Decoder. plain-English AI glossary

Annotation

● Core

Marking up data with the extra information a model needs to learn — labels, spans, boxes, rankings.

Think of it like

Like a copy editor marking a manuscript: highlighting names, flagging sections, noting what’s what for whoever reads it next.

Example

Annotators draw boxes around every pedestrian in dashcam frames so a detection model knows what to find.

How it actually works

Annotation is the broader craft around labeling: not just a class, but bounding boxes, entity spans, relationships, or preference rankings, guided by detailed instructions. Clear guidelines and inter-annotator agreement matter because ambiguous tasks yield inconsistent data. It’s where much of a dataset’s cost and quality is decided, well before any model runs.

For product teams

The unglamorous work that determines whether your data is worth training on at all.

For engineers

Enriching raw data with structured labels/markup per guidelines; consistency measured via inter-annotator agreement.

Labeling — A superset of labeling.
HITL — Often done with HITL.

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open Decoder Add to Chrome