kobeoseijnr/Few-Shot-Sentiment-Learning-with-Synthetic-Augmentation-in-Low-Resource-Settings

Few Shot Sentiment Learning with Synthetic Augmentation in Low Resource Settings # Few-Shot Sentiment Learning in Low-Resource Settings This project performs sentiment classification on Hausa language data using a few-shot learning setup enhanced with back-translation for data augmentation. ## Features - Multilingual Transformer Fine-tuning - Synthetic Augmentation via Back-Translation - Evaluation using Accuracy, F1, Confusion Matrix, BLEU, and BERTScore ## Setup ```bash pip install -r requirements.txt ``` ## Usage ```bash jupyter notebook few-shot-learning.ipynb ``` ## Structure - `few-shot-learning.ipynb` — Main notebook for training and evaluation - `data/` — Input dataset directory - `augmented/` — Back-translated data - `outputs/` — Evaluation results and plots ## License MIT License