Multi-Class Document Classification Using Lexical Ontology-Based Deep Learning

With the recent growth of Internet, volume data has also increased. In particular, increase in amount unstructured makes it difficult to manage data. Classification is needed order be able use for various purposes. Since manually classify ever-increasing purpose types analysis and evaluation, automatic classification methods are needed. addition, performance imbalanced multi-class a challenging task. As number classes increases, so does decision boundaries learning algorithm solve. Therefore, this paper, an improvement model proposed using WordNet lexical ontology BERT perform deeper on features text, thereby improving effect model. It was observed that success increased when 11 general lexicographer files based synthesis sets, syntactic categories, logical groupings. used feature dimension reduction. experimental studies, word embedding were without Afterwards, Random Forest (RF), Support Vector Machine (SVM) Multi-Layer Perceptron (MLP) algorithms employed classification. These studies then repeated with reduction performed by WordNet. addition machine model, experiments conducted pretrained The results showed that, unstructured, seven-class, dataset, highest accuracy value 93.77% obtained our