Prosopography: expansion plan

implemented-prior-stage · Depends on: initial audit

Completed scope: Explicit register product ownership replaces shared-directory discovery. Authority and connection outputs retain transitive receipts; cycle history has its own projection manifest over immutable cycle receipts and publisher versions. Finish checks both connection and register generations. End-to-end finish and post-finish manifest verification remain required for each cycle; historical receipts never certify future generations. Current quality cycle adds read-only release validation with exact stored-report and compressed parity, binds current checker/native adapter dependencies to a new generation, and preserves three already curated context return adapters in owning producers. Adversarial checks cover no-write behavior and stale/unsafe states; no new historical source admission is claimed. Publication projections now separate reviewed bibliographic metadata and printed-source notes from audit controls. Raw witnesses remain preserved; quotations, selectors, dates, people, office ranks, geometry and uncertainty retain paired conservation checks. This is a publication-boundary correction, with no new historical source admission.

Sources: Authority data; PIR candidate queue; Source inventories; Fasti aggregates

1. Build reproducible work/language and person/office coverage, including zero-reviewed strata.
2. Reject duplicate IDs, identity bases for other people or works, reversed dates, cyclic sequences and jurisdiction-presence conflation.
3. Detect stale aggregates against authoritative claim files.

Acceptance:

• Every output binds the input generation.
• Negative fixtures demonstrate that unsafe changes fail.
• No visible dashboard added to the main workspace.

implemented-prior-stage · Depends on: quality-foundation

Sources: Tacitus Agricola 7.5; 9.1–7; 18.1; 40.1–3

1. Admit exact-quote identity contexts and distinct career, status and judicial-practice statements.
2. Preserve less-than-three-year duration and before/after relations without fabricated calendar endpoints.
3. Keep the Syrian prospect separate from an actual appointment.

Acceptance:

• Quotes, Unicode selectors and source hashes resolve.
• Unknown regional geometry remains null.
• Pontificate is not pontifex maximus; conventus practice is not an identified assize event.

Greek-Josephus-concordance-implemented; date-adjudication-and-Dio-pending · Depends on: quality-foundation

Completed scope: All Greek reader units map to exact Niese section spans with partition conservation. One reader ordering anomaly retained. Work-owned capture date leads do not admit personal dates; the legacy indexes remain quarantined. English units have separate literal extraction and Whiston-marker checks. The content mismatch under Antiquities reader label 17.13 is explicitly documented; other marker disagreements remain uncollated warnings. Cycle continuation: English reader 17.13 is separately adjudicated against Niese 17.339–341 inside Greek reader 17.332. Chapter heading and paragraph event scope are distinguished; raw shared-label joins remain blocked.

Sources: Josephus grouped first-Niese reader units; Dio reader editions; Inherited chronology sidecars

1. Inventory work, language, edition and locator grammar for each sidecar.
2. Build explicit one-to-many section mappings from checked edition sections; never prefix-join an author collection.
3. Audit every proposed date at sentence/event scope; keep mismatches in quarantine.
4. Collate unresolved English extractions and marker disagreements; distinguish numbering conventions from actual content misalignment before admitting counterpart joins.

Acceptance:

• A mapping identifies both editions and exact spans.
• Every admitted conversion has its own source and semantic role.
• Boundary and retrospective cases pass; original sidecars remain preserved.

selected-official-identity-checks-implemented; full-career-collation-pending · Depends on: quality-foundation

Completed scope: Three Dacian office-person records and three priestly identity candidates checked in official public JSON. These records expose identity metadata, not career fields. Printed entry/career review remains incomplete.

Sources: 2021 official public snapshot; Current official public PIR API; Verified printed PIR editions

1. Select one governor cohort and one priestly cohort from the candidate queue.
2. Fetch bounded public records with dates, response hashes and access/error states. Compare with the frozen snapshot.
3. Verify printed articles and cited witnesses; separate PIR1, PIR2 and addenda. Retain negative/ambiguous identities.

