Ontology-Based Text Classification into Dynamically Defined Topics

We present a method for the automatic classification of text documents into a dynamically defined set of topics of interest. The proposed approach requires only a domain ontology and a set of user-defined classification topics, specified as contexts in the ontology. Our method is based on measuring the semantic similarity of the thematic graph created from a text document and the ontology sub-graphs resulting from the projection of the defined contexts. The domain ontology effectively becomes the classifier, where classification topics are expressed using the defined ontological contexts. In contrast to the traditional supervised categorization methods, the proposed method does not require a training set of documents. More importantly, our approach allows dynamically changing the classification topics without retraining of the classifier. In our experiments, we used the English language Wikipedia converted to an RDF ontology to categorize a corpus of current Web news documents into selection of topics of interest. The high accuracy achieved in our tests demonstrates the effectiveness of the proposed method, as well as the applicability of Wikipedia for semantic text categorization purposes.