Table 12 AraXLNet compared with AraBERT and SVM
From: AraXLNet: pre-trained language model for sentiment analysis of Arabic
Model | AraSenTi (%) | ASTD (%) | SemEval (%) | AJGT (%) |
|---|---|---|---|---|
AraXLNet with Farasa | 94.78 | 93.01 | 85.77 | 88.43 |
AraXLNet without Farasa | 93.07 | 88.03 | 83.64 | 84.33 |
AraBERT | 84.65 | 92.13 | 85.05 | 91.94 |
SVM | 63.21 | 69.49 | 81.08 | 81.94 |