A deep learning and large group consensus based cruise satisfaction evaluation model with online reviews

Abstract

Cruise satisfaction assessment is pivotal for both passenger experiences and the enterprise's industry competitiveness. To do that, this research proposes a deep learning based cruise satisfaction evaluation model by integrating cruise online reviews and a large group consensus model. In order to acquire the dimensions influencing cruise passenger satisfaction, attribute extraction is employed. Unlike traditional methods that rely on word-based attribute representation, this method employs sentence-level information to comprehensively capture attributes that are of genuine concern to users, including implicit ones. Additionally, it utilizes the pre-trained language model, BART, to accurately predict sentiment polarity by treating sentiment analysis as a generation task based on online reviews from the CruiseCritic website. Subsequently, the proposed consensus model is implemented to determine the final satisfaction level, taking into consideration subgroup weights and an ideal solution-based feedback mechanism. This mechanism facilitates the provision of reasonable adjustment recommendations tailored to the context of online evaluations. Ultimately, the research findings contribute valuable insights to enhance service quality within the Royal Caribbean International cruise line.