Mining meaning from online hotel reviews: identification of determinants of (dis-) satisfaction using Latent Dirichlet allocation

Abstract

User-generated content (UGC) is increasing, so it is critical to clearly understand what customers think about the positive and negative aspects of specific goods. Such information is crucial for customers as well as for businesses. Online reviews are a powerful tool for a deep comprehension of customer satisfaction and dissatisfaction. This research intended to recognize the sources of satisfaction and dissatisfaction among business travelers in city hotels in Serbia. A total of 5,596 online reviews from 120 city hotels in Serbia were collected. Keyword-based visualization was used to display the most frequent words related to the experience during the hotel stay, reflecting the satisfaction and dissatisfaction of consumers. Latent Dirichlet allocation (LDA) was used for grouping positive and negative reviews into various topics. Results showed that business travelers' primary satisfaction sources are ?Conference facilities?, ?Atmosphere?, ?Amenities?, ?Accessibility?, and ?Location?. On the other hand, the sources of dissatisfaction are ?Dirtiness? and ?Internet Connection?. Based on the obtained results, hotel managers can direct their activities to improve the spheres that have proven to be the sources of dissatisfaction for customers.