In this paper, we propose a method to recommend to a tourist (user) such a travel destination that is little known to many people, but of interesting for the user. To this end, we use two recommendation techniques, i.e. collaborative filtering and content-based filtering. We use the collaborative filtering method to predict the user's preference and select a destination that is well known and of interesting for the user. Then, with the destination as a clue, we make a final recommendation by finding out such a destination that is similar to the clue, but not well known itself by means of the content-based filtering method. To characterize travel destinations, we focus on many pieces of word-of-mouth information about them on the Internet, and use tf-idf values of keywords appearing in them to construct feature vectors for destinations. We conduct a user study and show that the proposed method is promising.