0
$\begingroup$

I am using the Conditional Random Fields CRF suite scikit-learn wrapper algorithm. I have read on the literature various approaches for feature selection, but I cannot find any on that package or, generally, available ones for CRF. Would you know any libraries (Python preferred) or easy to implement algorithms for this purpose?


Update

I tried using the scikit-learn feature selector's library but does not work for 2 reasons: 1) the CRF takes as an input list of lists of dicts whereas the other take tabular data 2) the CRF does not have a .coef or a feature_importances_ attribute in order to perform feature selection.

$\endgroup$
0
$\begingroup$

Since sklearn-crfsuite provides a scikit-learn compatible estimator, any scikit-learn feature selection class that takes an estimator could work.

| improve this answer | |
$\endgroup$
  • $\begingroup$ I tried, but I believe the type of data crashes the sklearn's selectors. That's because the CRF wrapper takes as an input nested lists whereas the sklearn selectors expect some sort of tabular data such as a numpy array. Then I believe if I convert my data into an tabular array the CRF will not work... $\endgroup$ – Grzegorz Nov 23 '19 at 15:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.