Application of Sentiment Analysis to Improve Online Hotel Reviews

Abstract

With the rapid growth of travel and e-commerce industries, online accommodation booking becomes very popular among travelers. Existing online travel sites do not only provide booking service, but also allow users to write and read reviews of the accommodation. Reviews on major websites, generally consist of a short text and an overall rating of the accommodation. Unfortunately, these formats are time-consuming to read and may create ambiguity as rating does not able to express the exact experience (Elango and Narayanan, 2014). Furthermore, users with specific preferences, such as a good location or service, will need to compare the reviews individually. Current approaches from several booking websites to allow manual input of specific aspects’ rating are proven to be inefficient with a very few percentage of user inputted reviews contain ratings of each aspects of the hotel. Therefore, this project develops a web application which allows users to view rating of certain aspects of the hotel and compare based on their preferences. This is developed with sentiment analysis, which is defined as an algorithmic process of identifying opinion to see user’s attitude towards subject expressed in text (Martin, 2018). This technique utilizes natural language processing and machine learning to extract and identify sentiment of a given aspect from existing user inputted reviews of the hotel. The processes include data processing, text processing, aspect extraction, web scrapping, feature extraction, imbalanced class processing, classification, as well as extensive evaluation procedures for the predicted classification results. The sentiment classification process includes extensive comparison and evaluation of different extraction models, processing models, and machine learning models to achieve the most credible results. Furthermore, the sentiment classification results of the processed reviews and ratings are showcased through a web application which also acts as a demo to show the improvement from current booking application. This improvement aims to simplify and improve current online hotel reviews by allowing users to compare and sort hotels not only based on the overall aspect, but also based on their own preferences in a minimum time possible.