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

List:       festlang-talk
Subject:    [festival-talk] OGI Lexicon and Capitalization
From:       klabbers () cslu ! ogi ! edu (Esther Klabbers)
Date:       2006-09-11 23:00:41
Message-ID: 4505EA99.8010809 () cslu ! ogi ! edu
[Download RAW message or body]

message from Esther Klabbers <klabbers at cslu.ogi.edu> to festival-talk
= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
Sorry, I don't have a solution here, but would it be possible to look up 
the whole word before it is split into syllables?

Esther

David Brown wrote:
> message from David Brown <dbrown at comcen.com.au> to festival-talk
> = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
> I am working on upgrades to Flinger, the festival singer module, and 
> am using the mwm5 voice.
> The lexicon the voice requires is OGI, which has, in its addendum, an 
> entry for 'UN'. That entry has phones which produce the sound of the 
> two letters capitalized, as it presumably should. (the common 
> abbreviation of 'United Nations').
> However, the lexicon looks for entries by doing a non-case sensitive 
> comparison of the input and entries. In song lyrics, as opposed to 
> normal text, there are often lyric tokens made up of single syllables 
> - for instance 'un-'  'like-'  'ly'  which spreads the word 'unlikely' 
> over three notes in a defined way. I have developed a means of getting 
> the correct phones for this type of thing, which involves getting the 
> lexicon output for each syllable, then the whole word, and 
> substituting phones from the latter into the former.
> Unfortunately, the lexicon output for the syllable 'un' is ((j U)0)((E 
> n)1)), which has twice as many phones as the entry for 'un' (which is 
> in the main body of the lexicon, not looked at if a match is found in 
> the addendum).
> My question is two-fold:
> (a) should the lexicon entry comparison be non-case sensitive, or is 
> this an error?
> (b) if the answer to (a) is 'yes, it is not an error', then is there a 
> way of directing the lexicon search to do a case-sensitive comparison, 
> on an individual basis?
> I am reluctant to change the lexicon entries themselves, as that 
> action would normally be beyond the scope of a user.
> Regards
> Dave Brown

-- 
Esther Klabbers
Center for Spoken Language Understanding
OGI School of Science & Engineering
20000 NW Walker Rd, Beaverton, OR 97006
phone: 503 748 3005 fax: 503 748 1306 www: http://www.bme.ogi.edu/~klabbers

= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
=    University of Edinburgh's Festival Speech Synthesis System       =
= http://festvox.org/festival      Sent Via festival-talk at festvox.org =
=                           To unsubscribe mail majordomo at festvox.org =
= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =



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