•  
  •  
 

Abstract

User-generated content (UGC) plays an increasingly important role for knowledge sharing among Internet users in recent years. Many travel sites encourage users to share their experiences and express their opinions on attractions, accommodations, restaurants, etc. The UGC about travel provide precious information to the staff in travel industry as well as the end users. Most of previous works in travel recommendation focus on identifying attractions/hotels or reviews that are useful to the customers by analyzing UGC. In this paper, we argue that reviews that are noteworthy for hotel management is critical to the success of hotels in the competitive travel industry. We employ two hotel managers to conduct an examination on Taiwan’s hotel reviews in Tripadvisor.com and they report that noteworthy reviews can be characterized by their content features, sentiments, and writing qualities. We propose three methods for representing content features. Through the experiments using tripadvisor.com data, we find that, in terms of content feature representation, LDA method achieves comparable performance to TF-IDF method with higher recall and much fewer features. In addition, all the three types of features are important in identifying noteworthy hotel reviews. Specifically, content features are shown to have the most impact on precisions, whereas sentiments and review qualities impact more on recalls.

Share

COinS