Two Contrasting Data Annotation Paradigms for Subjective NLP Tasks

Röttger, Paul; Vidgen, Bertie; Hovy, Dirk; Pierrehumbert, Janet B.

doi:10.48550/arxiv.2112.07475

Cited by 9 publications

(10 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For each language we recruited at least 12 annotators, so that in total more than 120 annotators worked on MHC. 5 Annotation was prescriptive (Röttger et al, 2021a), meaning that annotators were tasked with adhering to clear annotation guidelines reflecting our definition of hate speech, rather than sharing their personal view on what is and is not hateful. 6 Compared to the original HATECHECK, where four out of five annotators confirmed the gold label for 99.4% of all test cases, there was more disagreement on MHC (see Appendix D).…”

Section: Generating Test Casesmentioning

confidence: 99%

HateCheck: Functional Tests for Hate Speech Detection Models

Röttger¹,

Vidgen²,

Nguyen³

et al. 2021

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Confer

Self Cite

102

View full text Add to dashboard Cite

Detecting online hate is a difficult task that even state-of-the-art models struggle with. Typically, hate speech detection models are evaluated by measuring their performance on held-out test data using metrics such as accuracy and F1 score. However, this approach makes it difficult to identify specific model weak points. It also risks overestimating generalisable model performance due to increasingly well-evidenced systematic gaps and biases in hate speech datasets. To enable more targeted diagnostic insights, we introduce HATECHECK, a suite of functional tests for hate speech detection models. We specify 29 model functionalities motivated by a review of previous research and a series of interviews with civil society stakeholders. We craft test cases for each functionality and validate their quality through a structured annotation process. To illustrate HATECHECK's utility, we test near-state-of-the-art transformer models as well as two popular commercial models, revealing critical model weaknesses.

show abstract

Section: Generating Test Casesmentioning

confidence: 99%

HateCheck: Functional Tests for Hate Speech Detection Models

Röttger¹,

Vidgen²,

Nguyen³

et al. 2021

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Confer

Self Cite

102

View full text Add to dashboard Cite

show abstract

“…To validate the gold-standard labels assigned to each test case, we recruited three annotators with prior experience on hate speech projects. 5 Annotators were given extensive and prescriptive guidelines (Röttger et al, 2022), as well as test tasks and training sessions, which included examining real-world examples of emoji-based hate from Twitter. We followed guidance for protecting annotator well-being (Vidgen et al, 2019).…”

Section: Validating Test Casesmentioning

confidence: 99%

Hatemoji: A Test Suite and Adversarially-Generated Dataset for Benchmarking and Detecting Emoji-Based Hate

Kirk¹,

Vidgen²,

Röttger³

et al. 2022

Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Langua

Self Cite

View full text Add to dashboard Cite

Detecting online hate is a complex task, and low-performing models have harmful consequences when used for sensitive applications such as content moderation. Emoji-based hate is an emerging challenge for automated detection. We present HATEMOJICHECK, a test suite of 3,930 short-form statements that allows us to evaluate performance on hateful language expressed with emoji. Using the test suite, we expose weaknesses in existing hate detection models. To address these weaknesses, we create the HATEMOJIBUILD dataset using a human-andmodel-in-the-loop approach. Models built with these 5,912 adversarial examples perform substantially better at detecting emoji-based hate, while retaining strong performance on text-only hate. Both HATEMOJICHECK and HATEMOJI-BUILD are made publicly available. 1 Content Warning This article contains examples of hateful language from HATEMOJICHECK to illustrate its composition. Examples are quoted verbatim, except for slurs and profanity in text, for which the first vowel is replaced with an asterisk. The authors oppose the use of hateful language.

show abstract

“…One factor in settling how annotations can support the purpose of the task is how much subjectivity is desired. Röttger et al (2021a) distinguish two types of approaches to annotation: descriptive and prescriptive. Descriptive annotations encourage subjectivity of annotators (where inconsistency is not an issue), while prescription instructs annotators to strictly follow carefully defined criteria (the less subjectivity, the better).…”

Section: Task Specific Annotationsmentioning

confidence: 99%

Hate Speech Criteria: A Modular Approach to Task-Specific Hate Speech Definitions

Urja¹,

Vermeulen²,

Nalisnick³

et al. 2022

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

View full text Add to dashboard Cite

Offensive Content Warning: This paper contains offensive language only for providing examples that clarify this research and do not reflect the authors' opinions. Please be aware that these examples are offensive and may cause you distress.The subjectivity of recognizing hate speech makes it a complex task. This is also reflected by different and incomplete definitions in NLP. We present hate speech criteria, developed with perspectives from law and social science, with the aim of helping researchers create more precise definitions and annotation guidelines on five aspects: (1) target groups, (2) dominance, (3) perpetrator characteristics, (4) type of negative group reference, and the (5) type of potential consequences/effects. Definitions can be structured so that they cover a more broad or more narrow phenomenon. As such, conscious choices can be made on specifying criteria or leaving them open. We argue that the goal and exact task developers have in mind should determine how the scope of hate speech is defined. We provide an overview of the properties of English datasets from hatespeechdata.com that may help select the most suitable dataset for a specific scenario.

show abstract

Two Contrasting Data Annotation Paradigms for Subjective NLP Tasks

Cited by 9 publications

References 24 publications

HateCheck: Functional Tests for Hate Speech Detection Models

HateCheck: Functional Tests for Hate Speech Detection Models

Hatemoji: A Test Suite and Adversarially-Generated Dataset for Benchmarking and Detecting Emoji-Based Hate

Hate Speech Criteria: A Modular Approach to Task-Specific Hate Speech Definitions

Contact Info

Product

Resources

About