Skip to main content

Table 12 Relative differences of linguistic variables between comments predicted as hateful by XGBoost + All and those labeled as hateful in the ground truth

From: Developing an online hate classifier for multiple social media platforms

LIWC categoryRel. diff. (lower scores) (%)LIWC categoryRel diff. (higher scores) (%)
Parentha− 13.4Frienda+ 6.9
Quotea− 8.6Bodya+ 5.3
Dasha− 7.8Sweara+ 5.2
QMark− 5.6Sexuala+ 5.1
WC− 4.9Bio+ 4.7
Risk− 4.7Informal+ 4.6
anx− 4.5Anger+ 4.4
Work− 4.4Semic+ 4.0
Tone− 4.0Netspeak+ 3.6
  1. Relative difference is calculated as Cpredicted − Cground)/Cground, where C is a LIWC category
  2. aOutlier: > 1.5 interquartile ranges (IQRs) below the first quartile or above the third quartile