Evaluation of Content-based Tour Recommendation Systems

This paper reports preliminary experiments conducted to compare the performances of two content-based tour recommendation systems. These systems receive a tour review as a query, and then recommend related reviews written by others. To evaluate the relevancy of each review to the query, the two systems use the BoW (Bag-of- Words) model and the LSA (Latent Semantic Analysis) model, respectively. From the experiments, we confirmed that both the two models can recommend reviews which are related to the query review. In addition, we confirmed that while the BoW model tends to recommend reviews of tourist spots which are completely same as the query review, the LSA model tends to recommend information about more diverse tourist spots.