Multi-task Chinese aspect-based sentiment analysis framework for service improvement: a case study on BNB reviews

User-generated content has become an essential factor influencing customer choice in the selection of hotel and bed and breakfast (BNB) accommodation. It is important for entrepreneurs to effectively analyze customer reviews. From the perspective of BNB managers, it is necessary to know customer opinions relating to specific aspects of the services they have received. Aspect-based sentiment analysis (ABSA) evaluates customer sentiments and opinions and can be used to improve services. We propose a framework for automated ABSA of reviews. The framework contains modules for data preprocessing, a multi-task Chinese aspect-based sentiment analysis module, and a Kano module. The Kano module divided service attributes into several distinct categories including must-be, one-dimensional, and attractive attributes & mldr; and so on. This module is integrated into our analytical framework to elucidate the relationships between hotel service offerings and customer attributes. The framework is used for a dataset crawled from Google Maps hotel reviews with aspect categories labeled by domain experts. The experimental results demonstrate the excellent performance of this method. We categorize customer requirements based on the aggregate consumer preferences estimated by the Kano module. The effectiveness of the proposed framework is demonstrated in an empirical evaluation and the utility of the proposed Kano module demonstrated.