Skip to Main Content
Recommender systems automatically determine suitable items for users. Although preferences or context of users have been widely utilized in order to evaluate the suitability of the items for users, the surrounding context have little been considered. Focusing on that many ordinary human beings voluntarily report their observations of the current situation of the world to microblogs, this paper proposes a recommender system which not only recommends suitable restaurants to users based on their preferences and context but also provides the surrounding context information reported to microblogs which will further affect the users' restaurant selection behaviors. In particular, considering that such influential surrounding context information in microblogs includes keywords related to restaurant assessment, we propose a method for automatically determining the keywords to extract the context information by analyzing online reviews, which have been gathered also from ordinary human beings over a long period of time. The experiments by using Twitter as microblogs and Tabelog, a popular online restaurant review site in Japan, to obtain online reviews, indicated that the influential context information can be extracted from Twitter with the highest recall of 93.3% by using the area-related keywords. Additionally using the restaurant-related keywords was effective in removing irrelevant information obtaining the precision of 15.9%.