An Approach for Reducing the Numeric Rating Bias

There is bias in customer reviews and the associated ratings. We propose a method to identify and reduce such bias on the part of reviewers. There are three phases in our approach. Firstly, we conduct Rating-aware sentiment analysis for each review text accompanied with the numeric ratings. Then, we extract features of review text and do manual labelling for training learning machine. At last, SVM is used to train the learner to reduce the rating bias. We applied our method on the dataset obtained from amazon.com. Results suggest that biased ratings have effects on customer ratings significantly in recommender systems and that this bias can be substantially reduced by our model.