On Reasoning with RDF Statements about Statements using Singleton Property Triples

Vinh Nguyen, Olivier Bodenreider, Krishnaprasad Thirunarayan, Gang Fu, Evan Bolton, Núria Queralt Rosinach, Laura Inés Furlong, Michel Dumontier, Amit P. ShetharXiv 2015 · arXiv

Abstract

Representing metadata in the Resource Description Framework (RDF) has historically been a challenge, often requiring complex reification or non-standard extensions. This paper introduces a simple yet powerful mechanism to enable automated reasoning over "statements about statements" using the Singleton Property (SP) approach. By introducing a single primitive axiom, the authors demonstrate that existing Semantic Web reasoners can be adapted to recognize metadata-rich triples and infer underlying data. Experimental results show that this method maintains compatibility with current standards while incurring only a modest performance overhead of 10-12%.

Introduction

In the Semantic Web, RDF excels at representing binary relationships as triples (e.g., ProfessorA worksFor University1). However, real-world datasets frequently require contextual metadata, such as temporal validity (when the fact was true) or provenance (who asserted the fact). Standard RDF reification is often criticized for being verbose and, more importantly, lacking formal entailment—a reasoner does not automatically infer the truth of a reified statement.

The Singleton Property (SP) approach was proposed to address these limitations by creating a unique, one-time-use property for every specific context (e.g., worksFor#1). While SP provides a structure for metadata, the challenge remains: how can standard reasoners be "taught" to understand that these unique properties are instances of generic ones without modifying the underlying reasoning engines?

The Singleton Property Approach

The SP approach represents a statement about a statement by introducing a unique predicate for each occurrence of a relationship. This unique predicate is then linked to the generic property it represents.

Representation of metadata using Singleton Property triples.

As shown in Figure 3, a relationship like worksFor is instantiated as worksFor#1. Metadata such as start and end dates are then attached directly to this singleton property. This allows for a rich description of individual facts while keeping the graph structure relatively flat compared to traditional reification.

The Reasoning Mechanism

The core innovation of this work is a mechanism to bridge the gap between metadata and data using a new primitive axiom:

rdf:singletonPropertyOf rdfs:subPropertyOf rdfs:subPropertyOf

This axiom leverages the existing RDFS subPropertyOf hierarchy to enable a two-pass reasoning process:

  1. Property Inference: The reasoner identifies that a singleton property (e.g., worksFor#1) is a singletonPropertyOf a generic property (worksFor). Due to the new axiom, it infers that worksFor#1 is itself a subPropertyOf worksFor.
  2. Data Inference: Using standard RDFS inheritance rules, the reasoner then infers the original data triple (ProfessorA worksFor University1) from the singleton triple (ProfessorA worksFor#1 University1).

This approach ensures that metadata remains "transparent" to the reasoner, allowing it to derive base facts automatically. Furthermore, it integrates seamlessly with existing RDFS hierarchies.

Integration of inferred Singleton Property facts with standard RDFS reasoning.

As illustrated in Figure 5, once the base triple is inferred, standard RDFS rules continue to apply. If worksFor is defined as a sub-property of memberOf, the reasoner can further infer that ProfessorA memberOf University1.

Experimental Evaluation

To evaluate the performance and scalability of this approach, the authors developed LUBM-SP, an extension of the Lehigh University Benchmark (LUBM). LUBM-SP adds temporal metadata to a significant portion of the relationships.

The implementation was tested using Apache Jena and Oracle 12c. The largest dataset contained over 1.3 billion triples and 257 million singleton properties. Key findings include:

  • Reasoning Overhead: While the initial indexing (entailment) phase takes longer due to the massive increase in the property hierarchy, the process remains manageable for modern database-backed stores.
  • Query Performance: Query execution times on LUBM-SP were only 10% to 12% slower than on the standard LUBM. This indicates that the cost of maintaining rich, queryable metadata is relatively low once the initial reasoning is complete.
  • Compatibility: The method works with standard SPARQL, allowing users to query both raw data and metadata within the same pattern.

Conclusion

The Singleton Property reasoning mechanism provides a standard-compliant way to handle statement-level metadata in RDF. By using a clever "axiom trick," the authors enable powerful inference without requiring changes to existing reasoning software. Future work will focus on incorporating constraint enforcement (e.g., temporal logic) and further optimizing the reasoning process for massive property hierarchies.

Summary updated

Abstract

The Singleton Property (SP) approach has been proposed for representing and querying metadata about RDF triples such as provenance, time, location, and evidence. In this approach, one singleton property is created to uniquely represent a relationship in a particular context, and in general, generates a large property hierarchy in the schema. It has become the subject of important questions from Semantic Web practitioners. Can an existing reasoner recognize the singleton property triples? And how? If the singleton property triples describe a data triple, then how can a reasoner infer this data triple from the singleton property triples? Or would the large property hierarchy affect the reasoners in some way? We address these questions in this paper and present our study about the reasoning aspects of the singleton properties. We propose a simple mechanism to enable existing reasoners to recognize the singleton property triples, as well as to infer the data triples described by the singleton property triples. We evaluate the effect of the singleton property triples in the reasoning processes by comparing the performance on RDF datasets with and without singleton properties. Our evaluation uses as benchmark the LUBM datasets and the LUBM-SP datasets derived from LUBM with temporal information added through singleton properties.
BibTeX
@misc{nguyen2015reasoning,
  author = {Vinh Nguyen and Olivier Bodenreider and Krishnaprasad Thirunarayan and Gang Fu and Evan Bolton and Núria Queralt Rosinach and Laura Inés Furlong and Michel Dumontier and Amit P. Sheth},
  title = {{On Reasoning with RDF Statements about Statements using Singleton Property Triples}},
  year = {2015},
  eprint = {1509.04513},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1509.04513},
}
Figure