TY - GEN
T1 - Analyzing the traits and anomalies of political discussions on reddit
AU - Guimarães, Anna
AU - Balalau, Oana
AU - Terolli, Erisa
AU - Weikum, Gerhard
N1 - Publisher Copyright:
Copyright © 2019, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.
PY - 2019
Y1 - 2019
N2 - 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.
AB - 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.
UR - https://www.scopus.com/pages/publications/85070362915
UR - https://www.scopus.com/pages/publications/85070362915#tab=citedBy
M3 - Conference contribution
AN - SCOPUS:85070362915
SN - 9781577358060
T3 - Proceedings of the 13th International Conference on Web and Social Media, ICWSM 2019
SP - 205
EP - 213
BT - Proceedings of the 13th International AAAI Conference on Web and Social Media, ICWSM 2019
T2 - 13th International AAAI Conference on Web and Social Media, ICWSM 2019
Y2 - 11 June 2019 through 14 June 2019
ER -