Dataset · google-research-datasets
GoEmotions: emotions in Reddit comments
google-research-datasets/go_emotions
58K Reddit comments labeled across 27 emotions plus Neutral for classification.
What's inside
English Reddit comment texts labeled with 27 emotion categories or Neutral. Raw data and a simplified version with ready train/val/test splits are available. The dataset is designed for multi-class and multi-label emotion classification.
What you can do with it
- Fine-tuning BERT/RoBERTa for emotion classification
- Evaluating multi-label emotion classification
- Training moderation or sentiment analysis models
- Tasks
- Text classification
- Languages
- English
- Size
- 100K–1M rows
- Downloads this month
- 21.4K
- Likes
- 285
Models trained on it
- cirimus/modernbert-large-go-emotionsText classification 14.9K
- Horizon-Labs/multilingual-emotions-smallText classification 1.6K
- cirimus/ultrabert-380M-go-emotionsText classification 1.5K
- cirimus/ultrabert-150M-go-emotionsText classification 1.4K
- logasanjeev/bert-emotion-classifierText classification 1.4K
- cirimus/modernbert-base-go-emotionsText classification 1.1K
- ChrisLalk/German-EmotionsText classification 1.1K
- manjunathshiva/opendecider-nanoZero shot classification 1.1K
How to load it
from datasets import load_dataset
ds = load_dataset("google-research-datasets/go_emotions")