Sentiment Analysis of News Comments: A Comparison of Human and Automated Emotion Detection Methods

Document Type : Original Article

Author

Department of Al-Alsun, Faculty of Al-Alsun and Mass Communication, Misr International University (MIU)

Abstract

This paper conducts a sentiment analysis comparing human and automated sentiment annotation of Facebook comments associated with news articles likely to evoke the emotions of anger, fear, sadness, and happiness. The study finds that both human and automated methods assigned mostly similar sentiment polarities—negative for comments on the articles triggering anger, fear, and sadness, and positive for comments on the article evoking happiness. However, human annotators detected a wider range of emotional words, while the automated tool missed many of them and, at times, provided inaccurate descriptions of emotions. The study also employs Martin and White’s (2005) appraisal theory to examine the emotion-related language structures in these comments. It reveals that the affect dimension predominated in discussions of the sadness-related article, the judgment dimension was more prominent in discussions of the anger- and happiness-related articles, and the appreciation dimension featured more in discussions of the fear-related article.

Keywords

Files

Share

How to cite

Statistics