Sentiment Analysis as a Predictor for Scores of Persuasive Essays

Dongkwang Shin1,*   

1Gwangju National University of Education


This study aimed to investigate to what extent sentiment analysis can be applied to tone analysis of persuasive essays as a predictor for scores of the essays. To this end, the sentiment analysis of the text mining program ‘Orange3’ was utilized. This study first looked at the effects of Vader's sentiment indices (positive, negative, neutral, and compound values) on the total score of positive and negative essays. As a result, in the positive essay, all the sentiment analysis results had a significant effect on the total score. In particular, it was confirmed that the compound value had the greatest influence as a predictor of the total score of the positive essay as well as the scores in the four sub-scoring domains—task performance, content, organization and language use. On the other hand, it turned out that there were no significant relationship between sentiment indices and scores of negative essays. These results suggest that even if learners write a persuasive essay with either positive or negative tone, they could expect a higher score when comparing the two opposite positions rather than the one-sided one. This was also confirmed to some extent through word cloud analysis. Of course, there were some limitations in this study as well but it showed the applicability of sentiment analysis to predict scores of persuasive essays.

Figures & Tables

Fig. 1. Architecture of the argumentation structure parser (Stab & Gurevych, 2017, p. 635)