Review Mining from Online Media
Yang Liu · 2010
Online reviews are becoming an ever popular source of information. In this thesis, we study three important problems in the contexts of automatic review mining from online media, and propose a set of new techniques to address the challenges arising therein. Mining opinions and sentiments from reviews presents unique challenges that cannot be easily addressed by conventional text mining methods. Therefore, we first propose novel approaches that can provide a comprehensive understanding of the sentiments reflected in the reviews. Equipped with such approaches, we then develop models and algorithms that can use the extracted opinions and sentiments for predicting product sales performance. As a case study, we investigate how to predict box office revenues from Weblogs, which have recently received a lot of attention due to its high popularity. Orthogonal to the problem of identifying reviewer opinions, we consider how to automatically evaluate the helpfulness of reviews, and consequently develop novel methods to identify the most helpful reviews for a particular product. Properly used, we expect such models and algorithms to be highly helpful in various aspects of business intelligence.