Acceptance:

• Do not describe sampled live checks as a full current database.
• A repeated printed reference never collapses distinct numeric IDs.
• No private editorial comments or restricted articles enter public exports.

Dacia-Porolissensis-selected-attestations-implemented; wider-cohorts-queued · Depends on: pir-refresh

Completed scope: Seven reported diploma attestations for Livius Gratus, Flavius Italicus and Macrinius Vindex, plus Quintilianus’ CIL III 836 building attestation under 157 in Piso (entries 73–76). Explicit source-record identities are reviewed; literal procurator wording and modern presidial/provincial interpretation remain distinct. No anonymous-diploma identity, full tenure, immediate succession, origin or kinship inferred. Volusius’ Gilău diploma (entry 77.2) is one uncertain-date attestation under 161 or 162. Published day readings remain unresolved; a public source-record concordance is distinct from proposed military-tribune and Tingitana identities.

Sources: Thomasson Laterculi and supplements; Dacia II accessible local lead; Pflaum; Demougin; Regional fasti

1. Verify title, contents and year against the physical digital copy; deduplicate scans.
2. Start with a declared province, office type and period. Collate person/office clauses against PIR and cited inscriptions/diplomas.
3. Represent office tenure, attestation date and narrative context separately; retain competing reconstructions.

Acceptance:

• No assumed one-governor-per-year denominator.
• No promotion of procurator or quaestor to governor by keyword.
• Every accepted tenure bound has a cited basis.

partly-implemented · Depends on: quality-foundation

Completed scope: Carlisle Brit. 54.16 and Vindolanda Tab. Vindol. III 581 are distinct documentary objects, each with two dependent reading displays and its own source-local consul mention. The latter adds a 104 CE account-section heading with its corrected opening and missing iteration preserved. Neither document completes the ordinary pair, merges persons or locates a consul. Section-2 dates and its unnamed governor visit remain unadmitted as secure person/calendar statements.

Sources: CIL II²/14.2–4; AE/SEG references in PIR; EDH/EAGLE and edition repositories; P.Yale I 61; RIB Online, Brit. 54.16 (Carlisle stylus tablet); RIB Online, Tab. Vindol. III 581 b.20 (account-section consular heading)

1. Retrieve published records and underlying editions/images when available; record access failures rather than substituting web summaries.
2. Preserve surviving letters, restorations, alternative readings, findspot and monument history.
3. Distinguish one monument in several editions from several independent witnesses. Recheck current papyrus readings.
4. Next: collate one independent published epigraphic or papyrological consular date, or inspect the separate Trajan V heading in Tab. Vindol. III 581 a.9 with its own source and date scope. Same-object additions must retain the existing account dependency group.

Acceptance:

• Document, edition, reading and person identity have separate IDs.
• Findspot never automatically becomes personal presence.
• A bibliographic citation alone is not a verified document.

source-bound-Roman-and-Judaean-batches-implemented; full-college-rosters-still-needed · Depends on: documentary-witnesses

Completed scope: Joazar son of Boethus, Eleazar his brother and Jesus son of Sie are recorded from Antiquities 17.339 and 17.341. The Judaean institution is separated from Roman colleges. Calendar dates and external identities remain unknown. Tacitus Annals 3.58 and 3.71 add Maluginensis’ explicit priestly role, request, argument, reported rule, refusal and allocation elsewhere. The reviewed 22 CE consular frame dates these episodes, not an Augustan enactment or a full tenure. No Asian governorship or external identity is admitted. Two edited flaminica clauses for Calve[ntia] L. f. Marc[ellina] are linked to one AE 1974, 348 inscription and one reviewed PIR source record. The second divine recipient remains restored and queried; after-176 dating applies to the inscription only, with no latest or tenure bound.

