Data Annotation Guideline
Domain: Data & Datasets
🎯 Motivation
Rules for how to label data. Consistent guidelines produce consistent, high-quality labels.
📋 Concrete Example
Guideline: ‘Label as positive only if the text expresses clear happiness. Uncertain = neutral. Sarcasm = negative.’ All annotators follow the same rules.
🔗 Analogy
A grading rubric for teachers - ‘5 = excellent, 3 = average, 1 = poor. Specific criteria for each score.’ Annotation guidelines are rubrics for data labeling.