PGT-FG labels toxicity in 2,396 Punjabi song segments
Original title: The Punjabi Geet Toxicity Dataset (PGT-FG): Fine-Grained Multimodal Annotations for Toxic Content in Punjabi Songs
About 29% of segments are labelled toxic, violence is the most common category, and the resolved set is on Zenodo.
Researchers release the Punjabi Geet Toxicity Dataset (PGT-FG), a segment-level resource for toxicity in Punjabi songs drawn from public YouTube material. After filtering, the set has 2,396 unique segments from 392 songs, with 2,870 annotation records from four native annotators. Labels cover binary toxicity, eight fine-grained categories, severity, sentiment, mode, and uncertainty. Of the segments, 705 (29.42%) are toxic; violence appears in 539 of those. The resolved labels are on Zenodo; lyrics and copyrighted audio or video are not redistributed.
- An initial 620 songs yielded 465 downloads and 3,402 preliminary segments.
- Four annotators labelled a shared subset of 158 segments.
- Overall toxicity agreement is Fleiss' kappa 0.6521.
- Violence appears in 76.45% of toxic segments, ahead of gang and drug references.
- Copyrighted lyrics, audio, and video are not included in the release.
- Organizations
- Zenodo
From the original article
1 min read at arxiv.orgResources for toxicity detection in Punjabi remain limited, particularly for musical content that combines linguistic and multimodal information. We present the Punjabi Geet Toxicity Dataset (PGT-FG), a segment-level resource developed to support research on toxicity detection in Punjabi musical content. The dataset was constructed from publicly available Punjabi songs collected from YouTube.
By Lakhjeet Kaur · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.