Large-scale audio corpus containing 1,000 hours of Kurdish speech from 500+ speakers across different dialects, ages, and regions. Includes high-quality transcriptions and speaker metadata.
WAV audio files (16kHz, mono), Transcript files with timestamps, Speaker metadata (demographics, dialect), Phonetic alignments
Dr. Mohammad Ali , Dr. John Doe , Dr. Sara Ahmed (2023)
This research develops the first comprehensive ASR system for Kurdish, supporting both Sorani and Kurmanji dialects with domain adaptation techniques. Achieved 89.3% word accuracy on conversational …
Kareem, H., Ahmed, R., & Jamal, S. (2023). Kurdish Speech Recognition Audio Corpus. KaiLab Research Data Repository. https://doi.org/10.5281/kurd-speech-corpus.v1