U.S. Department of Energy

Pacific Northwest National Laboratory

Scalable Hypothesis Reasoning (SHyRe)

Objective

SHyRe is exploring the effects of streaming data on the Semantic application stack and how volatile streaming data affects well-known reasoning platforms within the Semantic application stack.

Approach

Our approach began with building on the results created during FY14 before SHyRe’s change of primary investigator. SHyRe currently includes both Description Logic ontologies developed in FY14 and a new Resource Description Framework Schema (RDFS) version for nuclear magnetic resonance (NMR) reasoning. SHyRe also has new test ontologies for the Strategic Surprise use case, including representations for line of business and change detection. Next, we focused on the stream components and how reasoning evolved with different complexity streams. We were able to show how competing reasoning packages were prioritized. Finally, we developed an architecture pattern that can be applied to ingesting streaming data independent of reasoning service.

Achievements

  • By consuming an undefined count of scans, we assembled an NMR run, modeled compounds within an ontology of background data, and then reasoned across this new combined model of compound and spectrum ontology
  • Modeled Description Logic background ontology of NMR metabolite classes and peaks (Pellet implementation)
  • Modeled RDFS background ontology/queries of NMR metabolite classes and peaks (StarDog/AllegroGraph implementations)
  • Created the reasoning structures that enable user intervention, interaction, and integration with other models
  • Compared reasoners and logic models across use cases and the effects of choices of these components on streaming data
  • Created a software design pattern for assembling a graph from a stream of data. Each graph was assembled from an undefined number of data packets

Impact

SHyRe is a unique project researching all aspects of the engineering pattern required to create a stream reasoning engine using the Semantic Web stack. Over the course of two years, SHyRe has published three papers, has two papers in preparation, deployed four live prototypes, created three ontologies across two use cases, and is currently collecting metrics on the performance of three different reasoners.

 

| Pacific Northwest National Laboratory