Retweet communities reveal the main sources of hate speech

We address a challenging problem of identifying main sources of hate speech on Twitter. On one hand, we carefully annotate a large set of tweets for hate speech, and deploy advanced deep learning to produce high quality hate speech classification models. On the other hand, we create retweet networks, detect communities and monitor their evolution through time. This combined approach is applied to three years of Slovenian Twitter data. We report a number of interesting results. Hate speech is dominated by offensive tweets, related to political and ideological issues. The share of unacceptable tweets is moderately increasing with time, from the initial 20% to 30% by the end of 2020. Unacceptable tweets are retweeted significantly more often than acceptable tweets. About 60% of unacceptable tweets are produced by a single right-wing community of only moderate size. Institutional Twitter accounts and media accounts post significantly less unacceptable tweets than individual accounts. However, the main sources of unacceptable tweets are anonymous accounts, and accounts that were suspended or closed during the last three years.

[1]  Luís Torgo,et al.  How to evaluate sentiment classifiers for Twitter time-ordered data? , 2018, PloS one.

[2]  Jacob Cohen.  Statistical Power Analysis for the Behavioral Sciences , 1969, The SAGE Encyclopedia of Research Design.

[3]  Igor Mozetic,et al.  Cohesion and Coalition Formation in the European Parliament: Roll-Call Votes and Twitter Activities , 2016, PloS one.

[4]  Jean-Loup Guillaume,et al.  Fast unfolding of communities in large networks , 2008, 0803.0476.

[5]  Animesh Mukherjee,et al.  Spread of Hate Speech in Online Social Media , 2018, WebSci.

[6]  Tomaž Erjavec,et al.  Datasets of Slovene and Croatian Moderated News Comments , 2018, ALW.

[7]  Dirk Hovy,et al.  Hateful Symbols or Hateful People? Predictive Features for Hate Speech Detection on Twitter , 2016, NAACL.

[8]  Leon Danon,et al.  Comparing community structure identification , 2005, cond-mat/0505245.

[9]  Marco Zaffalon,et al.  Time for a change: a tutorial for comparing multiple classifiers through Bayesian analysis , 2016, J. Mach. Learn. Res..

[10]  Julio Gonzalo,et al.  A comparison of extrinsic clustering evaluation metrics based on formal constraints , 2009, Information Retrieval.

[11]  Kathleen M. Carley,et al.  Characterizing network dynamics of online hate communities around the COVID-19 pandemic , 2021, Appl. Netw. Sci..

[12]  Santo Fortunato,et al.  Community detection in networks: A user guide , 2016, ArXiv.

[13]  Ariadna Matamoros-Fernández,et al.  Racism, Hate Speech, and Social Media: A Systematic Review and Critique , 2021, Television & New Media.

[14]  Daniel Remondini,et al.  International expert communities on Twitter become more isolated during the COVID-19 pandemic , 2020, ArXiv.

[15]  Giulio Rossetti,et al.  A Novel Approach to Evaluate Community Detection Algorithms on Ground Truth , 2016, CompleNet.

[16]  Walid Magdy,et al.  Abusive Language Detection on Arabic Social Media , 2017, ALW@ACL.

[17]  Petra Kralj Novak,et al.  Community evolution in retweet networks , 2021, PloS one.

[18]  Alexander Dekhtyar,et al.  Information Retrieval , 2018, Lecture Notes in Computer Science.

[19]  Björn Ross,et al.  Measuring the Reliability of Hate Speech Annotations: The Case of the European Refugee Crisis , 2016, ArXiv.

[20]  Animesh Mukherjee,et al.  Hate begets Hate: A Temporal Study of Hate Speech. , 2019 .

[21]  知秀 柴田.  5分で分かる!? 有名論文ナナメ読み：Jacob Devlin et al. : BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding , 2020 .

[22]  Elizabeth F. Churchill,et al.  Profanity use in online communities , 2012, CHI.

[23]  Igor Mozetic,et al.  Multilingual Twitter Sentiment Classification: The Role of Human Annotators , 2016, PloS one.

[24]  J. E. Hirsch,et al.  An index to quantify an individual's scientific research output , 2005, Proc. Natl. Acad. Sci. USA.

[25]  Guido Caldarelli,et al.  Echo Chambers: Emotional Contagion and Group Polarization on Facebook , 2016, Scientific Reports.

[26]  Ralf Peters,et al.  Detecting Cyberbullying in Online Communities , 2016, ECIS.

[27]  M. Cugmas,et al.  On comparing partitions , 2015 .

[28]  Preslav Nakov,et al.  SemEval-2020 Task 12: Multilingual Offensive Language Identification in Social Media (OffensEval 2020) , 2020, SEMEVAL.

[29]  Virgílio A. F. Almeida,et al.  Characterizing and Detecting Hateful Users on Twitter , 2018, ICWSM.

[30]  S. Battiston,et al.  Sentiment leaning of influential communities in social networks , 2015 .

[31]  David Lazer,et al.  Sustained Online Amplification of COVID-19 Elites in the United States , 2020, Social Media + Society.

[32]  Tomaz Erjavec,et al.  TweetCaT: a tool for building Twitter corpora of smaller languages , 2014, LREC.

[33]  Ingmar Weber,et al.  Automated Hate Speech Detection and the Problem of Offensive Language , 2017, ICWSM.

[34]  Igor Mozetič,et al.  Stance and influence of Twitter users regarding the Brexit referendum , 2017, Computational social networks.

[35]  Lucas Dixon,et al.  Ex Machina: Personal Attacks Seen at Scale , 2016, WWW.

[36]  Klaus Krippendorff,et al.  Content Analysis: An Introduction to Its Methodology , 1980 .

[37]  Igor Mozetic,et al.  Retweet networks of the European Parliament: evaluation of the community structure , 2016, Applied Network Science.

[38]  John Pavlopoulos,et al.  Deeper Attention to Abusive User Content Moderation , 2017, EMNLP.