Thousands of hotel reviews are available (TripAdvisor, Booking.com, Expedia), typically containing an overall (numerical) rating and a written review. Public ratings are an important factor in the demand of rooms in the hotel industry. Therefore, getting a good picture of the motivation behind a rating can be of great value.
DataQuest developed a customized web scraper to aggregate review data from Tripadvisor.com: Hotels in the Netherlands, Germany and the US were considered, 2250 hotels were scraped, over 1 million reviews were gathered.
The written hotel reviews are analyzed using sentiment analysis to reveal the opinion on hidden aspects such as the hotel’s service and location. Aspect ratings are often not (directly) visible, because the reviewers are only asked to give an overall rating. However aspect ratings can provide a lot of information. Using latent aspect rating analysis (LARA), the hidden aspects in the written review are extracted together with the relative weights the reviewer has placed on each aspect.
Please click here to read the PDF document on sentiment analysis of hotel reviews.