Sources: Rüpke annual lists/biographies not in local 37-page extract; Priestly acts and inscriptions; Hispanic municipal/conventus/provincial cult catalogues; Josephus Antiquities 17.339 and 17.341 · distinct Judaean high-priest institution

1. Locate complete authorised roster/biography access. Use the existing introductory extract only within its actual coverage.
2. Separate college membership, named function, co-option, inauguration and last attestation.
3. Compare cult levels and distinguish patron hypotheses reconstructed from calatores.

Acceptance:

• No lifelong membership interval from a single attestation.
• No fixed annual headcount assumption.
• Original and reconstructed affiliations remain distinguishable.

partly-incorporated · Depends on: documentary-witnesses, edition-concordance

Sources: Haensch Appendix IX and regional chapters; Dalla Rosa/Aristides; Lewis/papyri; Loukopoulou; Étienne

1. Retain the 51 undated synthesis rows as context. Inspect relevant body chapters for time-qualified seat changes.
2. Adjudicate place identity and interval separately; treat competing polygons as cited alternatives.
3. Keep judicial practice, individual session, district, resident association, cult institution and editorial corpus area distinct.

Acceptance:

• No capital coordinate substituted for a province, region or unresolved seat.
• Smyrna festival presence remains distinct from a judicial session.
• No movement line from ordered secondary episodes without personal movement evidence.

source-qualified-ordinary-roster-through-120; later-collation-and-suffects-open · Depends on: quality-foundation

Completed scope: Degrassi ordinary headings through 115 CE are source-qualified. The 106–115 batch was inspected in text and page images (printed 32–34); iterations, praenomen forms and source spelling variants are retained. AshLI 179 and 181 provide separately cited short-form comparisons for 114–115, without corroborating full names or personal presence. The continued second line of Manilius’ 114 heading is retained as part of his polyonymous name, without creating a third consul. Five ordinary pairs for 116–120 are now transcribed from pp. 34–35, retaining Rebilus/Rebulus, imperial iterations and both long 120 names. The 116 asterisk remains a source completeness claim, not completeness of the admitted suffect roster. IG X 2, 2, 487 provides a short-form 119 comparison with its numeral display normalization declared; it does not establish full tenure or a consul’s location.

Sources: Degrassi; Gallivan; CLRE; Consular documents and corrected regional fasti

1. Audit existing year ranges and held/disputed years.
2. Transcribe a bounded next chronological segment and independently collate names, iterations and sources.
3. Separate ordinarii, suffects, recognition jurisdictions and post-consular dating.
4. After a governor or priestly batch, inspect the next bounded ordinary segment from 121 CE; keep later nomenclature and suffects separate, and retain the held 13 CE identity/role dispute.

Acceptance:

• No fabricated pair in single/disputed years.
• Every year and office has a source locator and uncertainty.
• Suffect completeness has no guessed annual denominator.

queued · Depends on: quality-foundation

Sources: Greek/Latin readers not yet adapted; DPRR; Non-PIR source people; Women and local institutional personnel

1. Use the coverage matrix to select underrepresented works/languages and roles.
2. Create source-local authority records where external identity is unadjudicated.
3. Label a blind stratified evaluation set, including passages without recognised names, before claiming extraction recall.

Acceptance:

• Namespace boundaries remain explicit.
• Automatic aliases remain candidates.
• Report review coverage and search limits by stratum; no empire-wide completeness claim.

access-needed · Depends on: documentary-witnesses

Sources: PLRE; Connecting Late Antiquities; Late-antique consular and provincial studies

1. Verify access and data reuse terms. Define time and region overlaps with PIR.
2. Collate selected people and offices without forcing PIR identities or assuming identically named people are the same.

Acceptance:

• Cross-authority identification has an explicit decision and evidence.
• Late-antique governance is not forced into an early-imperial taxonomy.

partly-incorporated · Depends on: quality-foundation, agricola-career

Sources: Shared source resolver; Readers; Chronology spine; Fasti maps and exports

1. Add opt-in reader adapters only after locator/source-version checks pass.
2. Keep coordinated text, time, place and relationship views; show evidence roles and unresolved geometry.
3. Use About for methods and scope. Refresh affected hash-bound evidence before any later publication.

