Tourists' perceptions of Ubud as a world gastronomy destination: a modified ABSA-BERT analysis

Abstract

Although Ubud has been designated by the United Nations World Tourism Organization (UNWTO) as a global gastronomy tourism destination, existing research remains limited in assessing its strengths and challenges from a data-driven tourist perspective. Prior studies largely rely on qualitative, supply-side approaches and conventional sentiment analysis that overlook specific experiential aspects. Moreover, the integration of multi-platform data and the application of modified deep learning-based aspect-based sentiment analysis (ABSA) models remain underexplored in gastronomy tourism. This study addresses these gaps by employing a modified ABSA-bidirectional encoder representations from transformers (ABSA-BERT) framework on data from TripAdvisor and social media X to provide a more comprehensive evaluation of tourist perceptions. The model incorporates predefined aspect-based aspect term extraction (ATE), valence aware dictionary and sentiment reasoner (VADER) lexicon labeling, and a novel algorithm for extracting aspect-level opinions. The results demonstrate strong performance, achieving 93.50% accuracy on X data and 91% on TripAdvisor. Findings indicate an overall positive perception of Ubud as a global gastronomic destination, particularly for its authentic local cuisine and appealing natural dining atmosphere, although challenges such as overly spicy dishes and uncomfortable seating arrangements remain.