This document presents an approach for exploratory relationship search through hierarchical clustering. It aims to address the challenge of too many relationship search results by organizing them into a cluster hierarchy based on common relationship patterns. An evaluation with participants performing lookup and exploratory search tasks on DBpedia data found that the clustering approach outperformed simple listing and faceted categorization alternatives. User feedback suggested areas for improvement like more concise visualizations and cognitive support. The authors conclude it is a promising approach and future work could combine facets and clustering or explore alternatives.
7. Exploratory relationship search
• Exploring a set of relationships interactively and continuously
faceted categories
(RelFinder)
clustering
(our solution: RelClus)
9. Challenges
• How to meaningfully label a cluster?
• How to make sense of a cluster hierarchy?
• How to measure similarity between clusters?
Agglomerative hierarchical clustering
• Initially: relationships singleton clusters
• Then: progressively merge the most similar pair
12. How to meaningfully label a cluster?
• Using a leastest common relationship pattern
– Vertices: leastest common classes (or entities)
– Edges: leastest common properties
Person
P1
R4
R5
label({R4, R5}) = P1
13. How to make sense of a cluster hierarchy?
• subPatternOf (⊑)
– Vertices: s.t. subClassOf (or instance-type)
– Edges: s.t. subPropertyOf
P3
P2
P1
P2 ⊑ P3, P1 ⊑ P3
14. How to measure similarity between clusters?
• sim(Ci,Cj) = how many commonalities they share
which are exactly captured by label(Ci∪Cj)
– Measure: -log (probability of seeing label(Ci∪Cj))
i.e. the information content associated with label(Ci∪Cj)
– Probability estimation: based on the data set
P3
P2
P1