Hotel selection method based on online evaluations has become a hot research topic. The existing models based on online ratings or reviews from one website have a disadvantage of information being definite and information amount being small. Therefore, this paper proposes a hotel selection model based on Probabilistic linguistic Term Set (PLTS) which integrates online ratings and reviews from multiple websites: (1) Unifying the rating information’s evaluation attributes among different websites based on the PLTS similarity calculation method, putting forward the transformation method of linguistic scale to unify the rating information’s evaluation scale among different websites; (2) Analyzing the sentiment of review texts and putting forward the aggregation model of user reviews based on different groups' risk attitudes; (3) Improving the linguistic scale function to introduce the unbalanced effect of positive and negative evaluations; (4) According to preference differences among different groups, putting forward the attribute weight calculation method and providing recommendation results for different groups. Take four hotels on TripAdvisor, Ctrip and Hostelworld websites for case studies. The results show that information can be used to a greater extent by integrating online ratings and reviews from multiple websites, thus providing consumers with more objective and reliable decision-making results.