The Semantic Web is enabling application architects to use Linked Data in exciting new use cases. These applications will also enable humans to interact with this data. Common to these tools is the need to present data in useful views that are obvious and natural to humans. This typically entails the summarization and ranking of properties in a sensible order. Unfortunately, the Web Ontology Language (OWL) family does not cater for specification of property ranking information. Tool developers have no way to determine an appropriate (topic attuned) property ranking from a given OWL based ontology. We have researched this problem as to understand to what extent an ontology property ranking algorithm can approximate a human designed ranking to support automated view generation in Semantic Web browsers. We present three ontology property ranking algorithms that significantly improve over alphabetic ranking. These algorithms have been assessed using 19 key performance indicators to provide a holistic performance review by comparing the computed ranking against a ranking baseline that we extracted from Wikipedia Infobox templates. We conclude that our algorithms overall provide a moderate approximation of our ground truth set ranking, a significant result given the vast amount of possible ranking permutations.
展开▼
