Emotions in online debates: Tales from 4Forums and ConvinceMe

It is increasingly common for people to debate over various topics through online debate forums. While it has been shown that participants' emotional states affect debate processes and outcomes, it is unknown how different types of emotions are represented in online debates and what correlations exist between the emotions and other aspects of the debates such as their debate topic. We conduct a large-scale analysis of the emotions in two online debate forums, namely, 4Forums and ConvinceMe. Specifically, we first develop an emotion recognition algorithm that uses multiple channels BLSTM with a feedforward attention mechanism, which outperforms the state-of-the-art emotion recognition algorithm. Next, we label the emotions of each comment in the selected 4Forums and ConvinceMe discussions and analyze various aspects of the emotion's influence in the online debates. We observe that certain types of emotions are more likely dependent on the debate topic, and the prevalence of different emotions is independent of the individual discussions. We also observe emotion contagion between a comment and the immediately previous comment. We investigate the emotions of different types of respondents are less likely to express joy when they disagree and more likely to express disgust when they attack or disrespect to others.