dear sir, could you tell me does sphinxtrain suport MAP training ?
MAP adaptation you mean? Yes, it's supported.
i mean discriminant HMM, that we maximize a posterior probabliity instead of likelihood as conventional HMM in training process.
Do you mean discriminative training with MMI criterion which tries to increase posteriors on the lattice? It's not supported yet, but there are chances it will be supported soon
http://www.cs.cmu.edu/~sphinx/Sphinx2010/papers/107.unblinded.pdf
yes, thank in advance.
Log in to post a comment.
dear sir,
could you tell me does sphinxtrain suport MAP training ?
MAP adaptation you mean? Yes, it's supported.
i mean discriminant HMM, that we maximize a posterior probabliity instead of
likelihood as conventional HMM in training process.
Do you mean discriminative training with MMI criterion which tries to increase
posteriors on the lattice? It's not supported yet, but there are chances it
will be supported soon
http://www.cs.cmu.edu/~sphinx/Sphinx2010/papers/107.unblinded.pdf
yes, thank in advance.