Cyberbullying poses signicant threat to mental and physical health on its victims, leading to a worrisome social issue.
Previous research of automated cyberbullying detection on SNS is mainly textual-based, in which cyberbullying content
is identied through a set of textual features. Those methods are straightforward but are more likely to suer from the subjectiveness of the researcher. Moreover, each content is evaluated with same standard, not catering for individual's variance. Therefore, in this article we propose a automated cyberbullying detection method that utilises the parent-child relationship between comments to capture the reaction from a third party to detect cyberbullying comments. We were able to improve the eectiveness of cyberbullying detection using only publicly available data.