temeawoke/EthioMart-Extractor

based data ingestion and preprocessing pipeline to collect, normalize, and structure Amharic-language text and media from Ethiopian Telegram e-commerce channels. # EthioMart-Extractor based data ingestion and preprocessing pipeline to collect, normalize, and structure Amharic-language text and media from Ethiopian Telegram e-commerce channels. # EthioMart Telegram Channel Scraper ## Overview This project scrapes messages from selected public Telegram channels using the Telethon Python library. The data can be used for natural language processing tasks such as fine-tuning language models on Amharic text data. --- ## Features - Connects to multiple Telegram channels - Fetches recent messages (text only) - Saves data as JSON or Excel for further analysis - Easily extensible to add preprocessing steps --- ## Prerequisites - Python 3.7+ - Telegram API credentials (`api_id` and `api_hash`) - Installed packages (see `requirements.txt`) --- ## Installation 1. Clone the repository: git clone github.com cd ethiomart-telegram-scraper 2. Install dependencies: pip install -r requirements.txt 3. Obtain your Telegram API credentials from my.telegram.org. --- ## Usage 1. Update the `api_id`, `api_hash`, and `phone_number` in `scraper.py` file. 2. Run the scraper: python scraper.py 3. The messages will be saved in `telegram_messages.json`. --- ## Extending the Project - Add text preprocessing for Amharic language - Export data to CSV or other formats - Build a dataset for fine-tuning language models # Amharic Named Entity Recognition (NER) Labeling Project project_root/ │ ├── data/ │ ├── raw/ # Raw datasets (e.g., Excel, JSON) │ ├── processed/ # Preprocessed and cleaned data │ └── labeled/ # Labeled datasets in CoNLL format │ ├── notebooks/ │ └── amharic_ner_conll_labeling.ipynb # Annotation notebook │ └── data_ingestion_and_cleaning.ipynb # Annotation notebook │ ├── reports/ │ └── Amharic_NER_Labeling_Report.pdf # Generated PDF report │ ├── README.md └── requirements.txt ## Overview This project focuses on manually labeling a subset of Amharic text messages for Name …