Preference mining and fuzzy inference for hotel selection based on aspect-based sentiment analysis from user-generated content

Abstract

User-generated content (UGC) is a promising tool to assist consumers in hotel decision-making in an online environment. Existing studies on extracting sentiment opinions based on dependency parsing in aspect-based sentiment analysis (ABSA) are limited and rarely consider the importance of keywords and risk preferences of consumers to positive and negative ratings. To fill this gap, this study develops a decision-making model to select hotels based on UGC. First, the TextRank method is used to determine criteria and keywords, and 11 rules are proposed to identify potential sentiment opinions about specific keywords based on dependent parsing in ABSA. Then, a criteria value calculation method is proposed by combining sentiment scores and aspect-level ratings, where prospect theory is employed to portray different risk preferences of consumers for positive and negative ratings. Next, criteria values are input into a fuzzy inference system (FIS), and 29 linguistic rules are proposed to model the non-linear relationship between criteria and hotel performance. The applicability of the proposed method is verified by ranking 11 hotels in Chengdu city on Tripadvisor.com. Comparative analysis is given to demonstrate the effectiveness of the proposed model. Management implications are provided for practitioners to improve hotel management quality.