A Hybrid Fuzzy-BERT Framework for Aspect-Based Sentiment Analysis of Restaurant Reviews in Terengganu

View at publisher Summarise with VIRAT
Authors: Ashma' Roosyidah Ibrahim, Siti Nurul Hayatie Ishak, Shaiful Bakhtiar bin Rodzman
Cited by: 0
References: 14

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The rapid growth of food tourism in Terengganu has amplified the role of online reviews in shaping customer perceptions and business sustainability. However, existing review platforms provide only coarse ratings and lack actionable insights at the aspect level, leaving restaurant owners unable to identify targeted areas for improvement. To address this gap, this study develops and evaluates a fine-tuned BERT-based model for aspect-based sentiment analysis (ABSA) of restaurant reviews in Terengganu. A dataset of over$\text{2, 0 0 0}$English-language reviews was collected and preprocessed using a hybrid pipeline that integrates fuzzy aspect extraction, rule-based sentiment cues, and BERT-based classification. To mitigate class imbalance, Random Oversampling was applied, resulting in a balanced dataset for robust training. Experimental evaluation across multiple train-test splits demonstrates that the proposed model achieves up to 98.42 % accuracy with precision, recall, and$\text{F 1}$-score consistently above 98 %. Comparative analysis with baseline approaches confirms that the fuzzy-enhanced BERT architecture is particularly effective in handling noisy, informal, and mixed-language reviews common in this context. Beyond methodological advances, the system provides fine-grained, interpretable insights that support restaurant owners in enhancing service quality, while also contributing to sustainable food tourism in Terengganu. This work highlights the value of combining deep learning with domain-aware preprocessing strategies for domain-specific sentiment analysis and offers practical implications for the hospitality and tourism industry.

Fields of Study & Subjects

Sentiment Analysis and Opinion Mining Topic Modeling Hate Speech and Cyberbullying Detection

Publication details

Journal / source
International Conference on Cognitive Systems and Computer Interaction (ICoSCI)
Pages
1–6
Publication date
15 January 2026
Language
EN