Mining online hotel reviews using big data and machine learning: An empirical study from an emerging country

This paper presents a framework for collecting large datasets of hotel reviews (e.g., from Booking.com and TripAdvisor) and performing useful analytics from the data collected. This approach automates data collection, reduces manual effort, enhances data cleaning, and standardizes data processing. We compiled extensive datasets of 607,451 reviews from Booking.com and 782,584 from TripAdvisor, representing the most extensive emerging market-specific hotel review datasets. We conducted statistical analysis to evaluate the review distribution and customer satisfaction levels. Sentiment analysis assessed the polarity and subjectivity of English reviews and their impact on customers' overall satisfaction. Additionally, we used topic modeling with Latent Dirichlet Allocation (LDA) to identify key themes within the reviews to understand customers' real needs, providing helpful insights for hotel management. • Big datasets of customer online reviews are derived from Booking and TripAdvisor. • Review scores indicate customer satisfaction at each hotel star rating level, showing that budget accommodations often receive high satisfaction ratings. • Polarity and subjectivity enhance understanding of customer emotions, with most English reviews being positive and reflecting appreciation for good service. • The LDA intertopic distance map visualizes topics and their relationships, highlighting key customer priorities like excellent staff, comfortable rooms, convenient location, and cleanliness.