A little goes a long way: Improving toxic language classification despite data scarcity
This addresses data scarcity in toxic language detection, offering practical solutions for resource-constrained applications, though it is incremental in exploring augmentation methods.
The paper tackled the problem of toxic language classification with scarce labeled data by systematically studying data augmentation techniques, showing that combining three techniques, including GPT-2-generated sentences, enabled shallow classifiers to perform comparably to BERT on very scarce datasets.
Detection of some types of toxic language is hampered by extreme scarcity of labeled training data. Data augmentation - generating new synthetic data from a labeled seed dataset - can help. The efficacy of data augmentation on toxic language classification has not been fully explored. We present the first systematic study on how data augmentation techniques impact performance across toxic language classifiers, ranging from shallow logistic regression architectures to BERT - a state-of-the-art pre-trained Transformer network. We compare the performance of eight techniques on very scarce seed datasets. We show that while BERT performed the best, shallow classifiers performed comparably when trained on data augmented with a combination of three techniques, including GPT-2-generated sentences. We discuss the interplay of performance and computational overhead, which can inform the choice of techniques under different constraints.