Technical
Data Labeling
The process of annotating data with ground-truth labels used for supervised learning, often requiring human annotators to classify, tag, or rate examples.
Explained at five levels
Level 1
When people look at pictures or text and add notes to teach the AI what things are — like putting name tags on everything.
Level 2
The process of humans tagging data with correct answers so AI can learn from it — like marking photos as "cat" or "dog" so the AI learns the difference.
Level 3
The process of annotating data with ground-truth labels used for supervised learning, often requiring human annotators to classify, tag, or rate examples.
Level 4
The creation of annotated datasets through human or semi-automated annotation, establishing the supervision signal for model training — subject to inter-annotator agreement, label noise, and annotation guidelines quality.
Level 5
The production of labeled exemplars from an annotation protocol — a critical bottleneck in supervised learning, addressable via active learning, weak supervision, programmatic labeling functions, and human-in-the-loop workflows with calibrated annotator agreement metrics.
Definitions are educational summaries. Terminology can vary by source and context.
Sources
- NIST Trustworthy and Responsible AI Resource Center: Glossary — National Institute of Standards and Technology. Accessed 2026-07-20.