Table 4.

Table showing the performance of the proposed model with different fractions of the training data. The total number of labeled 20-second segments is 173160.

Training data percentage	# Labeled segments	Test precision	Test recall	Test F1
5%	8658	0.9350	0.7289	0.8192
10%	17316	0.9353	0.7166	0.8115
25%	43290	0.9281	0.7913	0.8543
50%	86580	0.9214	0.7956	0.8538
75%	129870	0.9184	0.7952	0.8524
90%	155844	0.9331	0.8007	0.8618