Learning Semantic Classes for improving Email Classification
Nicolas Turenne · 2003
In this paper we present a new term clustering method evaluated by a text classification task involving user profiles. The clustering algorithm is based on the extraction of graph patterns of terms in a training corpus. User profiles are then described by their areas of interest and the related term clusters. We demonstrate the utility of our approach in electronic mail classification. The e-mail folders represents the user areas of interest. They are described by the relevant term sets learned by clustering. A given message is routed to the relevant folder according to the distance between the