<p>Online reviews have become an important channel for people to share their experiences with tour groups. However, the complexity and uncertainty of the massive amount of online review information make it difficult for tourists to compare tour groups through personal browsing. Most studies usually focus on the semantic understanding of the review text content when extracting attributes, but they ignore the question of whether the review text content is credible. Therefore, based on 21,076 reviews of 6 tour groups from Ctrip.com, this paper proposes an online review-driven method considering credibility to assist tourists in choosing tour groups. Firstly, we use review credibility to measure the importance of words in reviews. Word2Vec and K-means are applied to cluster keywords and obtain attributes. Subsequently, sentiment analysis is conducted using Stanford CoreNLP. This analysis assesses the attitudes of reviewers towards each attribute, forming a nested probabilistic linguistic term set matrix to comprehensively retain the complex and multidimensional information present in the reviews. Finally, the EDAS-SIR method is proposed for ranking and selecting tour groups. Furthermore, comparative analysis with other methods demonstrates the good rationality and reliability of this method, which is applicable to other decision-making problems.</p>