Turkish Document Classification Based on Word2Vec and SVM Classifier

Abstract

In this study, Turkish texts belonging to different categories were classified by using word2vec word vectors. Firstly, vectors of the words in all the texts were extracted then, each text was represented in terms of the mean vectors of the words it contains. Texts were classified by SVM and 0.92 F measurement score was obtained for seven different categories. As a result, it was experimentally shown that word2vec is more successful than tf-idf based classification for Turkish document classification.