Contextual Recommendation based on Text Mining

The potential benefit of integrating contextual information for recommendation has received much research attention recently, especially with the ever-increasing interest in mobile-based recommendation services. However, context based recommendation research is limited due to the lack of standard evaluation data with contextual information and reliable technology for extracting such information. As a result, there are no widely accepted conclusions on how, when and whether context helps. Additionally, a system often suffers from the so called cold start problem due to the lack of data for training the initial context based recommendation model. This paper proposes a novel solution to address these problems with automated information extraction techniques. We also compare several approaches for utilizing context based on a new data set collected using the proposed solution. The experimental results demonstrate that 1) IE-based techniques can help create a large scale context data with decent quality from online reviews, at least for restaurant recommendations; 2) context helps recommender systems rank items, however, does not help predict user ratings; 3) simply using context to filter items hurts recommendation performance, while a new probabilistic latent relational model we proposed helps.