eden82/Amharic-News-Classification-

# 📰 Amharic News Classification using AfriBERTa Welcome to the **Amharic News Classification using AfriBERTa** project repository! This project focuses on building an intelligent Natural Language Processing (NLP) system capable of automatically classifying Amharic news articles using the powerful **AfriBERTa transformer model**. The repository contains datasets, preprocessing scripts, model training pipelines, evaluation results, and research implementations developed collaboratively by our project team members. --- # 👥 Group Members | Group Member | Student UGR ID | | :--- | :---: | | **Shalom Mesfin** | `UGR/25453/14` | | **Benjamin Endale** | `UGR/25484/14` | | **Salem Mesfin** | `UGR/25407/14` | | **Yaikob Wasihun** | `UGR/25556/14` | | **Bereket Daniel** | `UGR/25430/14` | --- # 📌 Project Overview This project applies **Deep Learning** and **Transformer-based NLP techniques** to classify Amharic news articles into 6 different categories automatically: 1. **ሀገር አቀፍ ዜና** (National News) 2. **ስፖርት** (Sports) 3. **ፖለቲካ** (Politics) 4. **ዓለም አቀፍ ዜና** (International News) 5. **ቢዝነስ** (Business) 6. **መዝናኛ** (Entertainment) Using **AfriBERTa**, a multilingual transformer language model pre-trained specifically on 11 African languages (including Amharic), the system achieves state-of-the-art accuracy for sequence classification tasks on low-resource Amharic text. --- # 📂 Project Structure ``` /home/yaikob-wasihun/Desktop/NLP/ ├── Amharic_News_Dataset.csv # Dataset file containing 51,483 news articles ├── requirements.txt # Package dependencies ├── main.py # Main pipeline executor (training + evaluation + sample prediction) ├── predict.py # CLI inference utility for predictions on new Amharic texts └── src/ ├── __init__.py # Module entrypoint exposing modules ├── data_loader.py # CSV loader with inspection functions ├── preprocessing.py # Cleaning text, encoding labels, and PyTorch datase …