Datasets properties.

Properties of the three datasets: number of comments , number of different users giving comments , number of discussions/threads , average valence in the dataset , probability of finding positive, negative or neutral emotion (respectively , and ) and values of exponents for positive, negative and neutral clusters (respectively , and ). In case of Blogs data it was impossible to quantify the number of different users and note also a low number of comments in this dataset. Each data set has a different overall average valence – BBC is strongly negative, Digg is mildly negative while Blogs are mildly positive.