Acceptance:

• Desktop/mobile and no-script evidence remain usable.
• Relative sequence does not become plotted exact dates or routes.
• Source replacement invalidates affected joins; export IDs remain stable.

The broad library inventory contains 552 discovery records. The inspected-source ledger and additional access probes overlap that inventory and each other. Selected-page inspection does not imply complete incorporation.

Source stages and duplicate groups · Work and language coverage · Passage review queue

Source-level stage records preserve the prior inspection scope. Filename matches remain leads and do not equate file bytes. Some bibliography candidates have checked passages in specialist fasti but are not matched to the supplied-file assessment ledger; their bibliography notes remain authoritative for that bounded scope.

1. Verify work, edition, completeness, licence/access and duplicate group
2. Record immutable raw input hash and exact printed/PDF or text locator
3. Extract without overwriting raw transcription; preserve editorial signs and raw dates
4. Generate name/office/place candidates with rule version and literal spans
5. Adjudicate identity and source dependence; retain alternatives and rejections
6. State what the source reports; separate proposals, negations and later interpretation
7. Attach role-qualified time, relative sequence, duration and geometry only where supported
8. Run semantic/source gates, review affected browser views and regenerate exports
9. Admit the bounded batch; record outstanding work and revalidate dependants on source change

• The collection date index has no per-work ownership, uses a different locus convention from reader units keyed by first Niese section, and must not be prefix-joined to all four Josephus works.
• The inherited entry does not agree with the current reader’s post-Augustan-death episode. The original indexing and edition concordance have not been revalidated; this is not attributed to an error by the external publisher.
• Only a selected discovery cohort has passage-level identity review; source and social selection biases remain.
• Source review is selective and bibliographic identification remains incomplete. Complete priestly and late-antique rosters remain unincorporated.
• Most provincial seats lack time-qualified validity. Null geometry is intentional; regional and contextual records cannot establish personal movement.
• The Greek unit under reader label 17.13 concerns Herod’s grandchildren. The English unit’s chapter heading announces Archelaus’s banishment, but its body describes earlier administration and priestly appointments. The shared reader label does not align the texts. A separately reviewed paragraph counterpart maps the English body to Niese 17.339–341 inside Greek reader unit 17.332. Raw reader/XML data are preserved; automatic same-label joins remain blocked. Other marker disagreements are uncollated warnings, not asserted content errors.
• Piso prints VII Kalendas Novembres alongside 24 October; the frozen PIR source field reports 26 October 161 (?). The day discrepancy remains unresolved. Only the alternative years 161 or 162 are used for searching. The current official PIR endpoint supplies name/reference metadata, not the career or date field. These publications concern one diploma.

Current quality audit · Structured plan · Relative chronology

1. Correct contradictions, stale aggregates and unsafe joins before expansion
2. Prefer recoverable source passages and currently useful office/jurisdiction cohorts
3. Balance governor, priestly and consular evidence; include women, non-elite and non-PIR people where sources support them
4. Use small, source-bounded batches; record rejection reasons and unsearched strata
5. Verify a stratified sample of automated non-matches and ambiguous matches before widening extraction

No overall accuracy or recall percentage is claimed from the editorial pilot. Before measuring recall, define source sections and manually label positives and negatives independently of the extractor. Report results by work/language, office family and ambiguity class. Duplicate publications and translations share dependency groups.

• BBAW PIR project methods and scope: Source-type distinctions and biographies grounded in non-literary witnesses as well as literary reports. PIR is a selective authority with a period boundary; missing fields do not imply an absent office.
• W3C PROV-DM, sections 2 and 5: Separate source entities, derivation activities and editorial responsibility. Record versions and dependencies. Local receipts are informed by PROV; no blanket conformance claim.
• EpiDoc, Previous Editor: Preserve evidence and responsibility for supplied readings. A previously read or restored letter is not silently treated as visible surviving text.