The data is a common storage of all research cartography results. However, an approximate abstraction is a large table with a row for each paper (ever) and a column for each property we might use to differentiate papers. In essence, this is a system for reliably desegregating pieces of research through a systematic process of adding columns.
Features are properties of a research effort that we want to document, e.g., title
, authors
, N
of experiments, effect_size
of each condition etc.
Parents frame the unit of analysis for a feature; for example, "effect_size
of each condition" has the parent condition
because each condition has a result for this feature. Similarly, condition
has the parent experiment
. The highest parent at this point is paper. We also sometimes call these scopes.
Features have validation information, such as ground truth ratings from researchers, which can be aggregated into performance metrics on a per-feature basis.
Projects — coming soon with #79
Projects are built from a bundle of features and papers (or a rule for including papers).
Effectively, a saved search that returns papers. The papers can then be added to a project individually or as a group. Updates to the results of the search instigate notifications to inform the user.
Papers are research output documents (such as published papers, arXiv preprints, proceedings, and white papers). They can incorporate other documents, such as supporting information (SI), appendices, registrations, data files, etc. Where possible, bibliometric indices, e.g., DOI, are used to uniquely identify papers in our data. Atlas can not be used to download or share papers.
Please submit a PR improving Atlas or adding new feature columns.