TABLE 3.

Results of classification with logistic regression and SVMs using different features evaluated using 10-fold cross validation of the training data with heuristic labels.

Approach	Precision	Recall	F1-score
Logistic regression (imbalanced labels) + TF-IDF	1.00	0.56	0.68
Logistic regression + USE (balanced labels)	0.83	0.51	0.59
Logistic regression (balanced labels) + TF-IDF	0.87	0.90	0.88
Linear SVMs + TF-IDF (balanced labels)	0.9	0.91	0.91