Table 2 Comparison between hULMonA and AraBERT [29, 30]
From: AraXLNet: pre-trained language model for sentiment analysis of Arabic
Dataset | hULMonA (%) | AraBERT v0.1/v1 (%) |
|---|---|---|
BERT | 95.7–95.7 | 96.2/96.1 |
RoBERTa | 67.7–69.9 | 92.2/92.6 |
XLNet | 51.1–52.4 | 58.9/59.4 |