Deep Sentiment Analysis: A Case Study on Stemmed Turkish Twitter Data

Sentiment analysis using stemmed Twitter data from various languages is an emerging research topic. In this paper, we address three augmentation techniques namely Shift, Shuffle, and Hybrid to increase the size of training data; then use key types deep learning (DL) models recurrent neural network (RNN), convolution (CNN), hierarchical attention (HAN) classify Turkish for sentiment analysis. The performance these DL has been compared with existing traditional machine (TML) models. TML affected negatively by data, but improved greatly utilization techniques. Based on simulation, experimental, statistical results deeming identical datasets, it concluded that outperform respect both training-time (TTM) runtime (RTM) complexities algorithms; most important factors as well average rankings.