dawit-senaber/amharic-ecommerce-data-extractor

# Amharic E-commerce Data Extractor *Transform Telegram commerce data into actionable business insights* ## Project Overview This solution addresses EthioMart's need to consolidate decentralized Telegram commerce channels by: 1. Extracting key entities (products, prices, locations) from Amharic messages 2. Analyzing vendor performance for micro-lending decisions 3. Creating a centralized vendor evaluation platform **Key Features**: - Telegram data ingestion pipeline - Fine-tuned Amharic NER models - Automated vendor scorecard - Model interpretability reports - Business intelligence reporting ## Business Impact ```mermaid graph LR A[Telegram Channels] --> B(Data Ingestion) B --> C[Entity Extraction] C --> D[Vendor Scoring] D --> E[Loan Decisions] E --> F[Increased Platform Engagement] ## File Structure ├── data/ # All project data │ ├── labeled/ # Labeled datasets (CoNLL format) │ ├── processed/ # Processed results and analytics │ └── raw/ # Raw scraped data │ ├── models/ # Trained ML models ├── reports/ # Generated PDF reports ├── results/ # Model evaluation outputs │ ├── scripts/ # Execution pipelines │ ├── data_ingestion.py │ ├── data_labeling.py │ ├── model_comparison.py │ ├── model_interpretability.py │ └── process_scraped_data.py │ ├── .gitignore ├── LICENSE ├── README.md # This document └── requirements.txt # Python dependencies ## Key Metrics Component Result Messages Processed 146 Entities Extracted 587 Best Model F1-Score 0.3571 Top Vendor Score 3.10 (ShegerOnline) ## Installation # Clone repository git clone github.com cd amharic-ecommerce-data-extractor # Create virtual environment python -m venv .venv source .venv/bin/activate # Linux/Mac .\.venv\Scripts\activate # Windows # Install dependencies pip install -r requirements.txt # Step-by-Step Execution 1. Data Collection …