Conditional Random Fields for Jointly Labeling Multiple Sequences (DCRFs)

[Cristian Victoria]

Hello all

It is possible to train a jointly labeling multiple sequences task in Factorie?

I want to implement the model from  Dynamic Conditional Random Fields: Factorized Probabilistic Models for Labeling and Segmenting Sequence Data http://homepages.inf.ed.ac.uk/csutton/publications/dcrf.pdf or http://homepages.inf.ed.ac.uk/csutton/publications/dcrf-nips03.pdf

Graphical model:

I define the template model, and assigned the values to the variables then i tried to train it. when i pass to the trainer a sequence of examples from only one sequence, it not learn the parameters for the transitions labels of the other sequence. when i pass to the trainer examples of the two sequence it learn all the parameters but i think that is not doing it jointly.

Can you help me, maybe you have any suggestions?

Thank you.