Difference between revisions of "Document the provenance of the results"

Revision as of 17:13, 16 April 2015

What This Task Involves

The training session and training materials indicate how to:

1. Capture the provenance of the results in a paper
2. Develop a workflow sketch, a formal workflow, or a provenance record that represent to different degrees of accuracy what the provenance of the results is
3. Publish the provenance and make it part of a publication

Training Materials

This training session was held on March 6, 2015:

• Presentation

Suggested Readings

• “Requirements for Provenance on the Web.” Paul Groth, Yolanda Gil, James Cheney, Simon Miles. International Journal of Digital Curation, 7(1), 2012.
  • A general overview of provenance'

• "A Primer for the PROV Provenance Model." Yolanda Gil, Simon Miles, Khalid Belhajjame, Helena Deus, Daniel Garijo, Graham Klyne, Paolo Missier, Stian Soiland-Reyes, and Stephan Zednik. Published as a W3C Working Group Note on 30 April 2013.
  • A brief and practical introduction to the PROV standard for provenance, showing examples of how to represent the provenance record in RDF through a simple notation called Turtle

• "Intelligent Workflow Systems and Provenance-Aware Software." Gil, Y. In Proceedings of the Seventh International Congress on Environmental Modeling and Software, San Diego, CA, 2014.
  • A brief introduction to workflows for scientists, giving examples and explanations of their benefits

What To Do

We described many options in the training. Here is a sketch of the most common approach:

1. At the very minimum, describe the workflow in the text (a "Methods" section) or in an appendix
   • Mention the datasets used, the software, and the data flow across the software components
   • Specify unique identifiers for data and software, mention the version used, credit all the sources
2. Develop a workflow sketch and show it in a figure or in an appendix
   • Capture high-level dataflow across components
3. To really capture the full provenance, specify the formal workflow or provenance record
   • The formal workflow shows all data flow across components, corresponding to the detailed command line invocations and parameter values used
   • Options:
     1. Describe it as a graph where the nodes are computations and the links show data and parameters
     2. Use the PROV provenance standard (start with a result and trace back how it was generated)
     3. Use a workflow system (e.g. WINGS) to create the data flow graph
   • Publish the formal workflow or provenance record, and assign a unique identifier
     • Cite it in the paper
     • Show the provenance graph

= Using the WINGS Workflow System to Document Provenance

Documentation on how to use the WINGS workflow system:

• A tutorial
• Introductory papers
• A portal where you can get an account and try it out
• Downloading the code to set up in a local machine

Some examples of workflows created with WINGS for GPF papers:

• Kyo's provenance