IndoBERTweet for Sarcasm: Evaluating Domain-Adapted Transformers for Indonesian Twitter Sarcasm Classification

This study investigated deep learning approaches for detecting sarcasm in Indonesian Twitter data.Fine-tuned transformer models including IndoBERTweet, IndoBERT, and IndoGPT were evaluated on a dataset of 8700 tweets and compared to an LSTM baseline.IndoBERTweet achieved the best performance with an F1-score of 89.11%, demonstrating the promise of domain-adapted models for this task.The findings provide useful insights into effective sarcasm classification strategies for low-resource languages.Further research on integrating contextual features is warranted to address limitations.