Structuring Opinions by Relative Characteristics for User-Opinion Aggregation

Abstract

We propose a scoring function for aspect ranking that facilitates user-opinion aggregation. The function uses the log-likelihood ratio to capture relative characteristics between opinions. Review documents contain many variant expressions because they are written by different web users. We also propose a re-ranking method that identifies variant expressions having the same meaning by using several clustering techniques. The experimental results indicate that the proposed scoring function and the re-ranking method are more effective than baseline methods.