An Uncertainty Representation for Set-Valued Attributes with Hierarchical Domains

In collected data information about a single property may be presented with variable resolution and focus. The present paper describes how hierarchically structured attribute domains support the transfer of knowledge between alternative frames of discernment allowing to flexibly serve information needs and facilitate the processing of inhomogeneous data. The approach is later extended to accommodate setvalued attributes, which have previously been employed to represent imprecision and have recently gained attention in text processing, hierarchy learning or multi-label classification.