Analyzing User-Generated Content in Yogyakarta Tourism Through Integrated TF-IDF, LDA, and Qualitative Descriptive Analysis

Abstract

User-generated content shapes how tourists choose a destination, yet Indonesian destination managers rarely turn it into usable evidence. This study identifies the lexical and thematic structure of tourist reviews for five destinations in Yogyakarta, and explains that structure through the accounts of tourism stakeholders. A convergent mixed-methods design was applied to 15,000 Google Reviews, of which 7,963 survived preprocessing. TF-IDF weighted the distinguishing words, LDA grouped them into latent topics, and two focus group discussions with twenty participants explained why those topics mattered. The models yielded two to four topics per site. Malioboro proved the most fragmented because parking conduct and vendor pressure formed a topic separate from night atmosphere. Three concerns recurred everywhere: interaction with local actors, destination management, and price. Lexical, thematic, and qualitative evidence can therefore guide destination development together.