Ranking Online User Reviews for Tourism Based on Usefulness

Publication Type:

Conference Paper


PCI '20: ACM Panhellenic Conference on Informatics, ACM, Athens, Greece (2020)


Machine Learning, natural language processing, Neural networks, Review helpfulness classification, User-generated content


The growth of Web 2.0 services has lead to an increase in the volume of user-generated content in the form of online user reviews. Web platforms offering users the ability to evaluate the services of hotels have increased in popularity, as a large percentage of travellers offer their feedback or read hotel reviews to assist their decision making process. Users usually do not have time to go through the sheer volume of available hotel reviews and would prefer to read the most useful ones, whereas review usefulness is subjective and depends on the reader's needs and preferences. Therefore, the need for automatically detecting hotel review helpfulness arises.

In this paper, we propose the use of features that capture both textual content and review metadata for predicting hotel review helpfulness of Greek and English reviews. A novel approach for representing text as a word embeddings-based vector is introduced and review association with certain hotel service aspects is mapped. Evaluating the performance of our approach using Machine Learning and Neural classifiers yields promising results for the review helpfulness classification task.

Download full text: