# Emotion Analysis with LSTM

This project implements an LSTM-based neural network for emotion analysis using the `emotion_sentimen_dataset.csv`.

## Features
- **Efficient Training**: Uses a subset (20,000 samples) of the large dataset for faster experimentation while maintaining good accuracy.
- **LSTM Architecture**: Embedding -> SpatialDropout -> LSTM -> Dense.
- **Deployment Ready**: Includes scripts for training and real-time prediction.

## Requirements
- Python 3.x
- TensorFlow / Keras
- Pandas, NumPy, Scikit-learn

## Files
- `train_emotion_lstm.py`: Script to preprocess data, build, train, and save the model.
- `predict_emotion.py`: Script to load the trained model and predict emotions from user input.
- `emotion_model.h5`: The saved trained model.
- `tokenizer.pickle`: Saved tokenizer for text processing.
- `label_encoder_classes.npy`: Saved label encoder classes.

## Usage

### 1. Training
To train the model (if you want to re-train):
```bash
python train_emotion_lstm.py
```
This produces `emotion_model.h5` and necessary artifacts.

### 2. Prediction
To use the model for prediction:
```bash
python predict_emotion.py
```
Type any sentence when prompted, and the model will classify its emotion.

## Performance
The model achieves high accuracy (>90%) on the test set even with a reduced dataset size, demonstrating efficiency.