ElshadayD/Sentiment-Analysis-for-Amharic-language

# Amharic Sentiment Classification with BERT This project fine-tunes **BERT (bert-base-multilingual-cased)** on the Amharic Sentiment Dataset to classify tweets as **positive** or **negative**. --- ## Features - Preprocessing and label mapping (`positive`, `negative`) - Fine-tuning BERT for sequence classification - Train/validation split with PyTorch `DataLoader` - Loss function: Cross-Entropy Loss (with padding ignored) - Optimizer: Adam - Custom **prediction function** for inference ---