ClintonBeyene/Amharic-NER-LLM-Model

This project aims to develop a robust Named Entity Recognition (NER) system to extract key entities such as products, prices, and locations from Amharic Telegram e-commerce messages. The project involves data ingestion, preprocessing, labeling, model fine-tuning, comparison, and interpretability. **Amharic Telegram E-commerce NER Project 🛍️** ===================================== **Overview** ------------ This project aims to develop a robust Named Entity Recognition (NER) system to extract key entities such as products, prices, and locations from Amharic Telegram e-commerce messages. The project involves data ingestion, preprocessing, labeling, model fine-tuning, comparison, and interpretability. **Table of Contents** ----------------- 1. Introduction 2. Dataset 3. Model 4. Results 5. Usage 6. Future Work 7. Team 8. License 9. Acknowledgments **Introduction** --------------- This project aims to develop a robust Named Entity Recognition (NER) system for the Ethiopian e-commerce market 🛍️. The system is designed to extract key business entities from Amharic text data, including product names, prices, and locations. **Dataset** ------------ * **Source**: Telegram e-commerce channels (e.g. @mertteka) * **Language**: Amharic * **Format**: Text data in conll format * **Size**: 10,000+ messages **Model** ------------ * **Model Type**: XLM-Roberta * **Task**: Named Entity Recognition (NER) * **Entity Types**: Product names, prices, locations * **Training Data**: Labeled dataset in CoNLL format **Results** ------------ * **Accuracy**: 99.99% * **Precision**: 1.00 * **Recall**: 1.00 * **F1-score**: 1.00 **Usage** ------------ 1. Clone the repository: `git clone github.com` 2. Install required libraries: `pip install -r requirements.txt` 3. Train the model: `jupyter notebook fine_Tune_XLM-Roberta.ipynb` 4. Evaluate the model: `jupyter notebook Fine_tuned_DistillBERT_1.ipynb ` **Future Work** ---------------- * Expand the model to handle images and documents * Explore the use of NER technology in other domains * Improve the model's performance on out-of-vocabulary words **Team** ---------- * Beabsira Yenezer(github.com) 🤖 * Clinton Beyene(github.com) 🤖 * Seifegebriel Mosisa …