A meta-learning framework with cross-domain episode sampling achieves higher anomaly detection AUC and F1 than fine-tuned BERT and one-class SVM on SMS spam, COVID-19 fake news, and hate speech tasks.
For each, shuffle the normal and anomaly examples (the model doesn’t know which is which a priori)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Anomaly Detection in Human Language via Meta-Learning: A Few-Shot Approach
A meta-learning framework with cross-domain episode sampling achieves higher anomaly detection AUC and F1 than fine-tuned BERT and one-class SVM on SMS spam, COVID-19 fake news, and hate speech tasks.