---
title: Emotion Intensity Prediction using Transformer Based Models
emoji: 🤩
colorFrom: purple
colorTo: indigo
sdk: streamlit
sdk_version: 1.46.1
app_file: app.py
pinned: false
---

# Multitask Emotion Prediction Space

This Hugging Face Space hosts a deep learning model that predicts emotions and their intensities from text.
It utilizes a BERT-based architecture combined with lexicon features for enhanced performance.

**Features:**
- BERT-based text understanding.
- Integration of NRC VAD, NRC Emotion Lexicon, and NRC Hashtag Emotion Lexicon.
- Multi-task learning for emotion classification (joy, sadness, anger, fear) and intensity regression.

**How to use:**
Enter your text in the input box below and click "Predict Emotions" to see the model's output.

**Model Details:**
- Trained on dataset SemEval-2018 El-reg
- Uses `bert-base-uncased` from Hugging Face.
- `lex_dim`: 21 (number of combined lexicon features)

**Files included:**
- `app.py`: The Streamlit application code.
- `best_multitask_multilabel_model.pth`: Trained model weights.
- `*_scaler.pkl`: Joblib-saved feature scalers for lexicon features.
- `NRC-*.txt`: Lexicon data files.

---
Feel free to duplicate this Space and experiment!