Semantic Relatedness as an Inter-Facet Metric for Facet Selection over Knowledge Graphs

Abstract

Faceted Browsing is a wide-spread approach for exploratory search. Without requiring an in-depth knowledge of the domain, users can narrow down a resource set until it fits their need. An increasing amount of data is published either directly as Linked Data or is at least annotated using concepts from the Linked Data Cloud. This allows identifying commonalities and differences among resources beyond the comparison of mere string representations of metadata. As the size of data repositories increases, so does the range of covered domains and the number of properties that can provide the basis for a new facet. Manually predefining suitable facet collections becomes impractical. We present our initial work on automatically creating suitable facets for a semantically annotated set of resources. In particular, we address two problems arising with automatic facet generation: (1) Which facets are applicable to the current set of resources and (2) which reasonably sized subset provides the best support to users?