labelable

Labelable is an adjective meaning capable of being labeled or categorized. In information management, data science, and knowledge representation, an item is labelable if a defined taxonomy or label space allows assigning one or more labels that describe its category or class. The label space may be single-label (one label per item) or multi-label (several labels per item). Labelability relies on a clear schema, explicit criteria for each label, and, for human tasks, acceptable inter-annotator agreement.

Labelable data are central to supervised learning, data annotation, and crowdsourcing. Datasets used to train models must provide labels; items lacking a definable label or with highly ambiguous content may be deemed non-labelable or require recoding to fit the taxonomy. In practice, labelability can be influenced by the granularity of the label set, cultural or domain biases, and the subjective nature of some categories.

Examples include image datasets labeled with object classes (dog, car), text corpora labeled by topics or sentiment, and medical records labeled by diagnosis codes. Spelling variants include labelable (American English) and labellable (British English). Related terms include labeling, annotation, classification, and taxonomy.