Skip to main content

Table 2 Average F1-macro scores with standard deviation by model architecture

From: AutoDiscern: rating the quality of online health information with hierarchical encoder attention-based neural networks

Model ArchitectureQ4: ReferencesQ5: DateQ9: How Treatment Works
Random Forest0.83 (4)0.70 (6)0.66 (10)
HE BERT0.72 (14)0.77 (3)0.66 (13)
HE BioBERT0.71 (13)0.78 (5)0.66 (11)
HEA BERT0.86 (3)0.77 (4)0.68 (10)
HEA BioBERT0.80 (7)0.82 (6)0.72 (10)
Model ArchitectureQ10: Tt. BenefitsQ11: Tt. RisksAll Questions Avg
Random Forest0.53 (15)0.72 (4)0.69
HE BERT0.60 (11)0.74 (4)0.70
HE BioBERT0.56 (11)0.80 (4)0.70
HEA BERT0.66 (2)0.76 (3)0.75
HEA BioBERT0.54 (9)0.81 (5)0.74