Loading…
Glossary term
Adding correct answers or tags to data so it can be used for supervised learning.
Data labeling means attaching correct answers to raw data: drawing boxes around cars, marking sentiment, or ranking model replies. People do it with guidelines and tools, and sometimes models pre-label for people to correct.
Quality control, such as having several people label the same item, is essential.
Supervised learning is only as good as its labels, and labeling is often the largest cost in an AI project.