Sentiment Analysis and Stance Detection in Arab versus non-Arab English News Comments

This paper conducts sentiment analysis and stance detection on Arab and non-Arab Facebook comments associated with news articles likely to evoke anger, fear, sadness, and happiness emotions. The study found that both human and automated methods largely aligned in sentiment classification, assigning negative sentiment to comments on the articles evoking negative emotions and positive sentiment to discussions of the happiness-evoking article, yet they differed in the sentiment intensity values assigned to the comments associated with each emotion. Hyland’s (2005) stance model was also applied to explore stance or attitude towards the articles’ topics. The analysis revealed that both groups of commenters showed a predominant ‘against’ stance in discussions of topics evoking fear and sadness and a predominant ‘in favor of’ stance in the topic evoking happiness, but the two groups significantly differed in their stances towards the topic evoking anger, since it raised a controversial political issue. The paper recommends the use of the stance framework in future sentiment analysis for more accurate opinion mining results.