Linear Chain Conditional Random Field with Latent Variables

[Varun]

Hello Everyone,

I am completely new to pystruct. I am working an application where I need to implement the parameter learning of a CRF with Latent variables. My model will be similar to Linear Chain CRF with variable (X_t1, Y_t2, H_t3) where X,Y,H corresponds to observed, unobserved and hidden variables respectively and t1, t2, t3 corresponds to lengths of respective sequences of X,Y,H of current data point.

After doing some search online I found out this amazing library. Just going over the documentation I am kind of guessing there is implementations available for ChainCRF and also for models with Latent Nodes, but I was wondering can I combine both somehow to achieve my task.

My question is, Can I achieve this parameter learning using pystruct ? If so kindly point me to the best place to start. 

Any pointers will be appreciated. 

Thanks in Advance.