Sentiment Classification on Persian Twitter dataset (test)
74.08AccuracyHybrid Pipeline
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Hybrid PipelineEmbeddings=ParsBERT + LaBSE, Classifier=Neural Model + Rule-Based Model, Decision Rule=Confidence >= 802026.03 | 74.08 | 72.02 | 72.03 | 75.99 | |
| Fine-Tuned Sentiment Analyzer2026.03 | 51.49 | 48.88 | 49.79 | 50.92 | |
| Transformers Encoder-Decoder2026.03 | 51.02 | 50.53 | 53.12 | 48.22 | |
| Random Forest2026.03 | 48 | 48.9 | 51.04 | 47.06 | |
| Baseline2026.03 | 20 | 20 | 20 | 20 |