Skip to content

Cartography (V2)

A literature cartography of contemporary science. SciField V2 maps how research topics arise, move between journals, and trend over time across a broad multi-disciplinary journal panel. It is exploratory cartography — a descriptive atlas of the literature — not a hypothesis test.

Open the interactive map →{ .md-button .md-button--primary }

The map is a single self-contained page with five tabs. A tab-by-tab walkthrough is on the Using the map page; the underlying tables are documented in the Data dictionary.

What this is

The cartography is computed over an expansion panel of 78 journals, from a harvest of roughly 1.49 million papers (V2-S08). The published cartography layers are computed on the assigned corpus and resolve into 149 leaf topics. On top of that topic landscape it derives four further views:

  • Cascade — how a topic moves between journals (who leads, who follows).
  • Journal roles — where each journal sits in that diffusion network (source / bridge / terminal) and how fast its papers get cited.
  • Novelty origins — where novelty comes from, by institution sector and by country, plus a combinatorial-recombination signal.
  • Trajectories — a descriptive projection of where each topic's share and volume have been trending, extrapolated five years to 2030 with 80% bands.

A headline finding of the scaled run (V2-S09) is that the generalist "source" layer splits by journal type: PNAS ranks #1 source, while the high-impact clinical-prestige journals — NEJM is the most terminal, with Lancet, JAMA, Nature, and Science also terminal — behave as sinks rather than seeds. The inter-journal cascade is validation-real (the journal roles are highly stable under a leave-one-out jackknife, rank-correlation ≈ 0.9993).

Read this first — headline caveats

These govern how every view on the map should be read. They are not optional footnotes. The same caveats, per-table, are in the Data dictionary.

  1. Origins are PANEL-CONDITIONAL. Every "origin", "first appearance", "source", or "seeding" signal means first/source within the 78-journal panel — not first in the world. Several true generalist sources of science (e.g. Nature, NEJM, Lancet, Science) are read here only as they appear inside this panel; they cannot be credited as world-origins. Read every origin claim as "earliest observed in this panel".

  2. The corpus is 1995 LEFT-CENSORED. Coverage starts in 1995. Any topic or diffusion that began earlier has its onset clipped to the panel start, so onset-timing signals (origin year, seeding, time-to-50%) inherit that bias. Flow- and betweenness-based signals are comparatively robust to it.

  3. share is panel-conditional. A topic's share is its fraction of panel papers in a topic-year — not a fraction of all of science.

  4. Trajectories are a DESCRIPTIVE projection, not a forecast. The trajectory layer (V2-S10) extrapolates each topic's observed share/volume +5 years to 2030 with 80% uncertainty bands using a state-space local-linear-trend model (149/149 topics fit). It is not causal, not a validated predictive model, and is explicitly not the V1 F3 emergence GNN — that predictive model was signed NULL at Gate G4. Do not read the bands as a tested forecast.

  5. This is exploratory cartography, not a confirmatory result. The whole V2 program was gated as publishable exploratory work at Gate G6 (PASS, 2026-06-09). It describes structure in the panel; it does not test a pre-registered hypothesis.

Where things live

Citation & license

The derived cartography tables (~1.7 MB) ship in the repository under V2/data_v2/ and are documented in the Data dictionary. They are released under the repository's Apache License 2.0. The raw 1.49M-paper corpus is not redistributed; it is reproducible from the pipeline in this repository. Please cite the Zenodo record (DOI minted on the v0.2.0 release) and the software repository; see CITATION.cff at the repo root.