Twitter Sentiment Analysis for Kurdish Language

Abstract

Sentiment analysis of text data has received a significant attention throughout Natural Language Processing stages. However, most of the focus has been on English language depriving many other languages from taking advantage of the state-of-the-art techniques most suitable to a particular language especially the Kurdish Sorani language. This paper is an attempt to bridge the gap between English and Kurdish language in sentiment analysis for social media text.  For this purpose, firstly a new Kurdish sentiment analysis dataset was curated and annotated then we tried different combinations of machine learning algorithms including classical machine learning algorithms such as Random Forrest, KNN, SVM, Naive Bayes bias and Decision trees and compared the results to Deep Learning techniques namely ANN, LSTM and CNN. In our experiments Naïve Bayes achieved the best results achieving an 78% accuracy.