Automatic expansion of manual email classifications based on text analysis

Abstract

The organization of documents is a task that we face as computer users daily. This is particularly true for management of email. Typically email documents are organized in directory structures, which reflect the users' ontology with respect to his daily communication needs. Since users' activities are continuously changing this may render email classifications increasingly inaccurate and manual maintenance is a painstaking task. In this paper we present an approach for integrating user-defined folder structures with classification schemes that have been automatically derived from the email content. This allows to automating the process of evolving and optimizing directory structures without sacrificing knowledge captured in manually created folder structures. A prototype demonstrating the feasibility and utility of our approach has been developed and evaluated. With our approach we address both an important practical problem and provide a relevant study on the application of various techniques for maintaining application specific ontologies. © Springer-Verlag Berlin Heidelberg 2003.