The Symbolic World: Verification-First Knowledge Extraction from Interpretive Corpora (dataset, pipeline, and paper)

A methodology for AI-assisted knowledge extraction in which trust is manufactured by construction and terminates in human judgement, demonstrated on a knowledge graph of C. G. Jung's Collected Works: 248 nodes, 669 edges, 699 references, each verbatim-anchored to a numbered paragraph, independently gated by a second model family, calibrated with seeded corruptions (n=200, published confidence intervals), cross-vendor audited (full-rubric and reduced-rubric), human-validated on a 50-item author gold set (E10), and replicated at pilot scale on a public-domain corpus (James's Varieties, complete package included), with live public dispute and confirmation channels. Includes the full transaction ledger, all experiment inputs/verdicts/adjudications (E0-E10), and the paper. Copyrighted source texts are not included; quotations are <=25 words under quotation exceptions (see docs/LEGAL.md). Live atlas: https://symbolicworld.observer