Sentiment Analyses for Kurdish Social Network Texts using Naive Bayes Classifier

Abstract

Language is a great tool to communicate and carry information. Moreover, it is used to express feeling and sentiment. These days sentiment analysis is one the most active field of research, to discover people's opinion about specific product, service or topic. The task of sentiment classification is to categories reviews of users as positive or negative from textual information of Social Networks like Facebook, Google+, Twitter and Blogs to determine the feeling of majority about specific topics. Kurdish language suffer from the unique and standard writing rules, grammar syntax and alphabet. Therefore, Kurdish people write their feeling in social networks in different ways. Some of them prefer to use the Arabic script style while others prefer to use Latin letters to express their feeling, further some people use their different accents and syntax and even sometimes they use English letters write their emotion. Therefore, for the purpose of analytics for Kurdish sentiment analyses its proposed to use data mining classification techniques such as Naive Bayes classifier because of its strong independence assumption. In Experimental results, the Social Network comments are classified into positive or negative polarities. The accuracy of sentiment analysis is obtained 66% by using Naive Bayes classifier for unigram feature on Kurdish text dataset.