A Latent Dirichlet Allocation Technique for Opinion Mining of Online Reviews of Global Chain Hotels

Abstract

The hospitality industry has faced unprecedented challenges with the outbreak of Covid-19, which has changed customers' expectations. Therefore, it is essential to identify customers' new perceptions and expectations that lead to positive and negative opinions towards the service providers. Accordingly, this study aims to perform topic modeling and sentiment analysis on 94,200 online reviews of five global chain hotels in South Asia. Topic modeling as a text mining, unsupervised machine learning technique can decipher topics from a corpus such as online reviews, online reports, news covers, etc. In this study, the data is extracted from Trip Advisor through web scraping. Topic modeling is performed using the Latent Dirichlet Approach (LDA) on the extracted data set to analyze the key topics mentioned by the customers in the online reviews. The analysis depicted that cleanliness, food, staff, and service were the main concerns of the hotel guests. Furthermore, the findings represented that the main issues impacting the hotel guests were service delays. However, food and services were the keywords with the maximum word count as depicted by topic modeling.