Table 5 Classification report of different models for offensive language identification.
Class label | Measures | Logistic regression | CNN | Bi-LSTM | RoBERTa | BERT | Adapter-BERT |
|---|---|---|---|---|---|---|---|
Not offensive | Precision | 0.82 | 0.76 | 0.95 | 0.91 | 0.88 | 0.89 |
Recall | 0.90 | 0.92 | 0.83 | 0.73 | 0.84 | 0.86 | |
F1-Score | 0.85 | 0.83 | 0.88 | 0.81 | 0.86 | 0.88 | |
Support | 2775 | 3049 | 3487 | 3049 | 3049 | 3049 | |
Offensive targeted insult group | Precision | 0.36 | 0.11 | 0.13 | 0.19 | 0.27 | 0.17 |
Recall | 0.29 | 0.05 | 0.17 | 0.41 | 0.41 | 0.38 | |
F1-Score | 0.32 | 0.07 | 0.15 | 0.26 | 0.33 | 0.29 | |
Support | 384 | 302 | 222 | 302 | 302 | 302 | |
Offensive targeted insult individual | Precision | 0.42 | 0.12 | 0.11 | 0.30 | 0.33 | 0.43 |
Recall | 0.30 | 0.05 | 0.12 | 0.39 | 0.35 | 0.38 | |
F1-Score | 0.35 | 0.07 | 0.12 | 0.34 | 0.34 | 0.32 | |
Support | 392 | 283 | 270 | 283 | 283 | 283 | |
Offensive targeted insult other | Precision | 0.00 | 0.00 | 0.00 | 0.06 | 0.00 | 0.00 |
Recall | 0.00 | 0.00 | 0.00 | 0.15 | 0.00 | 0.00 | |
F1-Score | 0.00 | 0.00 | 0.00 | 0.09 | 0.00 | 0.00 | |
Support | 34 | 48 | 0 | 48 | 48 | 48 | |
Offensive untargeted | Precision | 0.40 | 0.23 | 0.06 | 0.37 | 0.42 | 0.44 |
Recall | 0.32 | 0.08 | 0.23 | 0.49 | 0.41 | 0.43 | |
F1 Score | 0.36 | 0.11 | 0.09 | 0.42 | 0.41 | 0.44 | |
Support | 491 | 394 | 97 | 394 | 394 | 394 |
