@inproceedings{vidgen-etal-2021-learning,
title = "Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection",
author = "Vidgen, Bertie and
Thrush, Tristan and
Waseem, Zeerak and
Kiela, Douwe",
booktitle = "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)",
month = aug,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.acl-long.132",
doi = "10.18653/v1/2021.acl-long.132",
pages = "1667--1682",
abstract = "We present a human-and-model-in-the-loop process for dynamically generating datasets and training better performing and more robust hate detection models. We provide a new dataset of 40,000 entries, generated and labelled by trained annotators over four rounds of dynamic data creation. It includes 15,000 challenging perturbations and each hateful entry has fine-grained labels for the type and target of hate. Hateful entries make up 54{\%} of the dataset, which is substantially higher than comparable datasets. We show that model performance is substantially improved using this approach. Models trained on later rounds of data collection perform better on test sets and are harder for annotators to trick. They also have better performance on HateCheck, a suite of functional tests for online hate detection. We provide the code, dataset and annotation guidelines for other researchers to use.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="vidgen-etal-2021-learning">
<titleInfo>
<title>Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection</title>
</titleInfo>
<name type="personal">
<namePart type="given">Bertie</namePart>
<namePart type="family">Vidgen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tristan</namePart>
<namePart type="family">Thrush</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zeerak</namePart>
<namePart type="family">Waseem</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Douwe</namePart>
<namePart type="family">Kiela</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2021-aug</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)</title>
</titleInfo>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Online</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>We present a human-and-model-in-the-loop process for dynamically generating datasets and training better performing and more robust hate detection models. We provide a new dataset of 40,000 entries, generated and labelled by trained annotators over four rounds of dynamic data creation. It includes 15,000 challenging perturbations and each hateful entry has fine-grained labels for the type and target of hate. Hateful entries make up 54% of the dataset, which is substantially higher than comparable datasets. We show that model performance is substantially improved using this approach. Models trained on later rounds of data collection perform better on test sets and are harder for annotators to trick. They also have better performance on HateCheck, a suite of functional tests for online hate detection. We provide the code, dataset and annotation guidelines for other researchers to use.</abstract>
<identifier type="citekey">vidgen-etal-2021-learning</identifier>
<identifier type="doi">10.18653/v1/2021.acl-long.132</identifier>
<location>
<url>https://aclanthology.org/2021.acl-long.132</url>
</location>
<part>
<date>2021-aug</date>
<extent unit="page">
<start>1667</start>
<end>1682</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection
%A Vidgen, Bertie
%A Thrush, Tristan
%A Waseem, Zeerak
%A Kiela, Douwe
%S Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)
%D 2021
%8 aug
%I Association for Computational Linguistics
%C Online
%F vidgen-etal-2021-learning
%X We present a human-and-model-in-the-loop process for dynamically generating datasets and training better performing and more robust hate detection models. We provide a new dataset of 40,000 entries, generated and labelled by trained annotators over four rounds of dynamic data creation. It includes 15,000 challenging perturbations and each hateful entry has fine-grained labels for the type and target of hate. Hateful entries make up 54% of the dataset, which is substantially higher than comparable datasets. We show that model performance is substantially improved using this approach. Models trained on later rounds of data collection perform better on test sets and are harder for annotators to trick. They also have better performance on HateCheck, a suite of functional tests for online hate detection. We provide the code, dataset and annotation guidelines for other researchers to use.
%R 10.18653/v1/2021.acl-long.132
%U https://aclanthology.org/2021.acl-long.132
%U https://doi.org/10.18653/v1/2021.acl-long.132
%P 1667-1682
Markdown (Informal)
[Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection](https://aclanthology.org/2021.acl-long.132) (Vidgen et al., ACL 2021)
ACL