Table 2.

Classification accuracy for classifiers tested on the hand-annotated Corpus B. Scores in the right column are averages of 10-fold cross-validation.

	Train on A, test on B	CV on B
NB	.717	.959
LR	.742	.990
SVM	.748	.988
COS	.804	.982
RI	.762	.934
BSC	.610	.948
chance	.349