Improving Hungarian Text Categorization Using Domain-Specific Ontology

The aim of Text Categorization is to automatically assign documents to a set of predefined categories. The prevailing approach is making use of a collection of precategorized examples for the induction of a document classifier through learning methods. In this paper we introduce a method which combines state-of-the-art learning techniques with background knowledge. We have used KAON ontology for knowledge representation. We have developed a reasoning method which makes use of the relations in the ontology. Our experiments will show that the method substantially enhances the results of text categorization, it will be clear that a domain specific ontology can improve performance. The proposed method is applicable in the field of spam filtering, document reorganization and classifying news stories and e-mails.