🏷️ Data Labeling
5th Grade 🤖 Artificial Intelligence
Data labeling means adding a “label” — the correct answer — to each example in a dataset: writing “cat” next to a picture, marking a message as “spam” or writing down what was said in a recording. The model learns from these labels.
When you teach a little kid to recognize animals, you point and say, “That’s an elephant.” Data labeling does something similar for a computer, but thousands and millions of times. Lots of labeling is done by people, sometimes with tools that suggest a label for a person to approve or fix. Sometimes people also mark exactly where something is in a picture — for example, by drawing a box around every person walking.
When you mark a message as spam, or a website asks you to “select all the pictures with traffic lights,” you’re sometimes helping to label data. Good labeling has to be accurate and consistent — if different people label the same thing in different ways, the model gets confused.
A common mistake is thinking AI learns everything on its own, without any human work. Behind many systems is a huge amount of human work collecting, labeling and checking data. And if the labels are wrong or biased, the model will learn exactly those mistakes.
For example
- Writing “dog” or “cat” next to each picture in a dataset.
- Drawing a box around every car in pictures of a road.
- Typing out what was said in each short recording.
- Marking product reviews as “positive” or “negative.”
📌 Remember: Labeling = attaching the right answer to each example. Wrong labels → wrong model.
Quick quiz
What is data labeling?
Who labels a lot of the data?
What happens if the labels are wrong?