Opinion Mining and Visualization of Online Users Reviews: A Case Study in Booking.com
Isidoros Perikos, Argyro Tsirtsi, Konstantinos Kovas, Foteini Grivokostopoulou, Ioannis Daramouskas, Ioannis Hatzilygeroudis · 2018
The growth of web applications and portals on hotel booking have led to an enormous amount of consumer generated comments and reviews on various hotels and travel services. In this paper, we present a work on the automatic analysis of user reviews on online hotel reviews with the aim to understand public opinions towards facilities and services they receive. We follow an aspect based approach where initially Latent Dirichlet Allocation is utilized to model topic opinions. The aspects specified indicate the important characteristics of the services and the facilities that users address in their reviews. After that, natural language processing approaches are used to analyze textual reviews, specify the dependencies on a sentence level and assist in understanding users opinions. At the final stage of understanding users opinions, several classifiers are trained under different feature sets extracted from the textual reviews and combined in ensemble schemas and their performance is examined on recognizing the polarity of the users' opinions. The results are quite satisfactory and indicate that features such as sentence dependencies assisted the classifiers in achieving accurate performance and that ensemble schemas perform robustly better than individual classifiers.