Table 7. Hyperparameters employed for Bert, Roberta and Electra.
| Task | Model | Learning rate | Batch-size | Epochs | MSL |
|---|---|---|---|---|---|
| HOTEL REVIEW | BERT base | 3e − 5 | 10 | 5 | 512 |
| RoBERTa base | |||||
| ELECTRA base | |||||
| MOVIE REVIEW | BERT base | 3e − 5 | 10 | 5 | 512 |
| RoBERTa base | |||||
| ELECTRA base | |||||
| SENTIMENT140 | BERT base | 3e − 5, 2e − 5, 1e − 5 | 10 | 5 | 512 |
| RoBERTa base | |||||
| ELECTRA base | |||||
| CITATION SENTIMENT CORPUS(CSC) | BERT base | 1e − 5 | 5 | 5 | 512 |
| RoBERTa base | |||||
| ELECTRA base | 8 | 8 | |||
| BIOINFORMATICS CITATION CORPUS (BCC) | BERT base | 2e − 5 | 9 | 9 | 512 |
| RoBERTa base | 1e − 5 | 6 | 6 | ||
| ELECTRA base | 3e − 5 | 8 | 8 |