Similarity search for scientific workflows

• DOI: 10.14778/2732977.2732988
• Publication Date: 2015-05-12T15:37:52Z
• Authors: Johannes Starlinger; Bryan Brancotte; Sarah Cohen-Boulakia; Ulf Leser
• Keywords: Similarity (geometry)

With the increasing popularity of scientific workflows, public repositories are gaining importance as a means to share, find, and reuse such workflows. As sizes these grow, methods compare workflows stored in them become necessity, for instance, allow duplicate detection or similarity search. Scientific complex objects, their comparison entails number distinct steps from comparing atomic elements whole. Various studies have implemented workflow came up with often contradicting conclusions upon which algorithms work best. Comparing results is cumbersome, original mixed different approaches used evaluation data metrics. We contribute field (i) by disecting each previous approach into an explicitly defined comparable set subtasks, (ii) isolation taken at step comparison, reporting on unexpected findings, (iii) investigating how can best be combined aggregated measures, (iv) making available gold standard over 2000 ratings contributed 15 experts corpus almost 1500 re-implementations all we evaluated.