Opportunities Citation Resolution

Status: Implemented into schema.

8 new tables + 1 column modification:

artifact_identifiers — N:1 alias concordance (any identifier resolves to P-number) artifact_identifier_evidence — standard evidence pattern for disputed identifications artifact_identifier_decisions — standard decision pattern for contested ID assignments publications — structured bibliographic registry with series membership and supersession chains publication_authors — structured authorship linking publications to scholars table artifact_editions — artifact-publication bridge with is_current_edition consensus flag and per-artifact supersedes chain artifact_edition_evidence — standard evidence pattern for publication links artifact_edition_decisions — audit trail for is_current_edition designation annotation_runs — 1 new column: publication_id (nullable FK to publications, alongside existing publication_ref freetext) 4 new migration steps added

Schema Update Summary

Changes

8 new tables, 1 new column on annotation_runs Table count: ~45 (pre-citation) -> ~53 (with citation resolution) 3 core tables + 2 supporting + 3 trust (evidence/decision), following existing patterns exactly

Tradeoffs

Dual resolution paths? Yes — publication references stored in both artifact_identifiers (fast single-index lookup) AND artifact_editions (rich metadata with page numbers, edition_type, currency). Accepted duplication for query flexibility. Two supersession levels? Yes — publications.supersedes_id for series-level ("RIME 4 supersedes VAB 6"), artifact_editions.supersedes_id for per-artifact granularity (one pub may supersede another for some texts but not all). Freetext parsing? Partial coverage initially. CDLI publication_history is inconsistent freetext. Import what regex can parse, preserve raw string, flag unparsed entries. NO data loss. publication_id vs publication_ref? Both — nullable FK added alongside existing freetext. Incremental migration: NULL until matched, freetext stays as fallback for unresolved citations. Flat columns stay? Yes — artifacts.museum_no and excavation_no remain for display convenience. artifact_identifiers is authoritative for resolution queries.

What This Enables

Two resolution paths for any citation:

Path A (fast): artifact_identifiers WHERE identifier_normalized = 'rime 4.3.6.1' -> P363653 Path B (rich): artifact_editions WHERE reference_normalized = 'rime 4.3.6.1' -> P363653 + edition_type, pages, currency

Then: full edition history via recursive supersedes_id chain VAB 6.1 (1907, hand_copy) -> RIME 4.3.6.1 (1990, full_edition) -> Oracc/OBMC (2010, digital_edition)

Key Queries Enabled

• Resolve any identifier format to a P-number (museum_no, excavation_no, publication, accession, ARK)
• Find current edition of any artifact
• Get full edition history with supersession chain
• Find ambiguous identifiers (same string -> multiple P-numbers)
• Research gaps: texts without digital editions
• All publications by a scholar
• All artifacts published in a given series
• Direct bibliographic link from any token_reading/lemmatization/translation via annotation_runs.publication_id

Trust Dimensions

• artifact_identifiers: Trust that two identifiers refer to the same physical object
• publications: Trust in the scholarly record itself (supersession makes edition authority explicit)
• artifact_editions: Trust in edition currency (is_current_edition is the consensus flag with full decision audit trail)

Seed Data Sources

• artifacts.museum_no (~353k rows), excavation_no, primary_publication
• CDLI API publications[] array (structured bibtex)
• CDLI CSV publication_history (partial regex parse, raw preserved)
• CDLI accession_no, external_id fields
• ORACC project membership (one digital_edition per project)
• ORACC catalogue cross-references
• Manual curation for major series (RIME, SAA, RINAP, VAB, PBS, ATU)

What This Does NOT Yet Cover (Future Work)

• Citation Resolution API (REST/GraphQL endpoints)
• Fuzzy matching with Elasticsearch/Levenshtein
• Browser extension for auto-linking citations in web pages
• Citation generation (Chicago, APA, BibTeX formatting)
• Citation graph (which publications cite which texts)
• Subscription/notification system for new editions
• ML-based citation extraction from PDFs These remain in the vision doc (Assyriology Opportunities.md, section 6). The schema additions provide the data model foundation they will build on.