Classification of News Document in English Basedon Ontology

摘要

Aims: This paper aims to propose ontology method of news document classification. The common method of document classification is based on morphology of term, without considering the meaning. It is impact to the number of term-document and computational time. Furthermore, the performance is decrease, even though the number of training data is increase.Methodology: The main idea of ontology is to handle the similarity of terms that have different morphological form but the same meaning (synonym). The ontology is built using WordNet database to find similary of meaning among terms-document. The terms that have similar meaning are merged including their term frequency to be constructed in vector space model. After that, the unknown document is classified using cosine similarity measurement of the weight-term. The text document that is used is English news text in general topic, such as interest, money-fx, trade, and crude. The experiment is compared to the conventional method which is document classification without ontology.Results: Classification of news document can be implemented using cosine similarity method based on ontology. The performance measure of this method including precission, recall and f-measure has increased eventhough the number of terms is reduced.