Section: New Results

Quotient summarization of RDF graphs

We have continued our work on efficiently computing informative summaries of large, heterogeneous RDF graphs.

First, we have noticed that type information, when available, can be used to group RDF nodes in interesting, pertinent equivalence classes. However, the integration of type in our quotient summarization framework (presented in ISWC 2017) is not straightforward, since an RDF node may have zero, one, or more than one types. In [15], we have identified a sufficient, flexible condition under which we are able to propose a form of quotient summarization based on types, even if a node has multiple types, and even if they are not organized in a tree-shape classification, but instead in a directed acyclic graph (DAG).

In parallel, we have finalized a comprehensive survey of RDF graph summarization techniques which appeared in the VLDB Journal [8]. We have also completely re-developed our RDF graph summarization platform, in order to ensure correctness, to factorize common elements across all the summarization methods, and to implement new, incremental summarization algorithms [21]. This work has attracted significant visibility through an invited keynote at the ESWC conference [25], and through an ISWC “Resource” publication where our summaries are integrated in a LOD visual exploration portal developed by the Ilda team of Inria [17].