A hybrid architecture for text classification

SKIS, a prototype system that allows for the construction and use of text classification applications, is discussed. SKIS uses a combination of knowledge-based techniques, statistical techniques, morphological processing, and relevance feedback learning techniques to perform text classification. SKIS has been used to construct a prototype text classification application for the routing of customer service requests within customer support centers. The SKIS run-time architecture, the development and knowledge maintenance environment, and how SKIS is used are described. The benefits of combining knowledge-based and statistical techniques for text classification are discussed. SKIS is compared with other text classification systems.<<ETX>>