Adequate Class Assignments on Linked Data

Abstract

In recent years Semantic Web technologies and the Linked Data paradigm have allowed the emergence of large interlinked knowledge bases as Linked datasets. These databases contain information that associates Web entities (called resources) with a well-defined semantics that specifies how these entities should be interpreted. A way to perform this task is through a class assignment process where resources are identified as members of certain classes described in ontologies. In order to improve the quality of the "meaning" of the data contained in Linked datasets a key challenge in the Linked Data community is to detect, assess and eventually fix wrong class assignments. In this sense, this work proposes an interpretation for adequate class assignments considering three quality dimensions from a semantic perspective: redundancy, consistency and accuracy. For each dimension, a formal definition is presented, then applied to class assignments and finally used as guideline to show how quality metrics and data curation strategies can be defined.