Automated Identification of Verbally Abusive Behaviors in Online Discussions

Joksimovíc, Srécko; Baker, Ryan S.; Ocumpaugh, Jaclyn; Andrés, Juan; Tot, Ivan; Wang, Elle Yuan; Dawson, Shane

doi:10.18653/v1/w19-3505

“…Threads have been studied using preceding comments (Pavlopoulos et al, 2020;Karan and Šnajder, 2019), discussion titles (Gao and Huang, 2017), and counts for aggressive comments (Ziems et al, 2020). The position of a comment in a thread -start or end -has also been considered (Joksimovic et al, 2019). Finally, researchers have analyzed conversation graphs for topological indicators of abuse (Papegnies et al, 2017).…”

Section: Methods In Abusive Language Detectionmentioning

confidence: 99%

Enriching Abusive Language Detection with Community Context

Kurrek¹,

Saleem²,

Ruths³

2022

Preprint

1

0

View full text Add to dashboard Cite

Uses of pejorative expressions can be benign or actively empowering. When models for abuse detection misclassify these expressions as derogatory, they inadvertently censor productive conversations held by marginalized groups. One way to engage with nondominant perspectives is to add context around conversations. Previous research has leveraged user-and thread-level features, but it often neglects the spaces within which productive conversations take place. Our paper highlights how community context can improve classification outcomes in abusive language detection. We make two main contributions to this end. First, we demonstrate that online communities cluster by the nature of their support towards victims of abuse. Second, we establish how community context improves accuracy and reduces the false positive rates of state-of-the-art abusive language classifiers. These findings suggest a promising direction for context-aware models in abusive language research.Productive conversations containing slurs are common, and they take many forms (Hom, 2008). Research inspired by the #MeToo movement has focused on the detection of sexual harassment disclosures by victims (Deal et al., 2020), but this research has not been sufficiently integrated into the literature on abusive language detection. The distinction between actual sexist messages and messages calling out sexism is rarely addressed in the field (Chiril et al., 2020). A similar trend is seen with sarcasm. Humor and selfirony can be employed as coping mechanisms by victims of abuse (Garrick, 2006), yet they constitute frequent sources of error for state-of-theart classifiers (Vidgen et al., 2019). For example, the median toxicity score for language on transgendercirclejerk, a "parody [subreddit] for trans people", is as high as 90% (Kurrek

show abstract

“…Threads have been studied using preceding comments (Pavlopoulos et al, 2020;Karan and Šnajder, 2019), discussion titles (Gao and Huang, 2017), and counts for aggressive comments (Ziems et al, 2020). The position of a comment in a thread -start or end -has also been considered (Joksimovic et al, 2019). Finally, researchers have analyzed conversation graphs for topological indicators of abuse (Papegnies et al, 2017).…”

Section: Methods In Abusive Language Detectionmentioning

confidence: 99%

Enriching Abusive Language Detection with Community Context

Saleem¹,

Kurrek²,

Ruths³

2022

Proceedings of the Sixth Workshop on Online Abuse and Harms (WOAH)

2

0

View full text Add to dashboard Cite

Uses of pejorative expressions can be benign or actively empowering. When models for abuse detection misclassify these expressions as derogatory, they inadvertently censor productive conversations held by marginalized groups. One way to engage with non-dominant perspectives is to add context around conversations. Previous research has leveraged userand thread-level features, but it often neglects the spaces within which productive conversations take place. Our paper highlights how community context can improve classification outcomes in abusive language detection. We make two main contributions to this end. First, we demonstrate that online communities cluster by the nature of their support towards victims of abuse. Second, we establish how community context improves accuracy and reduces the false positive rates of state-of-the-art abusive language classifiers. These findings suggest a promising direction for context-aware models in abusive language research.

show abstract

“…Threads have been studied using preceding comments (Pavlopoulos et al, 2020;Karan and Šnajder, 2019), discussion titles , and counts for aggressive comments . The position of a comment in a thread -start or end -has also been considered (Joksimovic et al, 2019). Finally, researchers have analyzed conversation graphs for topological indicators of abuse (Papegnies et al, 2017).…”

Section: Methods In Abusive Language Detectionmentioning

confidence: 99%

Proceedings of the Sixth Workshop on Online Abuse and Harms (WOAH)

2022

1

0

View full text Add to dashboard Cite

We continue to emphasize the need for inter-, cross-and anti-disciplinary work on online abuse and harms. These include but are not limited to: NLP, machine learning, computational social sciences, law, politics, psychology, network analysis, sociology and cultural studies. Continuing the tradition started in WOAH 4, we invite civil society, in particular individuals and organisations working with women and marginalised communities who are often disproportionately affected by online abuse, to submit reports, case studies, findings, data, and to record their lived experiences. We hope that through these engagements WOAH can directly address the issues faced by those on the front-lines of tackling online abuse. Speaking to the complex nature of the issue of online abuse, we are pleased to invite Mona Diab, Murali Shanmugavelan, Gebre Gebremeske, Daniel Borkan, Lucas Dos Santos, Alyssa Lees, and Rachel Rosen to deliver keynotes. In addition to our invited keynotes, we received 47 submissions out of which 24 were accepted. Of the accepted papers, 20 were long papers and 4 were short papers. These papers will be presented in our poster session. We thank the reviewers for their dedication and efforts in providing in-depth and timely reviews. With this, we welcome you to the Sixth Workshop on Online Abuse and Harms. We look forward to a day filled with spirited discussion and thought provoking research!

show abstract

Automated Identification of Verbally Abusive Behaviors in Online Discussions

Cited by 9 publications

References 29 publications

Enriching Abusive Language Detection with Community Context

Enriching Abusive Language Detection with Community Context

Enriching Abusive Language Detection with Community Context

Proceedings of the Sixth Workshop on Online Abuse and Harms (WOAH)

Contact Info

Product

Resources

About