Enhancing Sentiment Analysis for Hausa Language Using Dataset Augmentation and Machine Learning Models

This study focuses on enhancing sentiment analysis for the Hausa language, a low-resource language, through dataset augmentation and machine learning models. Hausa faces significant challenges, including limited annotated datasets, linguistic complexity, and cultural nuances, which hinder the development of robust sentiment analysis systems. To address these issues, this research employs dataset augmentation techniques- synonym replacement and back-translation to create a larger, more diverse dataset. The augmented dataset is then used to train and evaluate various machine learning models, including traditional methods-Support Vector Machines (SVM) and Naive Bayes, as well as advanced deep learning models like BERT and LSTM. The results demonstrate that dataset augmentation significantly improves model performance, with BERT achieving the highest accuracy (90.22%) and F1-score (92.46%). Deep learning models consistently outperformed traditional methods, highlighting the importance of leveraging contextual embeddings and sequential learning for low-resource languages. The study also identifies challenges in handling sarcasm, irony, and regional dialectal variations, underscoring the complexity of sentiment analysis in Hausa. This research contributes to the advancement of NLP for Hausa by creating an augmented dataset, comparing machine learning models, and emphasizing the effectiveness of dataset augmentation. The findings provide valuable insights for future work in Hausa NLP and offer strategies adaptable to other low-resource languages, promoting inclusive language technologies and bridging the digital divide.