SciMaps.
Create an account

SciMaps guide

From Papers to Research Maps: A Reproducible SciMaps Example

Published by SciMaps · Updated

Reproduce co-citation and title-term networks from a frozen collection of 20 real science-mapping publications, with documented inputs, settings and analytical checks.

One collection, two different questions

Can we reproduce a research map from known inputs rather than an ever-changing search result? This example uses 20 real publications about science-mapping methods and software, frozen from OpenAlex on 15 October 2026. It compares references cited together with terms appearing together in titles.

This is a deliberately selected teaching collection, not a systematic review, a ranking of tools, or a claim about the whole field. The downloadable data and automated checks make its boundaries inspectable.

Download the collection and evidence

OpenAlex makes its metadata available under CC0. We retain identifiers, titles, years, citation counts and reference IDs; no paper full text or abstracts are included. The selection includes foundational coupling and co-citation papers alongside mapping methods and software studies. Topic-specific applications and unrelated meanings of “mapping” were excluded from this teaching sample.

Illustrative example

Unlike an invented A/B/C example, these counts come from the real SciMaps analysis services using the frozen collection. The co-citation output contains 20 paper nodes and 83 undirected edges, with isolated corpus members retained.

  • Kessler's Bibliographic coupling between scientific papers and Small's Co-citation in the scientific literature share seven citing works in the selected collection. The edge weight is 7; it does not mean the papers cite one another seven times.
  • A unified approach to mapping and clustering of bibliometric networks and the VOSviewer software survey share six citing works in this collection.

Both members of each pair are themselves selected corpus records. SciMaps' paper co-citation view does not expand all external cited references into nodes. An absent reference can therefore be outside the selected collection, rather than evidence that it was never cited. See Co-citation Analysis for the workflow and co-citation versus coupling for choosing the question.

Title terms describe a different relationship

Using Title terms, minimum document frequency 2, minimum co-occurrence 1, quality threshold 1, and the explicit exclusion between, the keyword output has 20 documents, four retained term nodes and two edges. Isolates are hidden, the top-term limit is 50, and no per-node edge cap is applied. K-means uses three clusters; cluster IDs and colors are not reproduction targets.

The term science mapping occurs in four titles; software tools occurs in two. They occur together in one title: Science mapping software tools: Review, analysis, and cooperative study among tools. Their edge weight is 1, counting a shared document, not the number of words or repetitions. The other edge connects citation (three documents) with scientific literature (two), with weight 1.

The real processing pipeline normalizes and filters terms and prefers some multiword phrases. We explicitly exclude “between” because that generic connector survives the current normalizer. These are extracted title terms, not author-supplied keywords or OpenAlex's newer generated keyword field. See Keyword Co-occurrence for interpreting the map.

Increasing minimum co-occurrence to 2 produces an empty keyword graph on this collection. This is a useful threshold lesson: a term can occur in several documents while no retained pair meets the stronger shared-document threshold. An empty graph is not proof that the subject has no structure.

Three relationships reproduced from the frozen science-mapping collection Co-citation: Kessler and Small share seven citing works, unified mapping and VOSviewer share six. Title terms: science mapping and software tools occur together in one title. Analytical excerpt, not an application screenshot or complete graph. Reproduced analytical relationships Co-citation · shared citing works within the selected collection Kessler (1963)Bibliographic couplingSmall (1973)Co-citationUnified mappingand clustering (2010)VOSviewersoftware survey (2010)7 shared citing works6 shared citing works Title terms · each retained term appears in at least two documents science mapping · 4software tools · 21 shared document Analytical excerpt; not an app screenshot. The full co-citation graph has 83 edges.
Three verified relationships, shown as an analytical excerpt rather than an app screenshot. The downloads contain the full co-citation edge list.

Reproduce the frozen result

  1. Download the collection, frozen metadata, settings, expected results and checksums. Keep the files together.
  2. For exact reproduction, use the SciMaps source release containing this case study and its ReproducibleBibliometricCaseStudyTest. Follow the README with Java 25, Maven and the normal installed project prerequisites.
  3. Run mvn -o -q test -Ppublic-site-fast -Dtest=ReproducibleBibliometricCaseStudyTest from the repository root after resolving build dependencies. The fixture supplies frozen records to the real analysis services; no database, API keys or live metadata requests are needed.
  4. Compare the verified counts and relationships with expected-results.json. The tests also check file hashes and the empty keyword result at the stronger threshold.

The production analysis baseline is revision 9d24f3010cef22ca437686f304186842679a4303; the case-study release adds the data and test to that baseline. The fixture mocks storage and external metadata boundaries while retaining the real co-citation computation, term extraction and normalization. It validates analytical results, not the browser's graph layout.

For a product walkthrough, create a bibliometrics project, add the DOI records from the CSV through the supported Sources flow, select them and apply the documented settings. Live DOI retrieval may return changed metadata and different counts. The downloadable JSON is a developer fixture, not a supported production import format. Exact reproduction uses the frozen-data test.

Individual research example

A researcher preparing a methods workshop can inspect why two foundational references are linked, then check the shared citing records against the frozen reference lists. They can separately inspect the title containing “science mapping” and “software tools”. Each relationship becomes a prompt to read the sources, not an automatic conclusion about agreement or importance.

Institutional example

A librarian can distribute the same CC0 metadata and settings to a class, run the offline checks, and demonstrate how a higher threshold removes the small title-term network. Every participant starts with the same selected corpus. This exercise does not evaluate institutions, staff performance or the entire literature.

Limitations

Provider reference coverage and matching can be incomplete: the frozen Kessler record has only one recorded reference. No reference lists were manually repaired and no citation edges were invented. The 83 co-citation edges are relationships within this intentionally selected corpus; they are not exhaustive field coverage.

A shared-document weight of one is weak evidence used transparently for a small teaching example. Titles omit concepts that abstracts or full text may discuss. Selection, missing metadata, exclusions and thresholds all affect the maps. Coordinates, cluster colors and cluster assignments are not asserted. The diagram is an analytical excerpt, not a SciMaps screenshot or a complete network.

References

Related tools

Related guides

Try the workflow in SciMaps

Build a project collection and select the sources for your analysis.

Create an account

Create your account, then complete verification to start using SciMaps. Already have access? Sign in to SciMaps.