Sentiment analysis study on artificial intelligence using twi̇tter data

Social media plays a significant role in our lives today. People share their feelings, thoughts, daily activities, political views, and much more on social media platforms. Therefore, it can be considered a vast source of data. In other words, social media provides a large amount of information about individuals, and this data can be used to draw inferences about society. These inferences can be obtained using natural language processing and sentiment analysis methods. Sentiment analysis can be performed using deep learning and machine learning methods. The data set used in the study consists of 14,657 Twitter data with the hashtag "#artificial intelligence". Within the scope of this study, the emotion of fear was also analyzed in addition to positive, negative and neutral emotions.SentiTurkNet emotion dictionary was used for labelling positive, negative and neutral emotions. The words labelled as fear in the NRC (Disgust, Fear, Trust, Joy, Anger, Anticipation, Surprise, Sadness) emotion dictionary, which has eight different emotion classifications, were adapted according to the dataset used in the study. Logistic Regression (LR), Naive Bayes (NB), Support Vector Machine (SVM) and Random Forest (RF) Algorithms, which are machine learning techniques, were used as sentiment analysis methods. Among the deep learning techniques, Long Short Term Memory(LSTM), Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) Algorithms were applied. The results obtained for the first data set are as follows; It was found to be 81,69% for LR, 82,78% for SVM, 78% for NB, 82,86% for RF, 83,80% for LSTM, 84,62% for CNN and 83,30%for RNN. The results obtained for the second data set are as follows; It was found to be 83,65% for LR, 89,16% for SVM, 78% for NB, 88,67% for RF, 90,25% for LSTM, 89,60% for CNN and 87,72% for RNN.