LoginSign Up
← Back to all sparks
R

refsplitr

ANALYTICS
Velocity5.0

R package for processing, organizing, and visualizing reference records downloaded from the Web of Science. - GitHub - ropensci/refsplitr: R package for processing, organizing, an

github.com/ropensci/refsplitr ↗

Author disambiguation for bibliometrics, still grinding on the hard part: which names are the same person.

bibliometricsauthor-disambiguationgeoreferencingropenscir-package
◆Current state
refsplitr parses Web of Science reference records into tidy data and tries to resolve which author strings belong to the same researcher, then georeferences their institutional addresses for network and map visualizations. The active work is squarely on the disambiguation core: 1.2.3 continues refining the author grouping algorithm and 1.2.1 adjusted ORCID ID matching. Earlier, 1.2.0 replaced the address parsing algorithm and changed the default georeferencing option for author institutions.
◆Where it's heading
Development has narrowed to the two operations that determine whether the output is usable — grouping author name variants and resolving addresses to coordinates. Everything else has been shedding: the maptools dependency was removed once that package was deprecated, and visualization changes are mostly about surfacing records the pipeline could not resolve, as with plot_net_country() returning fixable_countries so users can correct and rerun. Release notes are terse and defer to NEWS, so the changelog itself carries little detail.
◆Prediction
Expect further incremental passes on author grouping and ORCID matching rather than new outputs; that algorithm is the package's accuracy ceiling and the last several releases have all touched it.

◆Recent moves

  1. 28d ago

    Further refinement of the author grouping algorithm

    Another pass at the author grouping algorithm, the component that decides which name variants represent one researcher. The entry defers detail to NEWS, but this is the third consecutive release touching disambiguation accuracy, which is where the package's output quality is actually determined.

    View source ↗
  2. 29d ago

    Minor fixes and ORCID matching edits

    Small bug fixes plus adjustments to ORCID ID matching, released a day before 1.2.3. Described only in NEWS, and narrow enough that downstream results should be largely unaffected.

    View source ↗
  3. 1y ago

    New address parsing algorithm, changed georeferencing default

    Replaces the address parsing algorithm and changes the default option for georeferencing author institutions. Both sit upstream of every map and network output the package produces, so results generated before and after this release are not directly comparable.

    View source ↗
  4. 2y ago

    Unresolved countries surfaced, maptools dependency dropped

    plot_net_country() now returns authors that have coordinates but no resolved country via products$fixable_countries, so users can correct them and rerun rather than silently losing records from the graph. The deprecated maptools dependency was removed in the same release.

    View source ↗
  5. 6y ago

    rOpenSci release v0.9.0

    Marks the package's release following rOpenSci peer review, with 1.0.0 deferred until the accompanying publication details were added. A process milestone rather than a change in capability.

    View source ↗