BESSTIE: A Benchmark for Sentiment and Sarcasm Classification for Varieties of English

BESSTIE is a benchmark dataset for sentiment and sarcasm classification across varieties of English. It was created by researchers Dipankar Srirag, Aditya Joshi, Jordan Painter, and Diptesh Kanojia and presented at ACL 2025. The dataset is hosted by the organization unswnlporg on Hugging Face.

Use Cases

• Benchmarking model performance on sentiment classification across different English varieties.
• Evaluating sarcasm detection algorithms on diverse linguistic data.
• Training models to improve robustness to dialectal variations in text.
• Conducting linguistic research on how sentiment and sarcasm are expressed in different English-speaking communities.

Strengths

• Specifically designed as a benchmark for two challenging NLP tasks: sentiment and sarcasm classification.
• Focuses on linguistic diversity by including multiple varieties of English.
• Created by academic researchers and presented at a major conference (ACL 2025).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

unswnlporg

Freshness

Last updated 2026-05-22 03:56:02; freshness should be verified.