Co-author map: your research network on a world map

Co-author map draws the people you have published with on a world map. Give it your ORCID iD and it does the rest. It runs entirely in your browser: no account, no sign-up, no API key, and nothing you type is sent anywhere except to the public APIs it reads.

The map below starts on my own record. Put your ORCID iD in the box and press Load this researcher to see yours, right here on this page. Each circle is a city, and its area is proportional to the number of co-authors based there. Hover a circle for the names.

Why it starts from claimed works

The obvious design would be to search a bibliographic database for your name and map whatever comes back. That does not work.

Looking up one real ORCID iD in OpenAlex returns two author records for the same person, and one of them carries papers from 1920, 1972, and 1988 attached to a researcher who started publishing in 2019. Roughly a quarter of that record belongs to other people who share the name, and nothing in the response says so.

Works you have claimed do not have this problem, because you curated them. For the same researcher, ORCID returns 34 papers, all genuinely theirs, and only 2% of the author rows are missing an affiliation, against 24% via the author-record route. Searching by name is still available for people with no claimed works, and it warns you that the result will be noisy.

Three things that had to be fixed before the map was honest

One person, several author records

Bibliographic databases split people. On my own list, 21 of 166 co-author records were duplicates of someone already there, a 13% inflation that made the largest circle 27% too big. The split has a consistent shape: a record holding several papers, plus a record holding exactly one, usually a paper indexed recently that has not yet been folded back into the existing cluster.

Two records are treated as one person when they share an ORCID iD, or when all three of these hold: the same name, an organization in common, and no paper on which both appear. That last condition carries the weight. A paper never lists the same person twice, so two records appearing together on one paper are certainly two different people. It is the only certain test available, so it runs first.

Pins are cities, not institutions

OpenAlex stores institution coordinates at city granularity. Fifteen Tokyo institutions sit on exactly the same point, and querying ROR directly returns that identical city centroid. Separating institutions geographically is impossible with this data, and the tool does not pretend otherwise: a circle is a city, and the organizations inside it are listed in the tooltip.

Cities need merging too. Without a proximity rule, Oxford splits into three circles, because the university, the hospital, and a nearby company sit a few kilometers apart, and each one claims the same ten papers.

Consortium headquarters are not workplaces

Papers often list a national research consortium alongside the author's university. Those consortia have a registered head office, and a naive map puts the person at that address. On my own network this produced a cluster of seven people in one German city and four in another, none of whom work there; they are all in the same Munich department.

The fix is to prefer institution types that are actually workplaces, universities and hospitals, when the author lists one. That moved 18 people and removed three cities that never existed. Researchers who genuinely work at a company or an institute keep that affiliation, because the rule only applies when a university or hospital is listed as well.

One person, one circle

Each person appears exactly once, at the affiliation printed first on the paper, which is the convention the literature already follows. Without that rule, 38% of my co-authors appeared in more than one city, and one appeared in six at the same time.

Circle area is proportional to the number of people by default. The alternative is worth a look:

City	Co-authors	Papers
Munich	25	13
Kyoto	18	17
Bern	3	14

Sized by people, Munich dominates. Sized by papers, Bern does. Neither is wrong, and reading one as the other is easy, which is why the line under the map always says which one you are looking at.

What it costs to draw

Seven requests and about a second and a half, for a record with 34 papers and 145 co-authors. Results are cached for a day, so reopening the page costs nothing. Because the requests come from your own browser rather than a server in the middle, rate limits are never a shared resource and nobody collects your queries.

Limits worth knowing

• City granularity is the floor. Campus-level pins are not possible with the available data.
• The affiliation is the one printed on the paper, not where the person works now. The map records where the collaboration happened.
• Authors listed without any affiliation cannot be placed. The tool counts them and says so rather than dropping them quietly.
• Papers with no DOI are not included, because the DOI is what links a claimed work to the bibliographic record.
• Very large consortium author lists will dominate the circles. The corrections panel exists for that.

Putting it on your own site

The map above is not a picture, and it is not a cut-down preview either. Everything the tool does happens in it: load a record, correct what the matching got wrong, read the result as tables, and open Embed this map to get the block that produces exactly what you are looking at. Paste that into your lab page and you are done.

Two things about that block are worth knowing. It holds no publication data, so the map is rebuilt from your ORCID record every time someone loads your page: claim a paper next year and it appears, with its co-authors and their cities, without you touching the page again. And the credit line is not required — the code is MIT-licensed and asks for no link back, so remove or reword it as you see fit.

Source

The code is MIT-licensed and lives at github.com/ykfrkw/coauthor-map. Data comes from OpenAlex and the ORCID public record, both CC0, from researchmap, and from Natural Earth for the country outlines. No API key is needed for any of them.

If the map gets something wrong about your network, the corrections panel will let you fix it, and an issue telling me what it got wrong is welcome.