rufaeleshetu/amharic-ecommerce-ner

Extracting key entities from Amharic Telegram e-commerce messages to power vendor analytics and micro-lending. # amharic-ecommerce-ner # Amharic E-commerce NER Extracting key entities from Amharic Telegram e-commerce messages to power vendor analytics and micro-lending decisions. ## 📌 Project Overview This project develops a Named Entity Recognition (NER) pipeline fine-tuned on Amharic text from Telegram-based vendors. Entities include: - 🛍️ Product names - 💵 Prices - 📍 Locations - 🚚 Delivery information - ☎️ Contact info Extracted entities are used to compute a **Vendor Lending Score** for micro-loan eligibility. --- ## 🗂 Project Structure amharic-ecommerce-ner/ ├── data/ # Raw + labeled Amharic text from Telegram ├── notebooks/ # Jupyter notebooks for each task ├── scripts/ # Preprocessing, training, scoring ├── outputs/ # Trained models, plots, etc. ├── reports/ # PDF summary and blog ├── README.md ├── requirements.txt └── .gitignore --- ## 🚀 Key Features - Fine-tuned transformer models: `xlm-roberta`, `bert-tiny-amharic`, `afroxlmr` - CoNLL-format labeling and Hugging Face training - Model interpretability using SHAP & LIME - Vendor activity analytics and lending scorecard --- ## 🛠️ Installation ```bash pip install -r requirements.txt 🧪 Run Training python scripts/train_ner.py 📊 Run Vendor Scorecard python scripts/scorecard.py 📄 License This project is open source and available under the MIT License. --- ### ✅ 2. `requirements.txt` ```txt transformers datasets pandas numpy scikit-learn seqeval matplotlib shap lime jupyter ✅ 3. .gitignore gitignore # Byte-compiled / optimized / DLL files __pycache__/ *.py[cod] *.so # Data and models *.csv *.pkl *.pt *.bin *.json # Jupyter notebook checkpoints .ipynb_checkpoints/ # Virtual environments .venv/ env/ venv/ # Output outputs/