Table 2:

Machine learning results on different categories for held-out test set of 500 documents.

Category	Precision	Recall	F1
Clinical Reports	0.84	0.63	0.72
Comment/Editorial	0.73	0.69	0.71
Communication	0.50	0.25	0.33
Contact Tracing	1.00	1.00	1.00
Diagnostics	0.93	0.83	0.88
Drug Targets	0.67	0.50	0.57
Education	0.63	0.71	0.67
Effect on Medical Specialties	0.79	0.49	0.61
Forecasting & Modelling	1.00	0.88	0.93
Health Policy	0.50	0.32	0.39
Healthcare Workers	0.80	0.89	0.84
Imaging	0.93	0.72	0.81
Immunology	0.44	0.44	0.44
Inequality	0.83	0.71	0.77
Infection Reports	0.75	0.30	0.43
Long Haul	0.00	0.00	0.00
Medical Devices	0.71	0.63	0.67
Meta-analysis	0.75	0.86	0.80
Misinformation	1.00	0.67	0.80
Model Systems & Tools	1.00	0.20	0.33
Molecular Biology	0.72	0.68	0.70
News	0.50	0.40	0.44
Non-human	0.83	0.50	0.63
Non-medical	1.00	0.43	0.61
Pediatrics	0.79	0.79	0.79
Prevalence	1.00	0.56	0.71
Prevention	0.72	0.56	0.63
Psychology	1.00	0.64	0.78
Recommendations	0.70	0.54	0.61
Review	0.57	0.75	0.65
Risk Factors	0.65	0.51	0.57
Surveillance	1.00	0.50	0.67
Therapeutics	0.70	0.68	0.69
Transmission	0.75	0.75	0.75
Vaccines	1.00	0.33	0.50
MICRO	0.76	0.62	0.68
MACRO	0.76	0.58	0.64