mayanknagda / pos-tag-prediction-using-conditional-random-field

POS Tagging

Please find the attached .ipynb file.

We import the files and libraries for our POS Tags Prediction

sklearn_crfsuite was utilized to make models.

2 Types of Models are analysed.

Model 1 doesn't account for the prefix/suffix in the features

Model 2 accounts for prefix/suffix in the features

Performance metrix for both the models are produced in the end.

Model 1 gives a f1 score of .866

Model 2 gives a f1 score of .901 making is better than Model 1

Other performance measures such as precision and recall were also used to analyse the models.

l2sgd (SGD with L2 Regularization) is used while training both the models. The regularization constant is taken as 0.1