Data Provenance
Domain: Safety & Ethics
π― Motivation
Knowing where data came from, who created it, and how it has been modified. Essential for trust and compliance.
π Concrete Example
Provenance: βThis dataset: collected from Wikipedia (Jan 2025), filtered for English, deduplicated, labeled by 5 annotators, reviewed by 2 editors.β Full history.
π Analogy
A certificate of origin for a diamond - it proves where the diamond was mined, cut, and sold. Data provenance is the certificate of origin for training data.