What is betweenness centrality? It measures how often an author sits on the
shortest co-authorship path between two other authors who aren't directly connected. A high score
means this person structurally bridges parts of the network that would otherwise be far apart —
e.g. someone who collaborates across two research groups that rarely work with each other directly.
This is a different signal from being prolific: a "connector" isn't necessarily the author with the
most papers or the most co-authors, just the one whose position links otherwise-separate clusters.
Computed approximately (networkx, sampled) over authors with at least one repeated (weight≥2)
co-authorship, since one-off single-paper collaborations are noisier signal for this kind of
structural measure.
What is affiliation coherence? This export has no ORCID or email, so authors are
merged by name + affiliation text only. A common name can belong to several unrelated real people
who get wrongly merged into one node — affiliation coherence checks whether that node's affiliation
strings actually describe the same institution (allowing for wording drift), which is a proxy for
"is this really one person". Low-coherence nodes are excluded from this ranking entirely.