website demonstrating the use of espeak-ng-thai. You will need to build your own system of espeak-ng as described in 'README.md' and write your own 'bash' scripts that include the 'preprocessing components' to test your test data.

Sorry, espeak-ng-thai is meant for programmers rather end-users.