Combining few-shot fine-tuning with incremental learning and regularization lets XLM-R and mDeBERTa reach about 96% accuracy on Persian sentiment analysis across five domains.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Cross-lingual Few-shot Learning for Persian Sentiment Analysis with Incremental Adaptation
Combining few-shot fine-tuning with incremental learning and regularization lets XLM-R and mDeBERTa reach about 96% accuracy on Persian sentiment analysis across five domains.