Review selection based on topic models

The thesis addresses an important issue in e-commerce area that customers were overwhelmed by huge amount of online product reviews in attempting to find useful reviews. The thesis proposed a novel approach to represent the online reviews using features and related words identified by applying data mining and text mining methods, specifically, using probabilistic Topic Modelling methods. Then, based upon the derived related words, the thesis proposed a new method to select the most helpful reviews that discuss a specified feature of product