Deriving customer preferences for hotels based on aspect-level sentiment analysis of online reviews

An increasing number of travelers like to share their experience and feelings about hotel stays through social media, generating a sheer volume of online hotel reviews. The user-generated comments contain their preferences for different aspects of hotels, which are helpful for hoteliers to improve hotels’ services. The key of deriving customer preferences from online hotel reviews is to identify fine-grained sentiment towards hotel attributes. However, the existing fine-grained sentiment analysis approaches cannot address the implicit aspect-level terms extraction very well, which is necessary to deal with the common situation that some aspects are omitted in the online reviews. To better understand customer preferences, we propose an unsupervised approach for aspect-level sentiment analysis with the implicit hotel attributes into consideration by integrating word embedding, co-occurrence and dependency parsing. A method based on overall sentiment values of hotel attributes is used to measure the customer preferences to support the hotel services analysis. Finally, online hotel reviews crawled from Ctrip.com are used to verify the proposed approach, and the results show that the hybrid approach outperforms the individual included techniques with respect to the sentiment classification performance. The analysis of customer preference for Dalian Bayshore Hotel suggests that the hotel’s facility should be upgraded urgently, and different types of customers pay different attention to hotel attributes, such as price, hygiene, and location.