“Knowledge graphs are the next generation tool for helping businesses make critical decisions, based on harmonized knowledge models and data derived from siloed source systems. Due to the huge value generated by their data standardization and semantic modelling capabilities, knowledge graphs are most often associated with data integration, linking, unification and information reuse. As more and more organizations are turning to knowledge graphs for better data and content analytics, search and graph exploration become key requirements also.”

“For many years, two main advantages of labelled property graphs (LPG) have been pointed out:

1. They can deal with properties on edges in the graph, and
2. They are good for graph traversal.”

“These advantages seem less powerful now with leading triplestores supporting RDF-star, which offers a simple and efficient mechanism to attach metadata to the edges of a graph (e.g. weights, access restrictions and provenance information), and SPARQL extensions that allow for exploration of multi-hop relationships in graphs.”

“The support for these extensions of the RDF and SPARQL is not implemented as a patch allowing us to check the box. RDF-star is already used by tools downstream and evaluations that prove efficiency improvement in managing Wikidata. RDF-star goes beyond the expressivity of LPG offering, not just key-value pairs, but rather the full flexibility of making statements about statements.”

“Ever since version 1.1 SPARQL property paths support graph traversal, allowing you to discover relationships between resources through arbitrary length patterns. Property paths uncover the start and end nodes of a specific path, but not the intermediate ones. There are standard complaint extensions of SPARQL now, which offer exploration of the paths and support all the different variants of the task, e.g. shortest path vs. all paths. And there are RDF engines that take advantage of their reasoning capabilities to score well at the LDBC Social Network Benchmark.”

“Having graph exploration covered, let us go back to the core requirements for knowledge graph management. RDF is recognized as the better option for knowledge graphs, because its web-native syntax supports data exchange and sharing and because its formal semantics allows for easy alignment of meaning and structure across sources, unified views and unambiguous interpretation. On the other hand, LPGs lack many features that are an absolute must for enterprise data management, e.g. schema language, data serialization formats and federation. On the semantics side, they lack ontology modelling language and data validation. What’s most important, there are no standards in the LPG space to guarantee interoperability and reduce vendor lock-in.”

“RDF engines check all the boxes: simple-yet-powerful graph model, standard schema and query languages, formal semantics, efficient graph traversal, analytics and reasoning, packed with all the enterprise features. There are a couple of cases where LPGs still have an edge: a micromanaged exploration using Gremlin and heavy analytics for wardrobes with TBs of RAM.”

Read more on the event website

Aaron Bradley

Aaron Bradley is Senior Structured Content Architect at Telus Digital, and chief cook and bottle washer at The Graph Lounge.

thegraphlounge.com