Rosetta as printed

Names and terms in the same records

Rule: vocabulary term (aliases OR-folded) × correspondent, FOLDER-level co-occurrence: the term's whole-word hit on a folder text × the person's letters in that folder — never letter-level (letter boundaries are not delimited beyond the anchors)

Order: distinctive first (lift) — records * N / (deg_term * deg_person); N = total fts records; degrees = graph_term_records / graph_person_records counts; support floor 5.

On this page: Order is “distinctive first” above the reviewed floor of 5 records, then by count.

SELECT t.slug FROM graph_term_records t JOIN graph_person_records p ON p.slug = t.slug WHERE t.term = :term AND p.person = :person ORDER BY t.slug

Corresponded with

Rule: letters carrying both parties (the boundary sentence), grouped by the unordered pair of correspondent groups; see the builder A correspondence is counted only where both a sender and a recipient were read from the letter itself; letters with one party read are in no edge.

Correspondence weighting: Shared letters divided by the smaller of the two correspondents' totals of letters sent and received. Support floor 5. provisional; the EC alongside floor reused as the correspondence floor.

SELECT id FROM letters WHERE (sender_group = :person_a AND recipient_group = :person_b) OR (sender_group = :person_b AND recipient_group = :person_a) ORDER BY year, id