How reviews on a travel website shape public perception of hoteling brands using sentimental analysis: a case study on Booking.com?

Abstract

In today’s world, the internet has become a powerhouse for a huge source of data in the form of user opinions, emotions, views, and arguments about various brands, products, topics, and events. The sentiment shared by an individual about a brand has a massive influence on other readers and viewers who are constantly scrolling the internet to search for answers to individual queries and needs. However, these data from the internet generally remain in unstructured form for which sentiment analysis is very much important to give the data a good structure and to interpret this huge compilation of emotions so that they can be classified and then segmented into different classes in order to understand the major emotions associated with a brand. Till now, many kinds of research have been conducted using different techniques of sentiment analysis to classify textual data. In this study, sentiment analysis and deep learning techniques have been merged to perform the sentiment analysis of reviews from the popular booking site called Booking.com. Machine learning has the ability to differentiate text and topic and that can be measured through model evaluations. The deep learning model is used because it is effective for their automatic learning capability. The study proceeds with a topic modeling to identify the highlighting topics used in negative reviews from a customer’s point of view as they experience a stay in an accommodation. From topic modeling, this can be evaluated that topics like “room”, “staff”, “bathroom”, “bed” and “breakfast” are more highlighted topics that guests focus on when expressing negative emotions in the reviews. In addition, sentiment analysis is conducted using sentiment scores and then classified into further classes of best, good, bad, and worst and it is discovered that most of the hotels listed in the Booking.com site is good as the majority of the reviews are in the good and best classes which is about 64% of total number of 515,738 reviews. Machine learning classification models like Random Forest, SVM, Naïve Bayes are used to train the dataset and accuracy scores are evaluated. Along with this, CNN model is also trained which shows the highest accuracy of 90% when compared with the traditional Machine learning algorithms. Hence, it can be stated that the deep learning Convolutional Neural Network (CNN) model is able to outperform the regular machine learning algorithms. Lastly, the top 10 best and worst hotels are also identified based on the customer’s reviews. Booking.com should give suggestions to these worst hotels and come up with effective marketing plans to leverage them and the best hotels should also be announced in the website and application to make them more popular among the people.