An unsupervised method for information retrieval on E-commerce websites

摘要

In this paper, we present an unsupervised method to find the covert properties of the product on E-commerce web sites. Our method is generic because we do not have to depend on web site restricted pattern. The method works by 3 algorithms which are Page Type Recognition, List Page Clustering and Query Relationship Discovery. Experimental results show that it can acquire accurate rate of 96% and recall rate of 94%.