[SRILM User List] Language Model Adaptation

[Andreas Stolcke]

Luis Uebel wrote:
> How can I produce a LM adapted for a particular speaker?
You need (a lot) of speaker-specific data.  Then you adapt as you would 
to other LM conditions, like specific genres, topics, etc.
> What are the SRILM tools and parameters that I need to use?
Your question is too general.   Start by reviewing the literature on LM 
adaptation, or the overview paper by Bellegarda
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.91.4893&rep=rep1&type=pdf
and figure out what methods look promising.

SRILM implements several, but not all of the techniques described.  The 
options in the ngram(1) tool that are relevant include

-mix-lm
-bayes
-cache
-adapt-marginals
(and associated options described nearby in the man page).

Andreas

>
> I produce a LM restricted for words that appear in test set.
> Model was quite big and now I would like to adapt for a particular speaker
> using a set of sentence (text) that he pronunciated.
>
> Thanks,
>
>
> Luis
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> SRILM-User site list
> SRILM-User at speech.sri.com
> http://www.speech.sri.com/mailman/listinfo/srilm-user