NLPData Engineering
What is actually inside an Urdu speech dataset.
Exploratory notebook with recorded statistics.
pandaslibrosapydubseabornPCA
Overview
Metadata cleaning, demographic coverage, clip-duration statistics, and audio features (MFCC, chroma, spectral contrast, tonnetz) for Mozilla Common Voice Urdu.
Recorded results
- Validated clips
- 52,144
- Plus 6,839 unvalidated
- Source: i211697_laibakhwar_A2.ipynb
- Mean clip length
- 4,246 ms
- Across 196,558 clip durations
- Source: i211697_laibakhwar_A2.ipynb