Ofgeha-Gelana/Amharic-NER-for-Telegram-E-commerce-Messages

# Amharic-NER-for-Telegram-E-commerce-Messages ## Project Overview Amharic-NER-for-Telegram-E-commerce-Messages centralizes e-commerce activities from various Ethiopian-based Telegram channels, enabling real-time data extraction to create a unified platform. This project focuses on: - Extracting real-time data (text, images, documents) from Telegram channels such as `@ZemenExpress`. - Fine-tuning a large language model (LLM) for Amharic Named Entity Recognition (NER). - Identifying key entities such as products, prices, and locations in Amharic text. This repository contains code for data ingestion, preprocessing, and labeling for NER tasks, supporting EthioMart’s vision of a seamless e-commerce platform. ## Features ### Real-Time Data Ingestion - A custom scraper connects to Telegram channels and fetches messages, images, and metadata. ### Text Preprocessing - Includes tokenization, normalization, and handling Amharic-specific linguistic features. ### NER Labeling - Annotates text data for NER tasks in the CoNLL format, labeling: - **Products**: `B-Product`, `I-Product` - **Prices**: `B-PRICE`, `I-PRICE` - **Locations**: `B-LOC`, `I-LOC` ## Requirements Install the required libraries using the following command: ```bash pip install -r requirements.txt