Enhancing sentiment analysis in tourism reviews: A comparative study of algorithms in ASPECT-BASED SENTIMENT ANALYSIS and EMOTION DETECTION

Abstract

Information technology now enables utilizing online review data to support the Sustainable Development Goals (SDGs). However, traditional sentiment analysis often cannot capture the complexity of sentiment. This research aims to combine Aspect-Based Sentiment Analysis (ABSA) and emotion detection for a more in-depth analysis of tourism reviews in Palangka Raya City and compare the performance of various algorithms. Review data was taken from Google Maps and analyzed using BoW, LDA, NRC Emotion Lexicon, machine learning, and deep learning algorithms such as Logistic Regression (LR), Naïve Bayes (NB), Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Gradient Boosting (GB), Decision Tree (DT), and BERT. Most of the reviews are positive, with the dominance of the emotions of anticipation and joy. The combination of cross-validation with the best parameters from GridSearchCV resulted in the most significant increase in model accuracy. The SVM model performed better than other machine learning and deep learning algorithms, with accuracy and F1-score reaching 99.86%. The combination of ABSA and emotion detection improves the understanding of sentiment and emotion to support strategic decisions in tourism.