TOKEN: Task Decomposition and Knowledge Infusion for Few-Shot Hate Speech Detection

Auteur-es

Badr AlKhamissi, Faisal Ladhak, Srini Iyer, Zornitsa Kozareva, Xian Li, Pascale Fung, Lambert Mathias, Asli Celikyilmaz, Mona Diab

Nombre Auteurs

9

Titre

TOKEN: Task Decomposition and Knowledge Infusion for Few-Shot Hate Speech Detection

Année de publication

2022

Référence (APA)

AlKhamissi, B., Ladhak, F., Iyer, S., Stoyanov, V., Kozareva, Z., Li, X., Fung, P., Mathias, L., Celikyilmaz, A., & Diab, M. (2022). TOKEN: Task Decomposition and Knowledge Infusion for Few-Shot Hate Speech Detection.

résumé

Hate speech detection is complex; it relies on commonsense reasoning, knowledge of stereotypes, and an understanding of social nuance that differs from one culture to the next. It is also difficult to collect a large-scale hate speech annotated dataset. In this work, we frame this problem as a few-shot learning task, and show significant gains with decomposing the task into its "constituent" parts. In addition, we see that infusing knowledge from reasoning datasets (e.g. ATOMIC 20/20 ) improves the performance even further. Moreover, we observe that the trained models generalize to out-of-distribution datasets, showing the superiority of task decomposition and knowledge infusion compared to previously used methods. Concretely, our method outperforms the baseline by 17.83% absolute gain in the 16-shot case.

URL

https://research.facebook.com/file/8394145417323280/TOKEN--Task-Decomposition-and-Knowledge-Infusion-for-Few-Shot-Hate-Speech-Detection.pdf

doi

https://doi.org/10.48550/arXiv.2205.12495

Accessibilité de l'article

Libre

Champ

Artificial Intelligence, Natural Language Processing & Speech

Type contenu (théorique Applicative méthodologique)

Méthodologique

Méthode

The method involves breaking down the task of hate speech detection into smaller subtasks, using a few-shot learning approach to train the model on a small amount of labeled data, and infusing knowledge from reasoning datasets to improve performance.

Cas d'usage

N/A

Objectifs de l'article

The objectives of the article are to improve the performance of hate speech detection models, particularly in low-resource settings, and to explore the use of task decomposition and knowledge infusion in natural language processing tasks.

Question(s) de recherche/Hypothèses/conclusion

The research question is how can task decomposition and knowledge infusion be used to improve the performance of hate speech detection models?

The hypothesis is that breaking down the task of hate speech detection into smaller subtasks and infusing knowledge from reasoning datasets will improve the performance of the model.

The conclusions are that the proposed method outperforms previous approaches to hate speech detection and shows promising results in detecting harmful content on social media platforms.

Cadre théorique/Auteur.es

The theoretical framework of the article includes natural language processing and machine learning, with references to authors such as Sanguinetti, Assimakopoulos, and de Gibert.

Concepts clés

Few-shot learning, Task decomposition, Hate speech detection, Social media

Données collectées (type source)

datasets for evaluation, including HateXplain, HS18, and Ethos

Définition des émotions

Non

Ampleur expérimentation (volume de comptes)

12,838 examples

Technologies associées

Natural language processing, Machine learning

Mention de l'éthique

The article does not mention ethics explicitly, but it does discuss the potential impact of hate speech on individuals and society.