A corpus analysis of online news comments using the Appraisal framework

We present detailed analyses of the distribution Appraisal categories (Martin and White, 2005) in a corpus online news comments. The consists just over one thousand comments posted response to variety opinion pieces on website Canadian newspaper Globe Mail. annotated all with labels corresponding different framework. Analyses annotations show that are overwhelmingly negative, they favour two subtypes Attitude (Judgment Appreciation) third, Affect. paper contributes methodology for annotating Appraisal, results interaction negation, constructive (or not) nature comments, level toxicity found them. highly opinionated language is expressed as an objective (Judgement rather than emotional reaction (Affect). This finding, together interplay evaluative constructiveness can be applied automatic moderation