Habtamu91/Amharic-Ecommerce-NER

Amharic E-commerce Named Entity Recognition (NER) This project focuses on building a Named Entity Recognition (NER) model tailored for Amharic-language e-commerce data. It aims to identify and extract key entities such as product names, brands, prices, quantities, vendor names, and locations from unstructured Amharic text. # Amharic E-Commerce Named Entity Recognition A production-ready NLP system that extracts products, prices, and locations from Amharic Telegram posts to power vendor financing decisions. ## Features - **Amharic NER Model**: Fine-tuned BERT model for Ethiopian e-commerce text - **Vendor Scoring**: Automated lending score calculation - **Web Interface**: Streamlit dashboard for easy interaction - **API Ready**: Can be deployed as a microservice ## Installation ```bash git clone github.com cd amharic-ecommerce-ner pip install -r requirements.txt Usage 1. Training the Model - python src/ner_pipeline.py 2. Here's your enhanced **professional README.md** with all requested additions, formatted for optimal GitHub presentation: ```markdown # Amharic E-Commerce Named Entity Recognition (NER) A production-grade NER system for Amharic e-commerce messages, with full pipeline from data collection to interpretable predictions. ## 🌟 Key Features - **Telegram Scraper** for Amharic e-commerce chats - **Custom BERT Tokenizer** for Amharic script - **SHAP/LIME Integration** for model explainability - **Vendor Analytics Dashboard** (see `src/vendor_analytics.py`) ## 📌 Project Overview **Domain**: Ethiopian e-commerce message parsing **Input**: Amharic Telegram messages **Output**: Structured entity extraction: | Entity | Examples (Amharic) | Examples (Transliterated) | |-----------|--------------------------|---------------------------| | VENDOR | የሱፍ ሱቅ, �ህል ሻይ ቤት | Yusuf Shop, Hilal Tea House | | PRODUCT | �ይን ጠጅ, ሽፋን | Wine, Traditional Cloth | | PRICE | 250 ብር, 15 ዶላር | 250 Birr, 15 USD | | LOCATION | መስቀል �ህዳሴ, ሰበር ሆቴል | Meskel Hidase, Saber Hotel | ## 🗂 Repository Structure ``` amharic-ecommerce-ner/ ├── data/ # Dataset directories │ ├── raw/telegram_messages.csv # Scraped messages │ └── labeled/ner_labeled_data.conll # 12,000 annotated sen …