Aspect Based Sentiment Oriented Hotel Recommendation Model Exploiting User Preference Learning

Abstract

Due to the advancement of the technology, people tend to focus on the online content related to products and services available through websites and the opinions of others which are provided in the form of reviews and comments. In the tourism domain, travelers are more concerned with the place of accommodation, facilities provided by a hotel, the location or the environment that the hotel is situated and tend to find the hotels that fulfill their requirements. They have to go through each and every review or comment in order to get a clear idea about a particular hotel, according to the opinion of the previous reviewers, which is a difficult and time-consuming task. Therefore, through this research, the users are provided with a system which analyses hotel reviews and provides aspect based personalized hotel recommendations that help users to easily find the best hotel according to their preferences, without having to go through a lot of reviews. For that, the proposed system was implemented with four steps, namely data gathering, data pre-processing, aspect extraction and sentiment analysis, and visualization of the output. In the implemented system, hotel reviews were analyzed and extracted the overall opinion of the reviews as opinion units with their related aspects. Based on the sentiment of those opinion units and the preferences of the user, the best hotels were suggested enabling the users to get an insight about the hotels. For this approach, aspect-based sentiment analysis was used. In addition to that, a weighted average calculation method was used for the final recommendations of the hotels, which suit users' preferences in an accurate manner.