Computing multi-dimensional trust by mining E-commerce feedback comments

Reputation-based trust models are widely used in e-commerce applications, and feedback ratings are aggregated to compute sellers' reputation trust scores. The “all good reputation” problem however is prevalent in current reputation systems - reputation scores are universally high for sellers and it is difficult for potential buyers to select trustworthy sellers. In this thesis, based on the observation that buyers often express opinions openly in free text feedback comments, we have proposed CommTrust, a multi-dimensional trust evaluation model, for computing comprehensive trust profiles for sellers in e-commerce applications. Different from existing multi-dimensional trust models, we compute dimension trust scores and dimension weights automatically via extracting dimension ratings from feedback comments.

Based on the dependency relation parsing technique, we have proposed Lexical-LDA (Lexical Topic Modelling based approach) and DR-mining (Lexical Knowledge based approach) approaches to mine feedback comments for dimension rating profiles. Both approaches achieve significantly higher accuracy for extracting dimension ratings from feedback comments than a commonly used opinion mining approach. Extensive experiments on eBay and Amazon data demonstrate that CommTrust can effectively address the “all good reputation” issue and rank sellers effectively. To the best of our knowledge, our research demonstrates the novel application of combining natural language processing with opinion mining and summarisation techniques in trust evaluation for e-commerce applications.