Transforming RDF-star to Property Graphs: A Preliminary Analysis of Transformation Approaches

Ghadeer Abuoda, Daniele Dell'Aglio, Arthur Keen, Katja Hose

Abstract

The integration of Semantic Web technologies with high-performance graph analytics requires effective data transformation between the Resource Description Framework (RDF) and Property Graphs (PG). With the advent of RDF-star, which enables metadata on triples, the gap between these models has narrowed, yet standardized transformation methodologies remain elusive. This paper formalizes two primary transformation strategies—RDF-topology Preserving Transformation (RPT) and Property Graph Transformation (PGT)—and evaluates current tools against a comprehensive 15-case benchmark. Our findings highlight a fundamental trade-off between structural integrity and query efficiency, revealing significant implementation gaps in existing software.

Introduction: The Graph Model Divide

The graph database landscape is currently bifurcated. On one side, the Resource Description Framework (RDF) excels in formal logic, data exchange, and standardized schemas. On the other, Property Graphs (PG) are optimized for high-performance path traversal and connection analysis. Historically, RDF struggled to represent metadata about relationships (reification) without significant structural bloat. The introduction of RDF-star addresses this by allowing "statements about statements," effectively bringing RDF closer to the PG model.

Comparison of RDF-star and Property Graph models. Note how the metadata (certainty) in the RDF-star graph (a) is often omitted or difficult to map in a standard property graph (b).

Despite this convergence, moving data between RDF-star and PG remains challenging. There is no industry standard for this translation, often resulting in either loss of information or the creation of overly complex graphs that hinder query performance. This research categorizes existing transformation approaches and identifies critical failure points in current toolsets.

Formalizing Transformation Strategies

The authors identify and formalize two distinct philosophies for mapping RDF-star data to Property Graphs:

  1. RDF-topology Preserving Transformation (RPT): This strategy maintains a strict one-to-one mapping where every RDF triple becomes an edge in the PG. Metadata associated with an RDF-star triple is directly mapped to properties on the corresponding PG edge. This approach prioritizes data integrity and structural fidelity but can lead to "graph bloat."
  2. Property Graph Transformation (PGT): This "idiomatic" approach distinguishes between "object properties" (relationships between entities) and "datatype properties" (attributes of entities). Object properties become edges, while datatype properties are stored as key-value pairs within nodes. This results in a cleaner graph but often struggles with complex metadata.
An RDF-star graph representing a complex statement: the relationship between Apple Inc. and its CEO Tim Cook is annotated with a start date.

These methodologies were implemented as four specific algorithms (RPT, PGT, and their "star" variants) to provide a theoretical baseline for evaluating existing tools such as RDF2PG, RDF-star Tools, and Neosemantics.

Benchmarking and Tool Evaluation

To rigorously test these approaches, the researchers developed a systematic benchmark consisting of 15 complex test cases. These cases cover a spectrum of challenges, including nested statements (triples describing other triples), predicate identity (where a resource serves as both a relationship and a subject), and data fidelity issues involving language tags and blank nodes.

Key Insights

The evaluation revealed a clear trade-off between structural integrity and performance. RPT-based tools (RDF-star Tools, RDF2PG) are robust at preserving data shape but produce complex graphs that can degrade query performance. Conversely, PGT-based tools like Neosemantics produce cleaner, more "native" property graphs but often fail to capture RDF-star metadata, especially when dealing with literal values or nested statements.

A significant discovery was the "Predicate Identity Crisis." Most tools fail when a resource is used both as a predicate and a subject, creating disconnected elements in the PG and breaking the conceptual link. Furthermore, even advanced tools struggled with "double-nested" statements, indicating that software implementations have not yet fully caught up to the RDF-star specification.

Conclusion and Future Directions

This preliminary analysis underscores that the choice of transformation depends heavily on the application. PGT is preferable for attribute-heavy machine learning tasks, while RPT is essential for formal reasoning and data integration.

Future research will focus on developing schema-aware transformations that incorporate RDFS and OWL axioms, as well as quantitative performance benchmarking. The authors suggest that a "hybrid" approach—using PGT for instance data and RPT for the ontology—might offer the optimal balance for future graph data management systems.

Summary updated

Abstract

RDF and property graph models have many similarities, such as using basic graph concepts like nodes and edges. However, such models differ in their modeling approach, expressivity, serialization, and the nature of applications. RDF is the de-facto standard model for knowledge graphs on the Semantic Web and supported by a rich ecosystem for inference and processing. The property graph model, in contrast, provides advantages in scalable graph analytical tasks, such as graph matching, path analysis, and graph traversal. RDF-star extends RDF and allows capturing metadata as a first-class citizen. To tap on the advantages of alternative models, the literature proposes different ways of transforming knowledge graphs between property graphs and RDF. However, most of these approaches cannot provide complete transformations for RDF-star graphs. Hence, this paper provides a step towards transforming RDF-star graphs into property graphs. In particular, we identify different cases to evaluate transformation approaches from RDF-star to property graphs. Specifically, we categorize two classes of transformation approaches and analyze them based on the test cases. The obtained insights will form the foundation for building complete transformation approaches in the future.
BibTeX
@inproceedings{abuoda2022transforming,
  author = {Ghadeer Abuoda and Daniele Dell'Aglio and Arthur Keen and Katja Hose},
  title = {{Transforming RDF-star to Property Graphs: A Preliminary Analysis of Transformation Approaches}},
  booktitle = {QuWeDa@ISWC},
  year = {2022},
}
Figure