Re: Sindhi Voice of eSpeak-Ng

Reece H. Dunn
 

On Sun, Jun 11, 2017 at 09:46 pm, Dipendra Manocha wrote:

When we had tried to work on Sindhi E-speak, we faced a problem on what script to use for this:

 

Devanagari, Persio-Arabic or roman script.

 

/all three are in use at different geographical places and at many places all three are being used in same geographical locations too.

 

Which script are you using for this task?

The Sindhi language is currently using the Arabic script.

If the scripts have a 1:1 mapping (like Hiragana and Katakana in Japanese), the mapping could be done in the dictionary files. If it is more complex, my aim is to have a dictionary file for the language and script -- in this case, there would be three dictionary files (sd-Deva, sd-Arab and sd-Latn). I need to make some changes to espeak-ng in order to properly support this for the different multi-script languages (see https://github.com/espeak-ng/espeak-ng/issues/216 for details).

 Thanks

Dipendra

 

 

 

From: espeak-ng@groups.io [mailto:espeak-ng@groups.io] On Behalf Of Ejaz Shah
Sent: 12 June 2017 08:26
To: espeak-ng@groups.io
Subject: Re: [espeak-ng] Sindhi Voice of eSpeak-Ng

 

Thank you. I have used b, d., J and g for the time being. I'll experiment with other phonemes to see if they are more closer to the sounds in Sindhi.

By the way, Sindhi is pronounced sIndhi:

On 11/06/2017 9:05 PM, Valdis Vitolins wrote:

Thank you for your contribution!
 
I have merged your changes in my fork of espeak-ng.
https://github.com/valdisvi/espeak-ng/commit/2724442f4197cd4cdf9588ab7b
b877354695e066#diff-8397618d0750f47393471496bf7a0b57
 
You can test changes on: http://odo.lv/Espeak
 
I have added your name as maintainer for Sindhi language.
https://github.com/valdisvi/espeak-ng/blob/2724442f4197cd4cdf9588ab7bb8
77354695e066/espeak-ng-data/lang/inc/sd
 
You can use following approximation for these sounds:
 
ɓ could be b or B
ɗ — d, d2, d#
ʄ — J or J#
ɠ — g, g_
 
Look at phoneme definitions in phsource/phonemes file (or other
languages) and
and test them in command line, like (doubled square brackets informs to
use 
phonetic writing):
 
espeak-ng "[[ad'a:]]"
 
If you want to test some phoneme defined only in specific (e.g.
ph_latvian) language, add language code e.g.
 
espeak-ng -vlv "[[Ed'e:]]"
 
If phoneme is defined in phonemes file, you can use them directly in
dictsource/.._list or dictsource/..._rules files. If phoneme is defined
in other language, copy definiton to ph_urdu or ph_sindhi files and you
will be able to use them in Urdu or Sindhi also.
 
Valdis
 
hello,
 
I think there are some serious issues with the Sindhi voice of
eSpeak-NG. It seems that someone created the Sindhi voice using
mostly Urdu voice data. Sindhi language has many sounds that are not
present in Urdu. Even many letters of the Sindhi alphabet are not
supported. In the attached sd_list and sd_rules, I've included most
of the Sindhi letters of the alphabet and made other changes. I could
not find the eSpeak-NG phonemes for the IPA ɓ, ɗ, ʄ and ɠ.
 
I have updated the attached Urdu dictionary files also.I hope these
can be included in eSpeak-ng.
 
 
 
 

 

 

 

Join espeak-ng@groups.io to automatically receive all group messages.