identsets

Identsets are a concept used in data management and information systems to designate sets of attributes that can be used to uniquely identify a record in a dataset. An identset is a subset P of attributes such that the projection of the dataset onto P is injective; in other words, no two distinct records share the same values on P. This formalizes the idea of identifying a row by a combination of fields.

Minimal identsets, where no proper subset is itself an identset, are often called identset keys. These correspond closely to the conventional notion of candidate keys in relational databases, though some authors prefer identset terminology to emphasize practical identity resolution across heterogeneous data sources.

Examples illustrate the idea. Consider a table with attributes (ssn, email, name, dob). The identset {ssn} typically uniquely identifies each row, as does {email} in many systems. The set {name, dob} may fail to be an identset if different people share the same name and birth date. The set {ssn, email} is also an identset, but it is not minimal if {ssn} already suffices.

Applications and properties. Identsets are used in identity resolution, deduplication, and schema design to determine which attribute combinations guarantee unique identification. They help in designing unique constraints and efficient lookup structures. The collection of all identsets on a dataset has structural parallels to the lattice of keys under inclusion, with minimal identsets serving as the most compact identity anchors.

Relation to related concepts. Identsets are closely related to primary keys, superkeys, and functional dependencies in database theory, and they intersect with practical identity resolution and data integration tasks. See also candidate keys, primary key, superkey, and functional dependency.