Abstract:

［Purpose/Significance］Diet is one of the basic conditions for humansurvival and development. Exploring users' dietary preferences could explain their different dietary habits. Proposing a method for mining users' dietary preference could help broaden thedietary research routes. ［Design/Methodology］Users' reviews on "www.dianping.com" were used as data sources. By using sentiment analysis and keyword based vector space model, this paper developed a users' dietary preference model, which could be used to recommend restaurants for users. Moreover, an evaluation methodwas proposed to assess the effectiveness of users' dietary preference models. ［Findings/Conclusion］The model of users' dietary interests developed by their recent reviews could be used to predict their future dietary preference.Besides, recommending restaurant information based on users' dietary preferences can provide them with high quality information of their interests, and also meet their personalized dietary needs. ［Originality/Value］This paper explores users' dietary preference according to user generated contents in order to meet their personalized favorites from the perspective of sentiment analysis. In addition, the proposed assessment method of users' interest model can effectively evaluate the validity of models.

Key words: Dietary preference mining, Review mining, Sentiment analysis, Modeling of users'  interests, Information recommendation