DynaSent

Dataset Information
License
Unknown
Homepage

Overview

DynaSent is an English-language benchmark task for ternary (positive/negative/neutral) sentiment analysis. DynaSent combines naturally occurring sentences with sentences created using the open-source Dynabench Platform, which facilities human-and-model-in-the-loop dataset creation. DynaSent has a total of 121,634 sentences, each validated by five crowdworkers.

Source: DynaSent: A Dynamic Benchmark for Sentiment Analysis

Variants: DynaSent

Associated Benchmarks

This dataset is used in 1 benchmark:

Recent Benchmark Submissions

Task Model Paper Date
Sentiment Analysis GPT-4o Fine-Tuned (Minimal) ELECTRA and GPT-4o: Cost-Effective Partners … 2024-12-29
Sentiment Analysis GPT-4o-mini Fine-Tuned ELECTRA and GPT-4o: Cost-Effective Partners … 2024-12-29
Sentiment Analysis GPT-4o + ELECTRA Large FT (Prompt, Label, Examples) ELECTRA and GPT-4o: Cost-Effective Partners … 2024-12-29
Sentiment Analysis GPT-4o (Prompt) ELECTRA and GPT-4o: Cost-Effective Partners … 2024-12-29
Sentiment Analysis GPT-4o-mini + ELECTRA Large FT (Prompt, Label, Probabilities) ELECTRA and GPT-4o: Cost-Effective Partners … 2024-12-29
Sentiment Analysis GPT-4o-mini + ELECTRA Large FT (Prompt, Label) ELECTRA and GPT-4o: Cost-Effective Partners … 2024-12-29
Sentiment Analysis GPT-4o + ELECTRA Large FT ELECTRA and GPT-4o: Cost-Effective Partners … 2024-12-29
Sentiment Analysis GPT-4o-mini (Prompt) ELECTRA and GPT-4o: Cost-Effective Partners … 2024-12-29
Sentiment Analysis ELECTRA Large Fine-Tuned ELECTRA and GPT-4o: Cost-Effective Partners … 2024-12-29
Sentiment Analysis GPT-4o-mini + ELECTRA Base FT ELECTRA and GPT-4o: Cost-Effective Partners … 2024-12-29
Sentiment Analysis ELECTRA Base Fine-Tuned ELECTRA and GPT-4o: Cost-Effective Partners … 2024-12-29
Sentiment Analysis SVM Arabic Multi-Dialect Segmentation: bi-LSTM-CRF vs. … 2017-08-19

Research Papers

Recent papers with results on this dataset: