Frietag 2000 Maximum Entropy Markov Models for Information Extraction and Segmentation write up

This is a review of the paper Frietag_2000_Maximum_Entropy_Markov_Models_for_Information_Extraction_and_Segmentation by user:sgopal1.

• The paper slightly changes the hidden markov model and tries to adapt the new model to information extraction.Instead of a generative model where we estimate P(observation|class) , they estimate the P(class|prev_class,observation). They make the observation sequence a known value while trying to estimate the state transition probability. They fit an exponential model at each state that gives the probability of transitioning to another state from the current state.

• The need for defining such a model has been motivated very well. I think this is a really interesting and novel application of maximum entropy models.