sumeyaaaa/-Amharic-E-commerce-Data-Extractor

Amharic E-Commerce Entity Extraction is a machine learning pipeline that scrapes Amharic Telegram e-commerce posts and fine-tunes a multilingual transformer model to extract key business entities like Product, Price, and Location, helping EthioMart become the central hub for Telegram-based digital commerce in Ethiopia. # -Amharic-E-commerce-Data-Extractor Amharic E-Commerce Entity Extraction is a machine learning pipeline that scrapes Amharic Telegram e-commerce posts and fine-tunes a multilingual transformer model to extract key business entities like Product, Price, and Location, helping EthioMart become the central hub for Telegram-based digital commerce in Ethiopia. This project is part of a data annotation and modeling pipeline for Amharic Telegram e-commerce channels. It includes data scraping, preprocessing, manual annotation in CoNLL format, and visualizations. -- ## 📁 Directory Structure of AMHARIC-E-COMMERCE-DATA-EXTRACTOR ``` ├── .github/ # GitHub actions and workflows ├── .venv/ # Python virtual environment ├── data/ │ ├── processed/ │ │ ├── conull.csv # Final labeled data in CoNLL table format │ │ ├── telegram_scraped_data_cleaned.csv # Cleaned Telegram messages │ │ └── top_30_messages_per_channel.csv # Top 30 messages per channel for annotation │ ├── raw/ │ │ ├── images/ # Downloaded product images │ │ └── telegram_scraped_data.csv # Raw scraped Telegram messages │ ├── models/ # Folder for storing fine-tuned NER models │ ├── notebook/ │ ├── task-1/ │ │ ├── normalization_and_tokenization.ipynb # Preprocessing pipeline │ │ ├── scrapper_session.session # Telethon session file │ │ └── scrapping.ipynb # Telegram scraping script │ ├── task-2/ │ │ ├── coNull.ipynb # CoNLL labeling and analysis │ │ └── conll_ready_tokenized.txt # Tokenized text for manual labeling │ ├── src/ │ ├── config.py # Channel list, phone, and output paths │ ├── pre_processing.py # Amharic text cleaning and normalization │ ├── scrapper.py # Telegram scraping with Telethon │ ├── coNLL.py …