Estimating Online Review Helpfulness with Probabilistic Distribution and Confidence

摘要

Product review helpfulness information is useful knowledge for consumers in their online shopping decision processes. Unlike the traditional method using the simple voting percentages, this paper proposes a new method for estimating the degrees of helpfulness with two features. One is to take into account the helpfulness distribution information on all reviews of concern in determination of helpfulness degrees; the other is to construct confidence intervals (CIs) of helpfulness to distinguish different reviews with the same voting percentage. Both synthetic and real data experiments, along with an illustrative example, reveal that the proposed method is superior to the traditional one in light of estimation accuracy.