Towards Exploratory Relationship Search: A Clustering-Based Approach

摘要

Searching and browsing relationships between entities is an important task in many domains. RDF facilitates searching by explicitly representing a relationship as a path in a graph with meaningful labels. As the Web of RDF data grows, hundreds of relationships can be found between a pair of entities, even under a small length constraint and within a single data source. To support users with various information needs in interactively exploring a large set of relationships, existing efforts mainly group the results into faceted categories. In this paper, we practice another direction of exploratory search, namely clustering. Our approach automatically groups relationships into a dynamically generated hierarchical clustering according to their schematic patterns, which also meaningfully label these clusters to effectively guide exploration and discovery. To demonstrate it, we implement our approach in the RelClus system based on DBpedia, and conduct a preliminary user study as well as a performance testing.