A Tourist Spot Search Method using Similarity of Function based on Distributed Representations of User Reviews
Tomofumi Yoshida, Daisuke Kitayama, Shinsuke Nakajima and Kazutoshi Sumiya
Proc. of The 25th International MultiConference of Engineers and Computer Scientists (IMECS 2017),pp.473-478,2017年3月,査読有
論文PDF
概要
Touristspots have certain functions, such as “spot suitable for seeing the night view” or “spot suitable for meeting point.” In this paper, we propose a method for searching tourist spots that uses the similarity of their functional features. In our method, we extract distributed representations of a tourist spot as a paragraph vector of user reviews for the tourist spot. We extract distributed representations of location and category in the same way. Next, we extract the functional feature of a tourist spot by combining distributed representations of the tourist spot, locations, and categories. Finally, we search tourist spots using the extracted functional feature and distributed representations of other locations and categories. In this phase, the most important thingis the way in which locations and categories are represented. We can employ an extraction method using a paragraph vector for user reviews based on a location or a category (the conventional method). On the other hand, we can use the average vector of distributed representations of all spots in a location or a category. In this paper, we compared our method (average vector) with the conventional method (paragraph vector). By this experiment, we evaluated the effectiveness of our method for searching tourist spots.