Researchers have introduced MISO-BR, a new open corpus designed for detecting misogyny in Brazilian Portuguese social media text. The corpus contains 896 annotated social media posts, developed with a detailed construction protocol and reproducible baselines. Experiments show that transformer models like BERTimbau achieve significantly higher performance than classical classifiers, reaching a macro-F1 score of 0.833 compared to the classical models' range of 0.651-0.676. AI
IMPACT Provides a new resource for developing and evaluating hate speech detection models in a less-resourced language.
RANK_REASON The item is a research paper introducing a new dataset and evaluation of models for a specific NLP task. [lever_c_demoted from research: ic=1 ai=1.0]
- Bernoulli Naïve Bayes
- BERTimbau Base
- BERTimbau Large
- Brazilian Portuguese
- Hugging Face
- Multinomial Naive Bayes
- OLID-BR
- Pericles Miranda Dr
- random forest
- support vector machine
- X
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
