Hi,
First of all, thanks for your work on MALLET!
Quick question — is there existing support for using an asymmetric beta prior as a vector of length numTypes (i.e., word-specific priors across the vocabulary) in MALLET's LDA implementation?
I'm currently working on adding this functionality, but wanted to check first in case it's already implemented.
Thanks,
Tal.