Skip to content
BENIBIX Learn

Data Labeling

Also known as Data Labelling, Annotation (Machine Learning), Training Data Annotation

Each example gets a tag with the right answer. The model learns from it.

Think of it like

It is like putting name tags on photos. Then a child can learn who is who.

Example

A team labels 10,000 X-ray images as "broken bone" or "no break". Two doctors check each one, and they settle any disagreement.

Why it matters

A model can only be as good as its labels. Tagging is often the biggest cost in a project.

Definition

Data labeling is the work of adding the right answer to each example. A person looks at a photo, a sentence or a sound clip and tags it. The tag might be "cat", "spam" or "angry". These tags become the labels a model learns from. The work is slow and often split across many people. Wrong or mixed-up tags teach the model wrong lessons.
The label column is added by people. It holds the answer the model must learn.

Common mistake

Not writing clear rules for the people who label. Each one then tags the same thing in a different way.

Key takeaways

  1. 1
    Add the answer. One tag for each example.
  2. 2
    Clear rules. Everyone tags the same way.
  3. 3
    Check twice. Two people catch each other's mistakes.

Learn it properly