A classification system maps an input to a category from a specified set. It can use learned patterns to distinguish categories in text, images or other data. Some systems output a category directly, while others output scores that are interpreted using a decision rule.
Useful evaluation checks the types of errors as well as total accuracy. A system that mostly predicts the common category may look accurate while performing poorly on rare but important cases. The categories and scoring method should match the intended application.
ELI5
Classification is sorting information into named groups. An AI system learns clues that help it decide which group an item belongs in.
For example, a support team could sort incoming messages into billing, delivery and technical questions. If a customer asks where a parcel is, the system should choose delivery, so the request reaches the appropriate team.
Can an input belong to more than one category?
Yes. Some classification tasks allow multiple labels, while others require exactly one category.
Is classification the same as generating a chat answer?
No. Classification chooses categories; free-form generation produces new text rather than only selecting a label.
