rehabib/Sentiment-Amharic-XLNet

# Sentiment Analysis of Amharic Product Reviews Using XLNet ## Abstract This research addresses sentiment analysis of Amharic product reviews using the XLNet transformer model. Amharic, a morphologically rich and complex language, poses unique challenges for natural language processing (NLP). By leveraging XLNet’s attention mechanism and augmentation techniques, we achieved a notable accuracy of **98.10%**, outperforming other models like BERT. ## Key Findings - **Model Used**: Custom XLNet model fine-tuned for Amharic sentiment analysis. - **Accuracy Achieved**: 98.10% (XLNet) vs. 90.79% (BERT). - **Significance**: Demonstrates the feasibility of applying transformer-based models to morphologically rich languages. ## Methodology - Data preprocessing techniques tailored to Amharic language characteristics. - Augmentation strategies (e.g., random insertion, deletion) for dataset variability. - Comparison of XLNet with BERT on the same dataset. - ## Contact For access to the full repository, including code and detailed results, please contact: - **Remla Habib** - **remla.habib@gmail.com** - GitHub: github.com