Gebrehiwot-Tesfaye/Amharic-ecommerce-data-extractor

# Amharic E-commerce Data Extractor ## Project Overview This project implements a comprehensive Named Entity Recognition (NER) system for Amharic e-commerce data extraction from Telegram channels. The system is designed to support EthioMart's vision of becoming the primary hub for all Telegram-based e-commerce activities in Ethiopia. ### Business Context EthioMart aims to consolidate real-time data from multiple e-commerce Telegram channels into one unified platform, providing seamless customer experience for exploring and interacting with multiple vendors. This NER system extracts key business entities such as: - **Product Names/Types**: Specific product identifiers and categories - **Material/Ingredients**: Materials used in products - **Location Mentions**: Geographic locations and delivery areas - **Monetary Values/Prices**: Product pricing information - **Delivery Fees**: Transaction costs beyond product price - **Contact Information**: Phone numbers and Telegram usernames ### Key Objectives 1. **Data Collection & Preprocessing**: Automated ingestion from Telegram channels 2. **Data Labeling**: High-quality NER annotations for Amharic text 3. **Model Fine-tuning**: Transformer-based models for Amharic NER 4. **Model Comparison**: Systematic evaluation of different approaches 5. **Model Interpretability**: SHAP/LIME analysis for explainable AI ## Project Structure ``` Amharic-ecommerce-data-extractor/ ├── data/ # Data storage │ ├── raw/ # Raw scraped data │ ├── processed/ # Preprocessed data │ ├── labeled/ # Manually labeled data │ └── models/ # Trained models ├── src/ # Source code │ ├── data_collection/ # Telegram scraping modules │ ├── preprocessing/ # Data cleaning and preparation │ ├── labeling/ # Data annotation tools │ ├── models/ # NER model implementations …