Table 11 Performance of NER using only BERT architecture
Entity | Precision | Recall | F1 | |
|---|---|---|---|---|
weighted avg | 87.66% | 91.08% | 89.28% | BERT |
weighted avg | 87.62% | 90.67% | 89.08% | RoBERTa |
weighted avg | 87.71% | 91.14% | 89.34% | GujiBERT |
weighted avg | 87.68% | 91.15% | 89.35% | GujiRoBERTa |
weighted avg | 87.14% | 91.01% | 88.98% | SikuBERT |
weighted avg | 87.19% | 90.90% | 88.95% | SikuRoBERTa |
micro avg | 87.10% | 91.08% | 89.05% | BERT |
micro avg | 87.12% | 90.67% | 88.86% | RoBERTa |
micro avg | 87.15% | 91.14% | 89.10% | GujiBERT |
micro avg | 87.21% | 91.15% | 89.14% | GujiRoBERTa |
micro avg | 86.61% | 91.01% | 88.75% | SikuBERT |
micro avg | 86.56% | 90.90% | 88.68% | SikuRoBERTa |
macro avg | 79.79% | 86.52% | 82.92% | BERT |
macro avg | 80.09% | 85.84% | 82.79% | RoBERTa |
macro avg | 79.79% | 86.41% | 82.87% | GujiBERT |
macro avg | 80.06% | 86.15% | 82.92% | GujiRoBERTa |
macro avg | 79.56% | 86.35% | 82.73% | SikuBERT |
macro avg | 79.26% | 86.34% | 82.54% | SikuRoBERTa |
