Incorporating reviewer and product information for review rating prediction

摘要: Traditional sentiment analysis mainly considers binary classifications of reviews, but in many real-world classification problems, non-binary review ratings are more useful. This is especially true when consumers wish to compare two products, both which not negative. Previous work has addressed this problem by extracting various features from the text for learning a predictor. Since same word may have different effects used reviewers on we argue that it necessary model such reviewer and product dependent order predict accurately. In paper, propose novel framework incorporate information into based learner rating prediction. The reviewer, modeled as three-dimension tensor. Tensor factorization techniques can then be employed reduce data sparsity problems. We perform extensive experiments demonstrate effectiveness our model, significant improvement compared state art methods, reviews with unpopular products inactive reviewers.