Bias correction methodology for online product ratings using affective text mining and item response theory

Abstract

This work presents a methodology for bias correction of product ratings, specifically by considering affective text mining and item response theory (IRT). The novelties of this work are designing a methodology for bias correction in online product ratings, defining an experimental scenario using official data from Amazon Fine Food Reviews, and analyzing promising results obtained from applying the proposed methodology in the experimental scenario. Our experiments reveal that it is possible to conceive an automated method for bias correction in online product ratings using IRT and affective text mining, all the above within a unified methodology.