helina1abebe/Amharic-Ecommerce-Extractor

# **Amharic E-commerce Data Extractor** ## **Project Overview** The **Amharic E-commerce Data Extractor** is a tool designed to transform unstructured Telegram posts from Ethiopian e-commerce channels into structured, machine-readable data. This data is used to build a centralized platform for EthioMart, enabling better vendor analysis and decision-making for micro-lending opportunities. --- ## **Key Features** - Scrape data (text, images, metadata) from Telegram channels. - Preprocess Amharic text for Named Entity Recognition (NER). - Fine-tune pre-trained models for Amharic NER tasks to extract entities like: - Product names and types. - Prices and monetary values. - Location mentions. - Compare multiple models for performance and interpretability. - Create a vendor scorecard for business analytics. --- ## **Project Structure** ```plaintext Amharic-Ecommerce-Extractor/ ├── data/ # Contains raw and labeled datasets ├── models/ # Saved fine-tuned models ├── scripts/ # Python scripts for various tasks │ ├── scrape_telegram.py # Script to scrape Telegram data │ ├── preprocess_data.py # Script to preprocess and clean data │ ├── label_data.py # Script for annotating data │ ├── train_model.py # Script to fine-tune the NER model │ ├── evaluate_model.py # Script to evaluate and compare models │ ├── vendor_analytics.py # Script to generate vendor metrics and scorecard ├── reports/ # Generated reports and visualizations ├── requirements.txt # List of Python dependencies └── README.md # Project documentation