Quality assurance in annotation

The practices used to make annotation quality repeatable across a project.

Quality assurance covers instructions, reviewer preparation, evaluation methods and corrective processes. Quality control checks the actual outputs against requirements. A sound process uses both: it prevents avoidable mistakes and inspects the work that will enter a dataset.

In practice

A team clarifies ambiguous action definitions, tests reviewers on reference clips and reviews each delivered batch against agreed criteria.