Akademik Veri Yönetim Sistemi
9th International Conference on Electrical and Electronics Engineering (ICEEE), Alanya, Türkiye, 29 - 31 Mart 2022, ss.387-391
As a well-known Natural Language Processing (NLP) task, text classification can be defined as the process of categorizing documents depending on their content. In this process, selecting classification algorithms and tuning classification parameters are crucial for efficient classification. In recent years, many deep learning algorithms have been used successfully in text classification tasks. This paper performed a comparative study utilizing and optimizing several deep learning-based algorithms. We have implemented deep neural networks (DNN), convolutional neural networks (CNN), long shortest-term memory (LSTM), and gated recurrent units (GRU). In addition, we performed extensive experiments by tuning hyperparameters to improve classification accuracy. In addition, we implemented word embeddings techniques to acquire feature vectors of text data. Then we compared our word embeddings results with traditional TF-IDF vectorization results of DNN and CNN. In our experiments, we used an open-source Turkish News benchmarking dataset to compare our results with previous studies in the literature. Our experimental results revealed significant improvements in classification performance using word embeddings with deep learning-based algorithms and tuning hyperparameters. Furthermore, our work outperformed previous results on the selected dataset.