Study of feature word extraction and cluster in Chinese product reviews

Evaluation system is the basis of product reviews mining. This paper introduces an unsupervised method to establish product evaluation system based on aspects. We extract product feature words with the syntax parser and achieve 72.33% F-value. This paper analyzes the mobile reviews, clusters the labeled feature phrases in the SemEval task and achieves 71.5% precision, which verifies the effectiveness of the method. Finally we make mobile features' clustering result visible and draw some conclusions by analyzing the relationship between different aspects.