NuBerea/intertextuality

Tasks

• feature-extraction

Citation text from the card

`scripture_citations_all` adds 349,888 additive rows in seven arms, one arm per published citation member (`edge_kind`). The member configs remain published and unchanged. The row count is the sum of the member rows; it is not a count of unique citations, because `sch-gold-crossref` repeats the 55,869 `sch-scripture` edges and both arms are kept. `edge_id` is unique within each arm except the two Scholastic arms, where 128 ids repeat exactly as published in the members.

`osisRef` keeps each arm's native reference (Hebrew numbering on `talmud-ot` and `targum-ot`) and `osisRef_std` the English standard reference. Where the member left a reference empty it stays empty: 916 `talmud-ot` rows and 137 `targum-ot` rows have NULL in both, because their candidate set was ambiguous and was never guessed. `match_grain` is the union's grain envelope (`verse` or `segment`); the native grain of every source row is kept in `member_match_grain` (`verse`, `chapter` or `verse_citation`), and 127,416 rows are chapter-level citations.

Each arm keeps the licence class of its source component in the per-row `k` column: `computed` on the Reformation, Talmud and Targum arms and `public-domain` on the Scholastic arms. The table ships citation identifiers, labels, scores and bibliographic locators only; no Talmud, Targum or other source text is shipped. README metadata is not a grant for any component.

| Arm (`edge_kind`) | Source config | Rows | `k` | Source revision | Content |
|---|---|---:|---|-

All datasets marked Permissive or share-alike | JSON