Behind Closed Words: Creating and Investigating the forePLay Annotated Dataset for Polish Erotic Discourse

TL;DR

This paper introduces forePLay, a new Polish dataset for erotic content detection, demonstrating that specialized models outperform multilingual ones, especially in handling complex linguistic features and imbalanced data.

Contribution

The creation of the first large-scale annotated Polish erotic discourse dataset with a multidimensional taxonomy and evaluation of language models for content moderation.

Findings

Specialized Polish models outperform multilingual models.

Transformer architectures excel with imbalanced categories.

The dataset supports development of linguistically-aware moderation tools.

Abstract

The surge in online content has created an urgent demand for robust detection systems, especially in non-English contexts where current tools demonstrate significant limitations. We present forePLay, a novel Polish language dataset for erotic content detection, featuring over 24k annotated sentences with a multidimensional taxonomy encompassing ambiguity, violence, and social unacceptability dimensions. Our comprehensive evaluation demonstrates that specialized Polish language models achieve superior performance compared to multilingual alternatives, with transformer-based architectures showing particular strength in handling imbalanced categories. The dataset and accompanying analysis establish essential frameworks for developing linguistically-aware content moderation systems, while highlighting critical considerations for extending such capabilities to morphologically complex languages.