Relation-valued attributes in rank-aware databases and related concepts

We present an initial study of the relation-valued attributes in rank-aware databases. We introduce the grouping and ungrouping operations and show their basic properties. We use a model that generalizes the Codd model of data by considering tuples in relations annotated by scores indicating a degree to which tuples match queries. We utilize the complete residuated lattices as the structures of degrees. We argue that relation-valued attributes are a useful concept and play an irreplaceable role in rank-aware databases as they form a bridge between the rank-aware and classical model.