UrduSpeech: A 156-Hour Urdu Speech Corpus with 12-Dimension Paralinguistic Annotations

TL;DR

UrduSpeech is a comprehensive 156-hour Urdu speech corpus with detailed paralinguistic annotations, addressing resource scarcity and enabling advanced speech technology development for Urdu.

Contribution

The paper introduces a large, high-quality Urdu speech corpus with paralinguistic metadata, curated using an LLM-driven pipeline, and provides a benchmark set for research.

Findings

Mean Opinion Score of 4.6 confirms high quality

97.6% confidence in data curation pipeline

Balanced gender representation across 71,792 utterances

Abstract

Despite 230 million speakers, Urdu remains critically under-resourced in speech technology. We introduce UrduSpeech: a large high-fidelity Urdu corpus comprising 156 hours of audio with 12-dimension paralinguistic metadata, encompassing US-Std, US-CS, US-EngPk. To address Right-to-Left script constraints and frequent code-switching, we developed UrduSpeech, a LLM-driven pipeline to curate data across 12 diverse categories, including news, drama, and rare literary forms like Bait-Bazi. We also release a 9-hour US-Benchmark set, manually corrected by native annotators to serve as a standard. Human quality assessment of the primary 156-hour corpus yielded a Mean Opinion Score (MOS) of 4.6 (std = 0.7) with inter-rater reliability confirmed by a 0.68 Cohen's Kappa, validating our curation pipeline's 97.6% confidence score. The corpus maintains a 60-40 gender balance across 71,792 utterances. Our work represents a significant leap toward linguistic inclusivity in global AI. The corpus and code are open-sourced, and a demo page is available.