Table 6:

Comparative results on classifying PPD. These results were obtained prior to the removal of depression related terms from the corpus.

Method	Prec.	Rec.	F1	Acc
Stochastic gradient descent (l1 = 0.95, loss = hinge)	0.90	0.88	0.89	0.93
Multinomial naive Bayes (alpha = 1)	0.95	0.81	0.87	0.92
Perceptron	0.83	0.87	0.85	0.89
Passive aggressive (loss = sqrt_hinge)	0.90	0.89	0.9	0.93
RSDD trained model	0.95	0.43	0.6	0.8