A new 53,274-example synthetic Portuguese dataset with group and discourse labels exposes a mutual generalization failure between social-media and minority-focused hate speech classifiers.
Our approach emphasizes con- trolled representation of protected groups and lin- guistic diversity across both toxic and non-toxic expressions
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
ToxSyn-PT: A Synthetic Fine-Grained Dataset of Minority-Targeted Toxic Language in Portuguese
A new 53,274-example synthetic Portuguese dataset with group and discourse labels exposes a mutual generalization failure between social-media and minority-focused hate speech classifiers.