How is the marginal probability calculated?

[ctrado18]

Hey,

I want to extract the influence of each features on each word to be an entity.
I tried to find out how the marginal probability is calculated for each word, and looked also here: https://www.chokkan.org/software/crfsuite/tutorial.html#idp8849142672
There is to be read something abput ``sclaing` value. But this is not found in the pycrfsuite.

You can see the feature weight of course such that you can walk through the list of weights and check which feature is belonging to each word. But I would like to know how the probabilities out of those weights are calculated?

Thanks!