In information delivery systems, users (pure client, hyperserver, and virtual proxy server) typically acquire push-delivery information according to vaguely defined categories or communities; furthermore, these users provide information from the same community or category. In this paper, an information push-delivery system is proposed that uses fuzzy information retrieval and fuzzy similarity measurement.

Using the authors’ system, each document is translated into a term vector form, and then its degree of similarity is measured with a centroid vector. According to the threshold value of degree of similarity, the document can then be classified into categories.

The paper starts with an introduction. The second section, “Review of Literature,” presents some concepts and relevant methods used to improve information retrieval efficacy on the Internet. Information push-delivery systems and fuzzy information retrieval algorithms are detailed. Delivery agents, a personal information push-delivery system, are presented. Section 3 is dedicated to building an architecture for such a system.

Section 4 details an experiment, using 108 students as subjects, that was done at the De-Yeh University in Changhua (Taiwan). Unfortunately, the details of the study (presented over more than five pages) are mainly in Chinese, not in English, so the authors’ point is hard to follow. The authors declare the degree of user satisfaction found to be around 71 percent; that is possible, but, be aware, this is the result of only one experiment. I don’t know if another application in other conditions would result in the same degree of satisfaction.