TamHateX: A Multi-Level Annotated Tamil YouTube Comment Dataset for Hierarchical Hate-Expression Analysis
Data in Brief · 2026-11 · 113405
DOI: 10.1016/j.dib.2026.113405 ↗Kurz gefasst
Fachartikel in der Zeitschrift Data in Brief, 2026-11. Verfasst von Kathiravan Pannerselvam und 1 weiteren Autor:innen. Die Arbeit nennt 53 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
TamHateX is a five-level hierarchically annotated corpus of 5313 public Tamil YouTube comments collected between September 2025 and March 2026. It preserves Tamil script, Romanized Tamil, and Tamil-English code-mixed text, slang, spelling variation, punctuation, and emojis while masking identified personal identifiers. Experts from Computer Science, Psychology, and Social Work developed the annotation guidelines and selected one native Tamil annotator from each discipline following a common 100-comment qualification pilot. The three-annotator matrix identifies 951 comments with disagreement at one or more levels, and the adjudication log records their individual annotations and final labels. Each record contains a unique identifier, privacy-masked text, and labels for hate status, target, expression type, fine-grained abuse, and social bias. The official partitions contain 4251 training, 531 validation, and 531 test comments. The corpus supports hierarchical hate-speech classification, low-resource and code-mixed NLP, imbalance-aware learning, and content-moderation research.
Zitieren
APA
Pannerselvam, K., & Rajiakodi, S. (2026). TamHateX: A Multi-Level Annotated Tamil YouTube Comment Dataset for Hierarchical Hate-Expression Analysis. Data in Brief, 113405. https://doi.org/10.1016/j.dib.2026.113405
Vancouver
Pannerselvam K, Rajiakodi S. TamHateX: A Multi-Level Annotated Tamil YouTube Comment Dataset for Hierarchical Hate-Expression Analysis. Data in Brief. 2026:113405. doi:10.1016/j.dib.2026.113405
BibTeX
@article{Pannerselvam2026,
author = {Pannerselvam, Kathiravan and Rajiakodi, Saranya},
title = {TamHateX: A Multi-Level Annotated Tamil YouTube Comment Dataset for Hierarchical Hate-Expression Analysis},
journal = {Data in Brief},
year = {2026},
pages = {113405},
doi = {10.1016/j.dib.2026.113405}
}
EUVIMED