Understanding and identifying the use of emotes in toxic chat on Twitch
Document Type
Article
Publication Date
1-1-2022
Abstract
The latest advances in NLP (natural language processing) have led to the launch of the much needed machine-driven toxic chat detection. Nevertheless, people continuously find new forms of hateful expressions that are easily identified by humans, but not by machines. One such common expression is the mix of text and emotes, a type of visual toxic chat that is increasingly used to evade algorithmic moderation and a trend that is an under-studied aspect of the problem of online toxicity. This research analyzes chat conversations from the popular streaming platform Twitch to understand the varied types of visual toxic chat. Emotes were sometimes used to replace a letter, seek attention, or for emotional expression. We created a labeled dataset that contains 29,721 cases of emotes replacing letters. Based on the dataset, we built a neural network classifier and identified visual toxic chat that would otherwise be undetected through traditional methods and caught an additional 1.3% examples of toxic chat out of 15 million chat utterances.
Identifier
85119059166 (Scopus)
Publication Title
Online Social Networks and Media
External Full Text Location
https://doi.org/10.1016/j.osnem.2021.100180
e-ISSN
24686964
Volume
27
Grant
NRF-2017R1E1A1A01076400
Fund Ref
Ministry of Science, ICT and Future Planning
Recommended Citation
Kim, Jaeheon; Wohn, Donghee Yvette; and Cha, Meeyoung, "Understanding and identifying the use of emotes in toxic chat on Twitch" (2022). Faculty Publications. 3418.
https://digitalcommons.njit.edu/fac_pubs/3418