Sentiment analysis in arabic language using machine learning: Iraqi dialect case study

Abstract

People's perspectives about products, services, events and other persons are widely available on social media and they represent an important feedback for organizations and governments. In the last decade, with the increased amount of online posted opinions, sentiment analysis (SA) started gaining higher attention from researchers. The aim of SA, an application of natural language processing, is to analyze sentiments within text, and in the case of this work is to extract sentiments associated with positive or negative polarities. Arabic language is of high interest as it is one of the widely used languages. Unfortunately, Arabic SA has many challenges due the special characteristics of the Arabic language which limited the number of works in this area and those that concern the Iraqi dialect in particular. This paper proposes some enhanced classifiers based on hybrid machine learning algorithms combined with two proposed preprocessing steps. These preprocessing steps along with the proposed hybrid classifiers enhance the classification accuracy for the Iraqi dialect reviews which are collected from online movie reviews. The results show up to 12.5% enhancement in the classification performance as compared to the conventional preprocessing and classifiers.