A Novel Product Features Categorize Method Based on Twice-Clustering

Abstract

Recently, the number of freely available online reviews is increasing in a high speed. More and more aspect based opinion mining technique has been employed to find out customers' opinions. In this paper, we only focus on categorize product features that the customers have commented on. An unsupervised twice-clustering based product features categorization method is proposed. Opinion words in context of product features are chosen to represent the interrelationship among product features instead of full context information. The cluster result of active product features is used as constraints to improve the whole categorization quality. Our experimental results show that opinion words in context and their group information are very important features in measuring the semantic similarity of their associated product features. The twice-clustering strategy achieves better performance than single-clustering method.