Text Augmentation to Overcome Data Limitations in Sentiment Analysis for Bahasa Indonesia

This paper focuses on the challenges of sentiment analysis in Natural Language Processing (NLP), with a special emphasis on limited data scenarios. Sentiment analysis, a critical component of text understanding, faces specific difficulties when available data are scarce, leading to problems such as overfitting or underfitting. This study explores ensemble learning techniques, POS Tagging, and text augmentation techniques to overcome these data limitations. The experimental results demonstrate that text augmentation with the prefixes "me-", "ter-", "ber-", and "di-" is effective relative to increasing data variety and quantity, which contributes to improved sentiment analysis model performance. The ensemble learning model achieved an accuracy of 91.29% with significant improvements in precision, recall, and F1-score.