Methodology
This page describes exactly where the data comes from, how the connections between topics are decided, and which numbers on this site are measurements versus editorial judgement. Read it before citing anything here.
1. The taxonomy backbone
The hierarchy derives from OpenAlex, an open catalogue of scholarly work published under CC0. OpenAlex maintains a four-level classification which maps onto this site's five levels:
| OpenAlex | Approx. count | Here |
|---|---|---|
| Domain | 4 | Discipline |
| Field | 26 | Domain |
| Subfield | ~252 | Sub-domain |
| Topic | ~4,500 | Topic |
| — | manual | Sub-topic (level 5) |
Level 5 has no OpenAlex equivalent. It is hand-curated, and currently exists in depth only for Geography and Environment.
The importer runs nightly, is resumable via a stored cursor, and never overwrites curated content — it enriches matching entries with identifiers and counts, and parks unmatched topics in an unpublished branch for manual review.
2. How cross-discipline links are decided
Hierarchy is structural and uninteresting — it falls out of the parent relationship. The links that matter are the ones that cross branches, and they come from three methods with very different reliability.
Method A — keyword overlap (automatic)
Jaccard similarity between the keyword sets of two topics in
different fields. Above a threshold, a link is written with
confidence derived. Cheap, broad, and noisy: it finds real
adjacencies and also coincidences of vocabulary.
Method B — co-classification of works (automatic)
Counts how often two topics are assigned to the same work. Normalised, this
is a genuine strength signal rather than a lexical one, and it is what
correctly attaches remote sensing to both geography and computer science.
Also written as derived.
Method C — manual curation (authoritative)
A link drawn by hand, marked verified, and given a
rationale sentence stating why the two connect. You see that
sentence on hover in the map and under every link on a topic page. The
automatic methods will never produce something like "coastal
geomorphology constrains what disaster policy can realistically protect"
— that requires knowing the field.
Link weight
Weight drives line thickness, on a 1–10 scale:
raw = 0.45·co_occurrence + 0.25·keyword_jaccard + 0.20·demand_signal + 0.10·manual_boost
weight = 1 + 9·raw, clamped to 1.0–10.0
A thicker line means more evidence of a relationship, not that the relationship is more important.
3. Scores that are editorial, not measured
Two numbers appear throughout the interface and are frequently misread, so to be explicit:
-
demand_score(0–100) sets node size. It is a comparative index assigned by hand, informed by publication volume and job-market presence. It is not a measured quantity. -
growth_score(0–100) drives the growth badge. Same caveat.
Once the OpenAlex importer has run against the full topic set, both can be
re-derived from real works_count and recent-year trends. Until
that happens, treat them as ordering hints.
4. Enrichment layers
| Layer | Source | Method |
|---|---|---|
| Tools & software | Vendor sites, OSGeo, curated lists | Manual |
| Careers (global) | O*NET, ESCO, job-title surveys | Manual + CSV import |
| Careers (Bangladesh) | BDJobs, government circulars | Manual, quarterly |
| Employers | Organisation websites | Manual seed list |
| Universities | ROR, OpenAlex institutions, UGC list | Semi-automatic |
| Programmes | University websites | Manual |
| Statistics | OpenAlex, World Bank, BBS | Script + manual |
5. Salary figures — read this before using them
Bangladesh salary bands were compiled from ranges published in public job listings. They are not survey data, not adjusted for experience or location within the country, and not verified against employers. They indicate an order of magnitude. Do not use them to negotiate, and do not cite them as labour statistics.
6. Provenance
Every manually entered fact is supposed to carry a row in the sources table with a URL, a licence and a retrieval date. That rule protects the project legally and makes it checkable.
Currently 0% of published entries have at least one source attached (0 of 0). The sources register lists every source and what it backs.
Licensing rules followed here:
- OpenAlex — CC0. Free to bulk-use, attributed anyway.
- Wikidata — CC0, accessed via API rather than scraped.
- University sites — consulted manually, never scraped in bulk; the source URL is stored.
- Paywalled databases (Scopus, Web of Science) — not used.
7. Known limitations
- Most cross-discipline links are machine-derived and unverified.
- Level 5 depth exists only for Geography and Environment.
- Bangladesh coverage is thin outside environmental and geographic fields.
- Demand and growth scores are editorial indices.
- Salary data is indicative only.
- Many entries have no source attached yet.
These are listed rather than buried because a map that hides its uncertainty is worse than one that admits it. If you can close any of these gaps, contributions are welcome.
How to cite
Connectivity System — Interdisciplinary Knowledge Network. Retrieved 5 August 2026, from https://ecosystem.aljubaer.com/