Leveraging IndoBERT and DistilBERT for Indonesian Emotion Classification in E-Commerce Reviews

2025-09-18

AI Summary

To address low sentiment classification accuracy on Indonesian e-commerce reviews, this study systematically evaluates back-translation and synonym replacement—two data augmentation strategies—for enhancing text representations in low-resource language settings, using IndoBERT and DistilBERT as base models. Experimental results show that IndoBERT significantly outperforms DistilBERT (80.2% accuracy); back-translation yields the most consistent performance gain (+3.1 percentage points), confirming its critical utility in low-resource scenarios, whereas synonym replacement delivers marginal improvement and ensemble methods confer no substantial benefit. Crucially, this work provides the first quantitative analysis of the interplay between pre-trained model selection and data augmentation in Indonesian e-commerce sentiment analysis. It establishes a reproducible methodology for fine-grained sentiment classification in low-resource languages, offering empirically grounded guidance for model architecture and data enhancement decisions.

Technology Category

• Natural Language Processing: Text Classification & Sentiment Analysis
• Data Mining & Knowledge Management: Semantic Web
• Machine Learning: Learning Preferences or Rankings

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Understanding emotions in the Indonesian language is essential for improving customer experiences in e-commerce. This study focuses on enhancing the accuracy of emotion classification in Indonesian by leveraging advanced language models, IndoBERT and DistilBERT. A key component of our approach was data processing, specifically data augmentation, which included techniques such as back-translation and synonym replacement. These methods played a significant role in boosting the model's performance. After hyperparameter tuning, IndoBERT achieved an accuracy of 80%, demonstrating the impact of careful data processing. While combining multiple IndoBERT models led to a slight improvement, it did not significantly enhance performance. Our findings indicate that IndoBERT was the most effective model for emotion classification in Indonesian, with data augmentation proving to be a vital factor in achieving high accuracy. Future research should focus on exploring alternative architectures and strategies to improve generalization for Indonesian NLP tasks.

Problem

• Enhancing Indonesian emotion classification accuracy
• Leveraging IndoBERT and DistilBERT models
• Applying data augmentation techniques for performance

Innovation

• IndoBERT and DistilBERT for emotion classification
• Data augmentation with back-translation and synonym replacement
• Hyperparameter tuning achieving 80% accuracy