[Paper Review] Annex: Radon - Rapid Discovery of Topological Relations

Mohamed Ahmed Sherif, K. A. Dressler+2 more|arXiv (Cornell University)|Nov 18, 2016

Semantic Web and Ontologies13 references3 citations

TL;DR

Radon is a novel, high-performance system for rapidly discovering topological relations between geo-spatial RDF resources using a sparse tiling index and minimum bounding box (MBB) filtering. It achieves up to 3 orders of magnitude speedup over state-of-the-art tools like Silk and Strabon while maintaining 100% F-measure accuracy on real-world datasets.

ABSTRACT

Datasets containing geo-spatial resources are increasingly being represented according to the Linked Data principles. Several time-efficient approaches for discovering links between RDF resources have been developed over the last years. However, the time-efficient discovery of topological relations between geospatial resources has been paid little attention to. We address this research gap by presenting Radon, a novel approach for the rapid computation of topological relations between geo-spatial resources. Our approach uses a sparse tiling index in combination with minimum bounding boxes to reduce the computation time of topological relations. Our evaluation of Radon's runtime on 45 datasets and in more than 800 experiments shows that it outperforms the state of the art by up to 3 orders of magnitude while maintaining an F-measure of 100%. Moreover, our experiments suggest that Radon scales up well when implemented in parallel.

Motivation & Objective

• Address the lack of scalable, efficient methods for discovering topological relations between geo-spatial resources in RDF/Linked Data.
• Overcome the computational bottleneck of pairwise topological relation computation in large-scale geo-spatial datasets.
• Develop a method that maintains completeness and correctness (F-measure = 100%) while drastically reducing runtime.
• Enable real-time applications such as machine learning and question answering on large geo-spatial knowledge graphs.
• Provide a scalable, parallelizable solution suitable for modern, large-scale geo-spatial Linked Data workloads.

Proposed method

• Proposes a novel sparse space tiling index to efficiently organize and access geo-spatial resources in RDF datasets.
• Employs minimum bounding box (MBB) approximation to pre-filter pairs of geometries that cannot possibly have topological relations.
• Uses the DE-9IM model as the formal basis for defining and computing topological relations between two geometries.
• Applies a filtering strategy that leverages MBB intersections to avoid expensive full DE-9IM computations for non-overlapping or non-adjacent geometries.
• Integrates the sparse tiling index with MBB filtering to reduce the number of full topological relation checks.
• Supports parallel execution with a round-robin load balancing strategy, showing strong horizontal scalability in experiments.

Experimental results

Research questions

• RQ1 Can a sparse tiling index combined with MBB filtering significantly reduce the time required to compute topological relations between geo-spatial RDF resources?
• RQ2 How does Radon’s performance compare to state-of-the-art systems like Silk and Strabon in terms of runtime and scalability?
• RQ3 To what extent does the filtering strategy based on MBBs reduce the number of full DE-9IM computations?
• RQ4 Does Radon maintain completeness and correctness (i.e., 100% F-measure) while achieving high performance?
• RQ5 How does Radon scale when processing very large geo-spatial datasets, especially when compared to systems that time out on such workloads?

Key findings

• Radon outperforms the state of the art (e.g., Silk) by up to 3 orders of magnitude in runtime across 45 datasets and 800+ experiments.
• On average, Radon is 11.99 times faster than Strabon, with significantly better performance on all relations except 'intersects', where Strabon's R-tree index gives it an edge.
• In scalability experiments with the large CLC_m dataset, Radon completed in approximately 95.10 minutes, while Strabon timed out after 2 hours and would require an estimated 20.17 hours (under linear scaling) or 241x speedup (under O(n²) scaling) to complete.
• The MBB-based filtering strategy in Radon reduces the number of full DE-9IM computations more effectively than existing approaches, contributing to its performance gains.
• Radon maintains 100% F-measure across all experiments, proving it is both correct and complete in discovering topological relations.
• The system scales well in parallel execution, with potential for further performance gains through advanced load balancing techniques like particle swarm optimization.