🎭 Polish Twitter Emotion Classifier
This model predicts emotions and sentiment in Polish text using a fine-tuned PKOBP/polish-roberta-8k model.
Detected labels:
- Emotions: 😊 radość (joy), 🤢 wstręt (disgust), 😠 gniew (anger), 🤔 przeczuwanie (anticipation)
- Sentiment: 👍 pozytywny (positive), 👎 negatywny (negative), 😐 neutralny (neutral)
- Special: 😏 sarkazm (sarcasm)
The model uses multi-label classification - text can have multiple emotions/sentiments simultaneously.
Replace @username with @anonymized_account
0 1
Example Input
Examples
Model Performance
Metric | Validation Score |
---|---|
F1 Macro | 0.85 |
F1 Micro | 0.89 |
F1 Weighted | 0.89 |
Subset Accuracy | 0.89 |
How to Use
- Enter Polish text: Paste a tweet, social media post, or any Polish text
- Select mode:
- Calibrated (recommended): Uses temperature scaling and optimal thresholds per label
- Default: Uses a single threshold for all labels
- Adjust settings: Toggle mention anonymization, adjust threshold (Default mode)
- Click Analyze: Get emotion and sentiment predictions with confidence scores
Prediction Modes
- Calibrated Mode (Recommended): Uses temperature scaling and label-specific optimal thresholds for better accuracy and calibration. This mode is recommended for most use cases.
- Default Mode: Uses sigmoid activation with a single threshold across all labels. Useful for quick predictions or when you want uniform threshold control.
Limitations
- Model is trained on Polish Twitter data and works best with informal social media text
- May not generalize well to formal Polish text (news, academic writing)
- Optimal for tweet-length texts (not very long documents)
- Multi-label nature means texts can have seemingly contradictory labels (e.g., sarkazm + pozytywny)
Citation
If you use this model, please cite:
@model{yazoniak2025twitteremotionpl,
author = {yazoniak},
title = {Polish Twitter Emotion Classifier},
year = {2025},
publisher = {Hugging Face},
url = {https://huggingface.co/yazoniak/twitter-emotion-pl-classifier}
}
📄 License
GPL-3.0 License