Proceedings of the International AAAI Conference on Web and Social Media · 2019 · 32 citations · 25 references
Abuse DetectionEngineeringPolitical ProcessPolitical BehaviorCommunicationControversial DiscussionsJournalismText MiningComputational Social ScienceSocial MediaPolitical CommunicationDiscussion ThreadsPolitical DiscussionsAbnormal BehaviorContent AnalysisArgument MiningSocial Medium MiningSocial ComputingPolitical AgendaSocial Medium DataArtsPolitical ScienceOpinion AggregationPublic Debate
Online communities like Reddit feature intensive discussions on political topics, engaging many users who give and receive votes for their posts and comments. Prior work has focused on detecting controversial discussions and analyzing the role of troll-like users. This paper provides a systematic analysis of the characteristics of user actions, like replies and votes, in discussion threads, identifying different patterns that refine the notion of controversies into disputes, disruptions and discrepancies. Most importantly, our study connects the dimension of user actions with a rich feature space comprising also the sentiments expressed in post contents and the topical variation across the posts and replies. We use this framework to gain insight on the traits of different archetypes of discussions, and to statistically test a variety of hypotheses on expected and abnormal behavior.
25
New Effect Size Rules of Thumb
Shlomo S. Sawilowsky · Journal of Modern Applied Statistical Methods · 2009 · 3.1K citations · Full text
From Word Embeddings To Document Distances
Matt J. Kusner, Yu Sun, Nicholas Kolkin et al. · PolyPublie (École Polytechnique de Montréal) · 2015 · 1.5K citations