What is data labeling?

Data labeling means attaching a label from a fixed list to each item in a dataset: this email is spam or not, this image contains a cat or a dog, this sentence is positive or negative. It is the simplest and highest-volume form of data annotation.

Labeling versus annotation

The two words are often used interchangeably. When they are distinguished, labeling means choosing from a set list, while annotation also covers richer marks such as boxes, spans and free-text notes.

Who does it

Labeling is the usual entry point into AI training work. It needs care and consistency more than credentials, is often paid per task, and is offered by large crowd platforms as well as specialist vendors. Most listings on our jobs board that need no degree are labeling or transcription roles.

Example

A support-ticket project shows one customer message at a time and asks you to pick its intent from twelve options, such as refund request or password reset. Ambiguous tickets go to a second labeler, and disagreements to a reviewer.