AustroTox: A Dataset for Target-Based Austrian German Offensive Language Detection

TL;DR

AustroTox is a new dataset for offensive language detection in Austrian German, including token-level annotations, enabling improved interpretability and evaluation of language models in this dialect.

Contribution

The paper introduces AustroTox, the first dataset with token-level annotations for Austrian German offensive language detection, and evaluates model performance on this resource.

Findings

Fine-tuned models excel at detecting dialect-specific vulgar language.

Large language models outperform in overall offensiveness detection.

The dataset enables better interpretability of toxicity detection models.

Abstract

Model interpretability in toxicity detection greatly profits from token-level annotations. However, currently such annotations are only available in English. We introduce a dataset annotated for offensive language detection sourced from a news forum, notable for its incorporation of the Austrian German dialect, comprising 4,562 user comments. In addition to binary offensiveness classification, we identify spans within each comment constituting vulgar language or representing targets of offensive statements. We evaluate fine-tuned language models as well as large language models in a zero- and few-shot fashion. The results indicate that while fine-tuned models excel in detecting linguistic peculiarities such as vulgar dialect, large language models demonstrate superior performance in detecting offensiveness in AustroTox. We publish the data and code.