Labeling
Attaching the correct answer to each training example so a model can learn from it.
Think of it like
Like a teacher writing the right answer beside each practice problem before handing the worksheet to a student.
Example
Reviewers tag ten thousand emails as “spam” or “not spam” so a classifier has ground truth to learn from.
How it actually works
Supervised learning needs examples paired with target outputs, and labeling is the human-heavy process of producing them. It’s often the costliest, slowest part of a project, and label quality caps model quality — noisy or inconsistent labels teach noise. Guidelines, multiple raters, and agreement checks are how teams keep labels trustworthy.
For product teams
Usually the real bottleneck in an ML project — slow, expensive, and quality-limiting.
For engineers
Assigning target outputs to examples for supervised training; label noise directly bounds achievable accuracy.
Related
- Annotation — Closely tied to annotation.
- Supervised learning — The fuel for supervised learning.
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 DecoderAdd to Chrome