An initial evaluation of automated organization for digital library browsing

Abstract

In this article we present an evaluation of text clustering and classification methods for creating digital library browse interfaces, focusing on the particular case of collections made up of heterogeneous metadata records. This situation is common in "portal" style digital libraries, which are built by harvesting content from many disparate sources, typically using the Open Archives Protocol for Metadata Harvesting (OAI-PMH). By studying the activity of users in an experimental system, we find that taxonomies built or populated using machine-learning (or "AI") techniques provide a potentially useful avenue for browsing in this digital library scenario.