Toward an Information Quality Based Method for the Holistic Quality Assessment of Online Reviews

The extraction of relevant and high quality online reviews is a key to effective opinion mining processes. One of the difficulties of review quality assessment is the lack of a sound theory to guide the holistic assessment of the various dimensions of online reviews. Early attempts have explored supervised machine learning techniques to predict the helpfulness of online reviews. Nevertheless, the ground-truth of review helpfulness is often determined based on user-contributed helpfulness votes, which may suffer from the same quality problems of online reviews. A robust methodology for assessing the quality of online reviews should examine both the intrinsic (explicit and latent features) and the extrinsic factors. Unlike opinion mining research, there is a lack of benchmark dataset to evaluate the effectiveness of the quality assessment methods for online reviews. The proposed research project aims to addressing the aforementioned research problems regarding the quality assessment of online reviews.