Annotation
Domain: Data & Datasets
🎯 Motivation
The process of labeling data, often by human experts. Quality annotation is expensive but crucial.
📋 Concrete Example
Annotation: 5 experts read 10,000 texts and label each as ‘positive,’ ‘neutral,’ or negative. Inter-annotator agreement: 92%. High-quality labels.
🔗 Analogy
A museum curator labeling artifacts - each artifact gets a card with its name, era, and description. Annotation labels data for the model.