Product-review Classification Combining Multiple Clustering Algorithms

Abstract

As product reviews accumulate more and more at online shopping sites, customers begin to have an increasing demand for analyzing reviews automatically. In some previous studies, clustering algorithms have been proved to be effective in grouping reviews. However, most of the existing systems are built based on a single clustering algorithm which might make the system fragile. In this paper, we have proposed a method to combine multiple clustering algorithms. Evaluation experiments have shown the effectiveness of our approach.