Proceedings of the 11th ACM Workshop on Artificial Intelligence and Security 2018
DOI: 10.1145/3270101.3270103
|View full text |Cite
|
Sign up to set email alerts
|

All You Need is "Love"

Abstract: With the spread of social networks and their unfortunate use for hate speech, automatic detection of the la er has become a pressing problem. In this paper, we reproduce seven state-of-the-art hate speech detection models from prior work, and show that they perform well only when tested on the same type of data they were trained on. Based on these results, we argue that for successful hate speech detection, model architecture is less important than the type of data and labeling criteria. We further show that a… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
2
1

Citation Types

0
57
0
3

Year Published

2019
2019
2024
2024

Publication Types

Select...
3
3
1

Relationship

0
7

Authors

Journals

citations
Cited by 161 publications
(70 citation statements)
references
References 18 publications
0
57
0
3
Order By: Relevance
“…Generalisability of a model has also come under considerable scrutiny. Works such as Karan andŠnajder (2018) and Gröndahl et al (2018) have shown that models trained on one dataset tend to perform well only when tested on the same dataset. Additionally, Gröndahl et al (2018) showed how adversarial methods such as typos and word changes could bypass existing state-ofthe-art abusive language detection systems.…”
Section: Previous Workmentioning
confidence: 99%
See 1 more Smart Citation
“…Generalisability of a model has also come under considerable scrutiny. Works such as Karan andŠnajder (2018) and Gröndahl et al (2018) have shown that models trained on one dataset tend to perform well only when tested on the same dataset. Additionally, Gröndahl et al (2018) showed how adversarial methods such as typos and word changes could bypass existing state-ofthe-art abusive language detection systems.…”
Section: Previous Workmentioning
confidence: 99%
“…Works such as Karan andŠnajder (2018) and Gröndahl et al (2018) have shown that models trained on one dataset tend to perform well only when tested on the same dataset. Additionally, Gröndahl et al (2018) showed how adversarial methods such as typos and word changes could bypass existing state-ofthe-art abusive language detection systems. They also observed unimpressive results when using ULMFiT (Howard and Ruder, 2018) for abusive language detection, but argued that model architecture is less important than the type of data and the annotation scheme.…”
Section: Previous Workmentioning
confidence: 99%
“…"Es sei hervorgehoben, dass Internet-Ausfälle geografisch überall auftreten und weder mit den AfD-Likes auf Facebook noch mit der Anzahl flüchtlingsfeindlicher Delikte von Hass-Kriminalität in Zusammenhang stehen" kommentieren die Autoren ihr Vorgehen [9]. 3 Mit diesem Verfahren wurde nachgewiesen, dass die Steigerung der Hass-Kriminalität in Zeiten von mehr Hass-Kommentaren auf Facebook an Orten und Zeiten von Internet-Ausfällen völlig ausblieb. Man konnte weiterhin zeigen, dass der Effekt auf die Facebook-Nutzung und nicht auf die Internet-Nutzung im Allgemeinen zurückging.…”
Section: Facebook Und Die Afdunclassified
“…Man konnte weiterhin zeigen, dass der Effekt auf die Facebook-Nutzung und nicht auf die Internet-Nutzung im Allgemeinen zurückging. 4 Für einen kausalen Effekt von Facebook auf die Auswirkungen von Hass-Kommentaren im Sinne der Befeuerung von Hass-Kriminalität spricht weiterhin der Befund, dass landesweite Facebook-Ausfälle zu einem Verschwinden des Zusammen- 3 "Notably, internet disruptions are geographically dispersed and orthogonal to both AfD likes on Facebook and the total number of refugee attacks in a municipality." 4 Ausnahmsweise sei hier der gesamte Absatz im Original wiedergegeben, um den methodischen Detailreichtum der Arbeit zu illustrieren: "We find that, while the number of hate crimes increases during periods of higher anti-refugee salience, this effect disappears for municipalities experiencing an internet outage.…”
Section: Facebook Und Die Afdunclassified
See 1 more Smart Citation