Trigger Warning Assignment as a Multi-Label Document Classification Problem
37958Authors
nan
Abstract
AbstractA trigger warning is used to warn people about potentially disturbing content. We introduce trigger warning assignment as a multi-label classification task, create the Webis Trigger Warning Corpus 2022, and with it the first dataset of 1 million fanfiction works from Archive of our Own with up to 36 different warnings per document. To provide a reliable catalog of trigger warnings, we organized 41 million of free-form tags assigned by fanfiction authors into the first comprehensive taxonomy of trigger warnings by mapping them to the 36 institutionally recommended warnings. To determine the best operationalization of trigger warnings, we explore state-of-the-art multi-label models, examining the trade-off between assigning coarse- and fine-grained warnings, open- and closed-set classification, document length, and label confidence. Our models achieve micro-F1 scores of about 0.5, which reveals the difficulty of the task. Tailored representations, long input sequences, and a higher recall on rare warnings would help.
Keywords
Full Content
Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 12113–12134 July 9-14, 2023 ©2023 Association for Computational Linguistics Trigger Warning Assignment as a Multi-Label Document Classification Problem Matti Wiegmann 1 Magdalena Wolska 1 Christoper Schröder 2 Ole Borchardt 2 Benno Stein 1 Martin Potthast 2,3 1Bauhaus-Universität Weimar 2Leipzig University 3ScaDS.AI Abstract A trigger warning is used to warn people about potentially disturbing content. We introduce trigger warning assignment as a multi-label classification task, create the Webis Trigger Warning Corpus 2022, and with it the first dataset of 1 million fanfiction works from Archive of our Own with up to 36 different warnings per document. To provide a reli- able catalog of trigger warnings, we organized 41 million of free-form tags assigned by fan- fiction authors into the first comprehensive tax- onomy of trigger warnings by mapping them to the 36 institutionally recommended warn- ings. To determine the best operationalization of trigger warnings, we explore state-of-the-art multi-label models, examining the trade-off be- tween assigning coarse- and fine-grained warn- ings, open- and closed-set classification, docu- ment length, and label confidence. Our models achieve micro-F1 scores of about 0.5, which reveals the difficulty of the task. Tailored repre- sentations, long input sequences, and a higher recall on rare warnings would help.1,2 1 Introduction Media of any kind can address topics and situations that trigger discomfort or stress in some people. To help these people decide in advance whether they want to consume such media, so-called content warnings or trigger warnings can be added to them. Trigger warnings were originally used to help pa- tients with post-traumatic stress disorder. But after being picked up by various internet communities to also warn people tending to be “emotionally trig- gered” by a topic (e.g., to cry), the set of known trauma triggers has grown to include many more, such as abuse, aggression, discrimination, eating disorders, hate, pornography, or suicide. Today, the two terms are often used interchangeably, with “trigger” referring to the semantic cause. 1Code: https://github.com/webis-de/ACL-23 2Data: https://doi.org/10.5281/zenodo.7976807 Abuse Sexual abuse Child abuse Animal abuse Collective Death Kidnapping Violence Procedures Blood and gore Miscarriage Abortion Childbirth Pornography Incest Mental illness Addiction Suicide Self harm Dysmorphia Body shaming Ableism Religios Racism Sexism Classism Transphobia Homophobia Abusive Language A b u s e D i s c ri m i n a ti o n A g g r e s s i o n M e n t al H e al t h S e x u al M e di c al P r e g n a n c y Rare warnings (other) P h y si c al n at ur e of t h e h a r m P s y c h o l o g ic a l n a t u r e o f t h e h a r m N o in te nt t o h ar m A ct or h a s in te …
Content truncated to 500 words
Labeled Text
AbstractA trigger warning is used to warn people about potentially disturbing content. We introduce trigger warning assignment as a multi-label classification task, create the Webis Trigger Warning Corpus 2022, and with it the first dataset of 1 million fanfiction works from Archive of our Own with up to 36 different warnings per document. To provide a reliable catalog of trigger warnings, we organized 41 million of free-form tags assigned by fanfiction authors into the first comprehensive taxonomy of trigger warnings by mapping them to the 36 institutionally recommended warnings. To determine the best operationalization of trigger warnings, we explore state-of-the-art multi-label models, examining the trade-off between assigning coarse- and fine-grained warnings, open- and closed-set classification, document length, and label confidence. Our models achieve micro-F1 scores of about 0.5, which reveals the difficulty of the task. Tailored representations, long input sequences, and a higher recall on rare warnings would help.
Publication Details
- Paper ID:
- 37958
- Year:
- 2023
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan