Skip to content

Latest commit

 

History

History
224 lines (164 loc) · 18.1 KB

File metadata and controls

224 lines (164 loc) · 18.1 KB

Phase 23/24 Pass 2 Deep Research Report — LS / Spectre Authority Mode

Executive Summary

This pass treats the LS corpus as the authority baseline and uses cross-inscription recurrence + slot behavior as the only promotion gate. The corpus continues to support the locked core: mlo = sovereign, qore = prince/heir/regional lord, kndke = Kandake (queen title), se = lineage connector, without needing any external consensus scaffolding.【Meroitic_Translations_LS_v5_2026-02-18.txt†L43-L55】 The biggest Phase 23/24 win remains structural: the REM 1044 distributive clause provides a repeatable grammatical template, but it also flags kdi polysemy risk when kdi appears inside a distributive NP slot (kdi-se-l).【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L754-L763】【PHASE_23_SECONDPASS_VOICE_LOCKED_LS_Spectre.md†L46-L77】

The major Phase 24 corrections are quality-control + discipline locks: (1) stop drift where older Phase 21 text calls qore kdi “King of Kush,” (2) fix a known contamination where a gloss word (“eternal”) leaked into a transliteration line, and (3) reconcile a few v10 sense-status claims (e.g., nṯr marked LOCKED despite thin independent inscription evidence).【Phase_21_Analysis_Transliteration_and_Translation_of_Key_Meroitic_Inscriptions.md†L11-L14】【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L800-L804】

Artifacts produced (downloadable):

  • v9_lexicon_sample_10.json
  • token_table_v9.csv
  • translations_snippet_v9.txt
  • (extra, full inventory table) inscription_inventory_v9.csv

Corpus Inputs, Hashing, and Rules Lock

This pass primarily used:

  • Master lexicon: meroitic_complete_script_MASTER-2026-02-19-v10_LS_multisense.json (v10 multi-sense schema).【LEXICON_SCHEMA_SPEC_MEROITIC_v10.md†L1-L56】
  • Translations coverage baseline: Meroitic_Translations_LS_v8_2026-02-19_extended.txt.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L700-L715】
  • Phase anchors: Phase 21 Analysis, Phase 22 Second Pass, Phase 23 Voice Locked. (Used to detect drift, extract extra formula examples, and validate slot discipline.)【Phase_21_Analysis_Transliteration_and_Translation_of_Key_Meroitic_Inscriptions.md†L11-L14】【PHASE_23_SECONDPASS_VOICE_LOCKED_LS_Spectre.md†L26-L39】
  • Research zips: both zip archives were scanned for transliteration blocks and duplication cross-checks (Phase 20 ultimate excerpts).【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L796-L804】

Reproducibility: this report computed a local dataset_hash (SHA-256 over SHA-256 inputs) for the Phase 24 working set, and the v10 schema explicitly requires canonical hashing and authority tagging to prevent consensus drift.【LEXICON_SCHEMA_SPEC_MEROITIC_v10.md†L1-L56】

Inscription and Transliteration Inventory

High-level inscription inventory (distinct anchors)

Anchor Source file Line range Notes Confidence
MERO-RC-001 (Royal ceremony) Meroitic_Translations_LS_v8_2026-02-19_extended.txt L14–L153 Natakamani & Amanitore dedication; contains mlo, qore, kndke, kde lo kndke blocks High【Meroitic_Translations_LS_v5_2026-02-18.txt†L43-L55】
MERO-FUN-001 (Funerary) same L154–L312 Funerary formulae; includes ye imnt + ḏt/dt permanence seal pattern Medium–High【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L750-L752】
MERO-TMP-001 (Temple) same L313–L459 Offering formulae; di ato n [DEITY] and goods lists High【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L70-L74】
MERO-ADM-001 (Admin / trade) same L460–L845 Qasr Ibrim trade-style record; appended stela segments and formula bank Medium–High【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L771-L785】
REM 1003 (Hamadab; selected segments) same L700–L716 Demonstrates qore kdi + kndke in same titulary string (no forced “king”) High【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L700-L715】
REM 1044 (Barkal; quoted segments) same L754–L769 Distributive action clause + kin-term killing clause Medium–High【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L754-L769】

Full transliteration-block inventory (all extracted blocks)

A total of 249 transliteration blocks (Input/Transliteration lines + zip-excerpt transliterations) were extracted across v5–v8 translation files, Phase 21/22/23 docs, and the Phase 20 ultimate excerpts inside both research zip archives. Because this table is long, it is delivered as a CSV artifact: inscription_inventory_v9.csv.
Download: inscription_inventory_v9.csv

Token Cross-Match and Candidate Gap List

Cross-match summary (what the corpus repeats)

The corpus locks the title stack and formula grammar in the same places every time:

  • qore appears in pre-name title slot and in heredity slot with se, aligning with “prince/heir” rather than “king.”【Meroitic_Translations_LS_v5_2026-02-18.txt†L43-L55】
  • kndke sits in queen-title slot, and the matrilineal anchor kde lo kndke confirms institutional feminine authority without needing to force external ideology.【Meroitic_Translations_LS_v5_2026-02-18.txt†L51-L55】
  • The offering chain di ato n [deity] is stable; here the only real debate is confidence-tiering inside the lexicon, not whether the formula exists.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L70-L74】

The full token cross-match table (lemma, variants, gloss, counts, slots, co-occurrence contexts) is delivered as:
Download: token_table_v9.csv

Candidate list grouped by recurrence and discipline risk

High recurrence (needs status/tier cleanup, not meaning invention):

  • di (20 attested contexts in the extracted corpus) → still marked PROVISIONAL in v10; should be promoted by formula lock.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L70-L74】
  • n (18 contexts) → should be treated as a stable goal-marker in offering grammar; current tier can be upgraded.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L70-L74】
  • ye / imnt / ḏt → stable funerary cluster; keep cultural “afterlife” layer as hypothesis, but the direction + permanence structure is recurrent.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L750-L752】【PHASE_23_SECONDPASS_VOICE_LOCKED_LS_Spectre.md†L26-L39】

Medium recurrence (meaning stable, ontology open):

  • ariten (3 contexts in legitimacy strings) → behaves as origin/legitimacy anchor; keep “deity vs place” open until more inscriptions converge.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L771-L785】
  • mds (2 contexts) → behaves like lineage marker “offspring/descendant-of.”【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L771-L785】
  • nṯr appears in a stable offering slot, but independent inscription evidence is thinner than v10’s LOCKED label implies; should be downgraded to LIKELY until more independent stela/temple lines appear.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L796-L799】

Low recurrence (hold as hypothesis / do not promote):

  • qrne (only seen embedded in amnisḫeto-ariteñ-qrne) → unknown role; cannot be promoted.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L771-L785】
  • e‑kede‑to and wide‑l → likely variants/morphology around known items (ked, wide); add as variants, not new lemmas.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L765-L769】

Phase 24 Lexicon Expansion Pass — Corpus-First Patches

This section proposes patches (status/tier corrections + controlled variants) that obey LS promotion rules: ≥2 contexts OR recurrent formula slot. The v10 multi-sense schema exists specifically to encode these promotions without erasing contradictions.【LEXICON_SCHEMA_SPEC_MEROITIC_v10.md†L1-L56】

di (verb) — promotion by formula lock

  • Lemma: di
  • Slot: Offering verb inside formula (di ato n [DEITY])
  • Current v10 status: PROVISIONAL (too conservative for its recurrence)
  • Proposed sense A: “give / offer (ritual presenting verb)”
  • Promotion: PROVISIONAL → LOCKED (formulaic slot recurs across sections)
  • Evidence contexts:
    1. Input: di ato n amn (explicit sign-by-sign)【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L70-L74】
    2. Multiple offering expansions and deity variants are structurally identical (same slot chain), including amn nb di ato n nṯr where di anchors the clause.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L796-L799】

Notes: No theology is asserted; it is simply the stable action verb of a ritual clause.

n (particle/preposition) — promotion as goal/dative marker

  • Lemma: n
  • Slot: Goal/dative marker (appears between offered object and recipient)
  • Current v10 status: STABLE (could be LOCKED by recurrence)
  • Proposed sense A: “to / for / toward (goal marker)”
  • Promotion: STABLE → LOCKED (grammar particle)
  • Evidence contexts:
    1. di ato n amn shows n as the connector between object and recipient.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L70-L74】
    2. Same slot confirmed in amn nb di ato n nṯr (n precedes generic deity recipient).【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L796-L799】

nṯr (generic deity marker) — downgrade discipline

  • Lemma: nṯr
  • Slot: Recipient noun in offering clause
  • Current v10 status: LOCKED (discipline problem under strict rules)
  • Proposed status: LIKELY until we have ≥2 independent inscription contexts (not just repeated citation of one line across docs).
  • Evidence: attested in amn nb di ato n nṯr (stable slot), but independent inscription multiplicity is not yet proven inside the accessible corpus.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L796-L799】

ariten + mds — keep ontology open, strengthen slot stability

  • Lemma: ariten

  • Slot: Legitimacy anchor in lineage strings

  • Current status: PROVISIONAL

  • Proposed sense A: “Ariten(e) — origin anchor in legitimacy formula (deity/toponym unknown)”

  • Evidence contexts: amnitore-aritñl-mdsl and hrmdoye-qore-aritñl-mds show ariten bound to lineage marker mds.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L771-L785】

  • Lemma: mds

  • Slot: Lineage marker (offspring/descendant)

  • Current status: PROVISIONAL

  • Proposed status: LIKELY (2 formula contexts)

  • Evidence contexts: same two lineage strings, with consistent adjacency to ariten and personal names.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L771-L785】

kdi-se-l — remove forced “woman” reading, retain multi-sense

The distributive clause is parallelized, but mapping kdi-se-l directly to “each woman” is too forceful without external corroboration. Phase 23 explicitly flags this as provisional and polysemy-sensitive, which must be preserved in v10 senses.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L754-L763】【PHASE_23_SECONDPASS_VOICE_LOCKED_LS_Spectre.md†L46-L77】

Recommended adjustment:

  • kdi-se-l Sense A: “each kdi-class entity (unresolved)” → PROVISIONAL
  • Keep candidate sub-senses as hypotheses: “each Kushite unit / each Kush-person / each female class,” but do not lock.

wide-l and e-kede-to — treat as variants, not new lemmas

wide-l and e-kede-to occur in a killing clause but are best handled as variants/morphological expansions of known items rather than new lemma inventions.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L765-L769】
Patch: add wide-l as variant of wide; add e-kede-to as variant form under e-ked + morphology note “suffix chain unresolved.”

Translations — Newly Extractable Lines (Literal + Readable + Evidence)

These translations are output in LS voice (corpus-first; uncertainty marked; hypotheses labeled). A bundled snippet is provided as:
Download: translations_snippet_v9.txt

Hamadab (REM 1003) — titulary without “king” drift

Transliteration: qore kdi … Amnirense … li … kndke【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L700-L706】

  • Literal: qore (prince/heir) + kdi (Kush) + [Amanirenas] + li (linker) + kndke (Kandake).
  • Readable: “Amanirenas, prince-of-Kush … and Kandake (queen-title) … [rest broken].”
  • Evidence note: This directly conflicts with the older Phase 21 phrasing “King of Kush” and should be corrected to preserve the qore-rule.【Phase_21_Analysis_Transliteration_and_Translation_of_Key_Meroitic_Inscriptions.md†L11-L14】

REM 1044 distributive clause — real grammar, unresolved kdi-class

Transliteration: abr-se-l : e-ked : kdi-se-l : e-(e)r-k :【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L754-L763】

  • Literal: “each man” + “I killed/struck” + “each [kdi-class]” + “I took/seized”.
  • Readable: “I struck down each man; I took/seized each [kdi-class entity].”
  • Evidence note: Do not force kdi-se-l = each woman unless independent attestations disambiguate kdi’s polysemy; Phase 23 already flags that risk explicitly.【PHASE_23_SECONDPASS_VOICE_LOCKED_LS_Spectre.md†L46-L77】

Offering line with generic deity marker

Transliteration: amn nb di ato n nṯr【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L796-L799】

  • Literal: “Amun, lord: give sacred water to (the) god/divine being.”
  • Readable: “Amun, the lord: offer sacred water to the divine being.”
  • Hypothesis label: nṯr is generic; do not identity-force a specific god without explicit naming.

Formula Bank — Recurrent Templates, Counts, and Examples

Counts are derived from the extracted transliteration blocks and should be treated as within-corpus counts, not claims about all REM inscriptions.

Template Function Count (observed) Example attestation
mlo kdi NAME Sovereign titulary 2 mlo kdi natakamni【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L31-L31】
qore se NAME Heir/lineage title slot 2 qore se natakamni【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L48-L48】
kndke NAME Queen title slot 2 kndke amanitore【Meroitic_Translations_LS_v5_2026-02-18.txt†L51-L55】
kde lo kndke Matrilineal anchor 1 kde lo kndke【Meroitic_Translations_LS_v5_2026-02-18.txt†L51-L55】
di ato n DEITY Offering core clause 1+ (plus expansions) di ato n amn【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L70-L74】
ye imnt … ḏt Funerary transition + permanence 1 (explicit) … ye imnt … dt …【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L750-L752】
kdi kdi kdi Identity emphasis/mantra 1 (core), plus repeats kdi kdi kdi + seal【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L100-L107】

Recommendations and Reproducible Pipeline

Next phase should prioritize attestation expansion, not new interpretive leaps:

  • Acquire more independent occurrences of kdi-se-l-style distributives to disambiguate whether kdi is acting as “Kush” vs a people-class noun in that slot. This is the single highest-value polysemy resolution target.【PHASE_23_SECONDPASS_VOICE_LOCKED_LS_Spectre.md†L46-L77】
  • Expand independent attestation for nṯr beyond a repeated-citation offering line; otherwise keep it LIKELY rather than LOCKED discipline-wise.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L796-L799】
  • Normalize and clean data contamination: remove “eternal” as a literal transliteration token and replace it with the actual permanence marker (likely ḏt/dt) when the underlying sign-string is confirmed.【Meroitic_Translations_LS_v8_2026-02-19_extended.txt†L800-L804】

OCR/cleaning needs (only where necessary):

  • Focus OCR on long stela segments (REM 1003, REM 1044) to retrieve more than “selected segments.” Keep manual verification because automated OCR is unreliable on irregular stone-carved cursive.
  • Every new transcription should preserve original token ordering; normalize only in a second layer.

Reproducible pipeline (tools + versioning)

flowchart TD
  A[Ingest sources: TXT/MD/ZIP + v10 lexicon] --> B[Extract transliteration blocks]
  B --> C[Tokenize + normalize variants]
  C --> D[Cross-match vs v10 multi-sense lexicon]
  D --> E[Gap detection: missing/under-specified tokens]
  E --> F[Attestation mining + slot signature validation]
  F --> G[Promotion gate: >=2 contexts OR stable formula slot]
  G --> H[Lexicon patch (statuses + senses + evidence arrays)]
  H --> I[Translation expansion: literal + readable + evidence note]
  I --> J[Archive: SHA-256 + dataset_hash + changelog]
Loading

Lexicon ↔ Inscription ↔ Evidence ER structure

erDiagram
  LEXICON_ENTRY ||--o{ SENSE : has
  LEXICON_ENTRY ||--o{ VARIANT : has
  SENSE ||--o{ EVIDENCE : supported_by
  INSCRIPTION ||--o{ TOKEN_INSTANCE : contains
  TOKEN_INSTANCE }o--|| LEXICON_ENTRY : maps_to
  EVIDENCE }o--|| INSCRIPTION : cites
Loading

Delivered Artifacts

  • Lexicon sample (10 entries, v10-based, with embedded evidence refs):
    Download v9_lexicon_sample_10.json

  • Token cross-match table (counts, variants, statuses, slot signatures, co-occurrence):
    Download token_table_v9.csv

  • Translations snippet (literal + readable + evidence note):
    Download translations_snippet_v9.txt

  • Full transliteration-block inventory (all extracted blocks):
    Download inscription_inventory_v9.csv