6. Ulusal Yüksek Başarımlı Hesaplama Konferansı, Ankara, Turkey, 07 October 2020, pp.1-6
In the age of globalization where everything is data driven, complex data generation has increased. Internet is flooded with all types of textual data. Consequently, extraction of information from such bulks of data has become very important and text classification has made this task relatively easier. In this paper, deep CNN, deep LSTM and their hybrid versions are proposed for binary and multi-class classification of text data. Basic deep models are compared with the deep hybrid ones using two distinct datasets. Our study aims to find out whether the basic deep models are more effective in handling the task of classifying text than the deep hybrid models. The proposed models provide better accuracy and training time against each other, where the better one differs with respect to dataset. The evaluated results indicate that the deep hybrid models yield better results in binary classification with 89.43% accuracy. Whereas the basic deep models perform better in multi-class classification with 71.4% accuracy.