Automatic text categorization of turkish news with machine learning and deep learning techniques

Categorization of text news is the process of determining the types of the news according to their contents. We construct two types of automatic learning process of the Turkish news on the same dataset. In the first of these two methods, the data was in the document form and classified with Support Vector Machines SVM, and in the second method, data is in the label format for classification with Recurrent Neural Network RNN. In the sample dataset, there are three categories of news including total 9000 data. Turkish characters are also included in creating a model for classification of Turkish news. After the training and testing of these models and obtaining the accuracy results categorizing news in text, we compare the accuracies of these two models. The results showed that the accuracy of RNN model, which is 0.98 better than the results of accuracy of SVM model that is 0.96. The results of the comparison between algorithms, accuracy, performance measures for RNN were better than SVM but the speed ratio of SVM model was faster than in RNN model. If a little time is required, we can use the SVM algorithm to classify the text. In case of high accuracy in the classification, we use the RNN classification method.