[hts-users:01215] Re: WER Comparison of HTK and HTS

[Junichi Yamagishi <jyamagis@xxxxxxxxxxxx>]

Hi,

> Can you tell us the total numbers of leaf nodes in your phonetic  
> decision trees built by HTK and HTS?

I've already deleted the models and log files. Sorry.
I will train the model and report them again.

> I expect that this difference was caused by modifications in tree- 
> based clustering part.

I suspect flooring of variance also has effects more or less.

Junichi