Data Labeling
Pelabelan Data
Attaching the right answer to each example so a model can learn from it.
Data labeling is the human work of attaching the correct answer to each example: which of these emails is spam, which sentence in this contract is the termination clause, what is in this image.
It is the hidden majority of most AI projects. The model is a solved problem for many tasks; producing enough correctly labelled examples is a staffing, tooling, and quality problem that usually takes longer.
Label quality has a ceiling. If two careful annotators would disagree on a case, the task definition is not clear enough yet, and no amount of extra annotators fixes it. Measure the agreement rate before scaling, because it predicts the model quality more reliably than the label count does.
Related terms
Dataset
A collection of examples used to train, evaluate, or ground a system.
Supervised Learning
Learning from examples that come with the right answer attached.
Evaluation
Measuring whether the system does the job, on a fixed set of cases, repeatedly.
Golden Dataset
A fixed, reviewed set of real cases the system is measured against.
Data Governance
The rules and habits that decide what data exists, who may use it, and how long it lives.
Want this working inside your team?
It starts with one conversation about your workflow.