doffn/Building-an-Amharic-E-commerce-Data-Extractor

# 🛍️ EthioMart Amharic E-commerce NER Project A 10 Academy AI Mastery Week 4 challenge to build a multilingual **Named Entity Recognition (NER)** system that extracts business-critical entities from Amharic Telegram messages. This project supports **EthioMart's vision** to centralize e-commerce activity and enable smart vendor evaluation for micro-lending. --- ## 📌 Project Summary Telegram has become a powerful marketplace in Ethiopia. However, the lack of structure across vendor posts makes automation and analysis difficult. We aim to solve this by: * Extracting entities like **Product Name**, **Price**, and **Location** from unstructured Amharic messages. * Building a **Vendor Scorecard Engine** to help EthioMart assess business performance for **micro-lending** decisions. --- ## 📁 Project Structure ``` ├── data/ │   ├── raw/                 \# Scraped Telegram posts (text, images) |       ├── channels.txt      \# Channels used for scrapping |       ├── labeled\_data\_from\_df.conll   \# samples for labeled data |       ├── telegram\_data.csv            \# Scrapped Datas │ ├── models/               \# Ideally include the models ( too large) ├── notebooks/               \# Jupyter notebooks (EDA, training, interpretability) │   ├── task1\_2.ipynb        \# Notebook for task 1 and 2 │   ├── task3\_4ipynb.ipynb   \# Notebook for fine-tuning and model comparison │ ├── photos/                  \# Photos scraped ├── scripts/                 \# Python scripts for scraping, labeling, training, etc. │   ├── telegram\_scraper.py   \# Script to extract message from telegram channels │ ├── .github/workflows/       \# GitHub Actions CI/CD workflows │ ├── .env                     \# Environment variables (API keys, etc.) ├── requirements.txt         \# Project dependencies └── README.md                \# This file ```` --- ## ✅ Key Tasks & Goals ### Task 1: Data Ingestion & Preprocessing * Scrape messages from 5+ Telegram channels * Extract text, images, timestamps, …