Sentiment Analysis of Azerbaijani twits Using Logistic Regression, Naive Bayes and SVM

In the work, the roadmap of sentiment analysis of twits in Azerbaijani language has been developed. The principles of collecting, cleaning and annotating of twits for Azerbaijani language are described. Machine learning algorithms, such as Linear regression, Naïve Bayes and SVM applied to detect sentiment polarity of text based on bag of word models. Our suggested approach for data processing and classification can be easily adapted and applied to other Turkish language. Achieved results from different machine learning algorithm have been compared and defined optimal parameters for the classification of twits.