emegua19/amharic_ecommerce_extractor

# Amharic E-commerce Data Extractor - Interim Report ## Project Overview This repository contains the interim submission for the **10 Academy Artificial Intelligence Mastery Program**, focused on the **"Building an Amharic E-commerce Data Extractor"** project ( **18 June – 24 June 2025**). The project's objective is to build a system for **EthioMart** to consolidate **Telegram-based e-commerce data in Ethiopia**, extracting key entities like: - Product - Price - Location using **Named Entity Recognition (NER)**. The data will eventually support a **FinTech vendor scorecard** for micro-lending. ### This interim report covers: - **Task 1**: Data Ingestion and Preprocessing from Telegram channels - **Task 2**: Labeling a subset of the dataset in CoNLL format --- ## Interim Submission Details - **Date:** June 22, 2025, 20:00 UTC (11:00 PM EAT) - **Deliverables:** - GitHub repository with code for Tasks 1 and 2 - 1–2 page PDF summarizing data preparation and labeling at: `reports/interim_report.pdf` --- ## Folder Structure ``` /amharic\_ecommerce\_extractor ├── /src │ ├── data\_handler.py # Telegram scraping & preprocessing │ ├── labeler.py # CoNLL labeling │ ├── ner\_model.py # NER model logic (pending) │ ├── interpreter.py # Interpretability (pending) │ └── vendor\_analytics.py # Vendor scorecard logic (pending) ├── /utils │ ├── config.py # Config management │ ├── logger.py # Logging utility │ └── amharic\_processor.py # Amharic text normalization/tokenization ├── /data │ ├── /raw # Raw Telegram data │ ├── /processed # Preprocessed and labeled CoNLL data │ └── /metadata # Telegram post metadata ├── /models │ ├── /finetuned # Fine-tuned models (pending) │ └── /checkpoints # Training checkpoints (pending) ├── /tests │ ├── test\_data\_handler.py # Unit tests for scraping │ ├── …