We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CL

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: Analyzing Hate Speech Data along Racial, Gender and Intersectional Axes

Abstract: To tackle the rising phenomenon of hate speech, efforts have been made towards data curation and analysis. When it comes to analysis of bias, previous work has focused predominantly on race. In our work, we further investigate bias in hate speech datasets along racial, gender and intersectional axes. We identify strong bias against African American English (AAE), masculine and AAE+Masculine tweets, which are annotated as disproportionately more hateful and offensive than from other demographics. We provide evidence that BERT-based models propagate this bias and show that balancing the training data for these protected attributes can lead to fairer models with regards to gender, but not race.
Comments: Accepted at "4th Workshop on Gender Bias in Natural Language Processing", NAACL 2022
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Cite as: arXiv:2205.06621 [cs.CL]
  (or arXiv:2205.06621v2 [cs.CL] for this version)

Submission history

From: Antonios Maronikolakis [view email]
[v1] Fri, 13 May 2022 13:13:46 GMT (448kb,D)
[v2] Wed, 18 May 2022 13:16:54 GMT (373kb,D)

Link back to: arXiv, form interface, contact.