Categorisation for small-medium sized information systems - an exploration

Abstract

This thesis is an exploratory study, investigating the causes and mechanisms of categorisation problems in information systems. Looking at both cognitive functions in the brain and the context of information systems shows that categorisation is far from simple. Individuals vary so greatly as to make the design of a perfect categorisation scheme impossible. At the same time however, category structures in the mind are not arbitrary or random, and there are many commonalities between people. Hence a good categorisation scheme will find a balance between accommodating individual differences and encouraging conformity. The investigation process revealed a gap in the literature concerning assumptions about how to solve the problem. Most proposed solutions assume that an expert administrator is available to maintain category structures, that the items to be categorised will be primarily textual, and that the dataset will be very large. The reality is however, that these assumptions do not always hold true. This thesis proposes that by using tag clouds and clustering techniques, folksonomies can be adapted to suit smaller information systems where no dedicated administrator is available to maintain the category scheme. This was demonstrated through a number of experiments evaluating these approaches.