[prev in list] [next in list] [prev in thread] [next in thread] 

List:       festvox-talk
Subject:    Re: new limited domain voice & question re: phoneset etc
From:       "Nickolay V. Shmyrev" <nshmyrev () yandex ! ru>
Date:       2011-11-30 21:25:05
Message-ID: 1322688305.24229.6.camel () localhost ! localdomain
[Download RAW message or body]


On 29/11/2011 at 21:02 -0500, Dahlia Thompson wrote:
> Hi Everyone,
> 
> Thanks in advance for your patience & your response(s)  


Hello

> I am working on creating a new limited domain voice (not time) in a
> new language.  I have installed and compiled Festival, Festvox and
> Speech Tools.  I have sent the necessary  environment variables, set
> up my directory structure, created my prompt list etc. The default
> voice is currently set to kal_diphone.
> 
> 
>       * How do I get Festival to access the a) phone set (phoneme
>         inventory), b ) lexicon and c) lts rule set for my language?
> I do not wish to use the English phoneset for annotation as the phone
> sets are not similar:

Describe everything for your target voice in corresponding scheme files
using required format. Put those files in the ldom voice folder. Load
them with (require). Select them with (lex.select) and analogs.

> 
>       * Do I need to modify the current radio_phone.scm  and cmu_lex
>         files to reflect my language or should I just add my own files
>         somewhere and if so where? Inside the festival/lib directory
>         or inside the festvox directory or ...?

It's better to avoid modifying files from the distribution. You can just
put files in voice scm folder.

> I have tried adding my own files but the system does not recognise
> them.  For example, when I build my prompts and examine the utterance
> files, I see that my phoneset is not being used..  
> Where should I put my phoneset (file) to ensure once I run festival -b
> festvox/build_ldom.scm '(build_prompts "etc/myutterancelist")", the
> system uses the phoneme inventory for my language and not English? 

You need to load file with (require). You need to select lexicon with
(lex.select). You can find some examples of this in other voices. You
need to do all those actions inside (voice_INST_LDOM_VOX_ldom) function.


["signature.asc" (application/pgp-signature)]

[prev in list] [next in list] [prev in thread] [next in thread]