Purpose
This paper aims to develop a model to predict online review ratings from multiple sources, which can be used to detect fraudulent reviews and create proprietary rating indexes, or which can be used as a measure of selection in recommender systems.
Design/methodology/approach
This study applies machine learning and natural language processing approaches to combine features derived from the qualitative component of a review with the corresponding quantitative component and, therefore, generate a richer review rating.
Findings
Experiments were performed over a collection of hotel online reviews – written in English, Spanish and Portuguese – which shows a significant improvement over the previously reported results, and it not only demonstrates the scientific value of the approach but also strengthens the value of review prediction applications in the business environment.
Originality/value
This study shows the importance of building predictive models for revenue management and the application of the index generated by the model. It also demonstrates that, although difficult and challenging, it is possible to achieve valuable results in the application of text analysis across multiple languages.
