Table 5.

Comparison of using different numbers of documents in the fracture task

Mayo Clinic Smoking Status Classification
	tf-idf	topic modeling	word embeddings
SVM	0.69	0.73	0.80*
RF	0.69	0.72	0.81*
Proximal Femur (Hip) Fracture Classification
	tf-idf	topic modeling	word embeddings
SVM	0.85	0.91	0.95*
RF	0.77	0.92	0.93*

The asterisk indicates that difference between word embeddings and other features is statistically significant