Published papers often provide insufficient information about the data, protocols, software, and overall method used to obtain the new results in computational experiments. A major challenge in reproducibility is the lack of appropriate support for authors to capture exactly how experiments were performed. Once the work is finished, authors write an account in their articles by retrospective reconstruction of the work that was actually done, relying on their memory and notes kept along the way. However, we should have automated tools that ensure that the descriptions that are written about computational experiments are in fact accurate and provide enough detail for transparency and reproducibility.
In our previous work we added to WINGS the functionality to capture the full provenance and specification of scientific workflows. Now we have used that information to generate data narratives of results, i.e., textual explanations that describe them at different levels of granularity.
Data narratives have three major components:
1) A record of events, that describe a new result through a workflow and/or provenance of all the computations executed;
2) Persistent entries for key entities involved through DOIs for data, software versions, and workflow;
3) a narrative account (or several), that is an automatically generated human-consumable rendering of the record and entities and may include visualizations or summaries.