feat(packs): en/he/el pack manifests, orthography, corpus manifests [Batch 3]
This commit is contained in:
parent
a7349e4ea3
commit
ddce4d83e0
9 changed files with 213 additions and 0 deletions
15
packs/el/corpora/manifest.yaml
Normal file
15
packs/el/corpora/manifest.yaml
Normal file
|
|
@ -0,0 +1,15 @@
|
||||||
|
corpus_id: el-tier0
|
||||||
|
status: draft
|
||||||
|
|
||||||
|
sources: []
|
||||||
|
|
||||||
|
license_review: pending
|
||||||
|
hashes: []
|
||||||
|
|
||||||
|
notes: |
|
||||||
|
Populate with vetted Koine Greek source manifests.
|
||||||
|
Priority tier-0 sources: SBL Greek New Testament (SBLGNT) or
|
||||||
|
Nestle-Aland 28th edition morphologically tagged text.
|
||||||
|
Both have verse-level D0 boundary markers.
|
||||||
|
Each source must include: title, edition, license, SHA-256, and
|
||||||
|
segmentation strategy. No LLM-extracted content.
|
||||||
27
packs/el/orthography.yaml
Normal file
27
packs/el/orthography.yaml
Normal file
|
|
@ -0,0 +1,27 @@
|
||||||
|
language: el
|
||||||
|
script: Greek
|
||||||
|
dialect: koine
|
||||||
|
|
||||||
|
normalization:
|
||||||
|
unicode_normal_form: NFC
|
||||||
|
lowercase_for_lookup: true
|
||||||
|
strip_diacritics: configurable
|
||||||
|
# Polytonic diacritics (breathing marks, accents, iota subscript) are
|
||||||
|
# preserved for morphological disambiguation but stripped for lemma lookup
|
||||||
|
polytonic_policy: preserve_for_morphology
|
||||||
|
diacritic_strip_for_lookup: true
|
||||||
|
|
||||||
|
punctuation:
|
||||||
|
strategy: preserve
|
||||||
|
ano_teleia: "\u00b7" # Greek semicolon (middle dot)
|
||||||
|
erotimatiko: ";"
|
||||||
|
|
||||||
|
spacing:
|
||||||
|
strategy: whitespace-delimited
|
||||||
|
|
||||||
|
aspect:
|
||||||
|
# Koine Greek verbal aspect is semantically load-bearing and must be
|
||||||
|
# distinguished from tense. Morphological analysis must surface aspect
|
||||||
|
# independently of temporal reference.
|
||||||
|
explicit_in_features: true
|
||||||
|
aspect_values: ["aorist", "present", "perfect", "imperfect", "pluperfect", "future"]
|
||||||
33
packs/el/pack.toml
Normal file
33
packs/el/pack.toml
Normal file
|
|
@ -0,0 +1,33 @@
|
||||||
|
pack_id = "el"
|
||||||
|
name = "Koine Greek Depth Pack"
|
||||||
|
version = "0.1.0"
|
||||||
|
status = "draft"
|
||||||
|
script = "Greek"
|
||||||
|
default = false
|
||||||
|
|
||||||
|
[contracts]
|
||||||
|
lemma_first = true
|
||||||
|
shared_field_target = true
|
||||||
|
local_readback = true
|
||||||
|
alignment_required = true
|
||||||
|
|
||||||
|
[gates]
|
||||||
|
schema = false
|
||||||
|
lexical = false
|
||||||
|
morphology = false
|
||||||
|
lift = false
|
||||||
|
readback = false
|
||||||
|
determinism = false
|
||||||
|
alignment = false
|
||||||
|
coverage = false
|
||||||
|
|
||||||
|
[notes]
|
||||||
|
role = """
|
||||||
|
Depth language pack 2. Koine Greek was chosen because John articulated the
|
||||||
|
Logos — the creative speech act through which all things were made — in
|
||||||
|
Koine Greek (John 1:1-2), which must have been a nod from the Holy Spirit.
|
||||||
|
Koine Greek carries semantic depth in its verbal aspect system (aorist,
|
||||||
|
perfect, present) that is distinct from tense and must be represented
|
||||||
|
natively. The aorist of egeneto (\u03b5\u03b3\u03ad\u03bd\u03b5\u03c4\u03bf) — 'it came into being' — is not
|
||||||
|
the same field state as the present of eimi (\u03b5\u03b9\u03bc\u03b9) — 'I am'.
|
||||||
|
"""
|
||||||
13
packs/en/corpora/manifest.yaml
Normal file
13
packs/en/corpora/manifest.yaml
Normal file
|
|
@ -0,0 +1,13 @@
|
||||||
|
corpus_id: en-tier0
|
||||||
|
status: draft
|
||||||
|
|
||||||
|
sources: []
|
||||||
|
|
||||||
|
license_review: pending
|
||||||
|
hashes: []
|
||||||
|
|
||||||
|
notes: |
|
||||||
|
Populate with vetted, deterministically segmentable source manifests.
|
||||||
|
Each source must include: title, origin, license, SHA-256 of raw file,
|
||||||
|
and structural segmentation strategy (e.g. paragraph-boundary, sentence-boundary).
|
||||||
|
No LLM-extracted content. Segmentation must be D0 or D1.
|
||||||
21
packs/en/orthography.yaml
Normal file
21
packs/en/orthography.yaml
Normal file
|
|
@ -0,0 +1,21 @@
|
||||||
|
language: en
|
||||||
|
script: Latin
|
||||||
|
|
||||||
|
normalization:
|
||||||
|
unicode_normal_form: NFC
|
||||||
|
lowercase_for_lookup: true
|
||||||
|
strip_diacritics: false
|
||||||
|
|
||||||
|
punctuation:
|
||||||
|
strategy: preserve
|
||||||
|
sentence_terminals: [".", "!", "?"]
|
||||||
|
quotation_marks: ["\u201c", "\u201d", "\u2018", "\u2019"]
|
||||||
|
|
||||||
|
spacing:
|
||||||
|
strategy: whitespace-delimited
|
||||||
|
collapse_multiple_spaces: true
|
||||||
|
|
||||||
|
special_forms:
|
||||||
|
contractions: expand
|
||||||
|
# "don't" -> "do not" before lemmatization
|
||||||
|
# "it's" -> "it is" before lemmatization
|
||||||
30
packs/en/pack.toml
Normal file
30
packs/en/pack.toml
Normal file
|
|
@ -0,0 +1,30 @@
|
||||||
|
pack_id = "en"
|
||||||
|
name = "English Base Pack"
|
||||||
|
version = "0.1.0"
|
||||||
|
status = "draft"
|
||||||
|
script = "Latin"
|
||||||
|
default = true
|
||||||
|
|
||||||
|
[contracts]
|
||||||
|
lemma_first = true
|
||||||
|
shared_field_target = true
|
||||||
|
local_readback = true
|
||||||
|
alignment_required = true
|
||||||
|
|
||||||
|
[gates]
|
||||||
|
schema = false
|
||||||
|
lexical = false
|
||||||
|
morphology = false
|
||||||
|
lift = false
|
||||||
|
readback = false
|
||||||
|
determinism = false
|
||||||
|
alignment = false
|
||||||
|
coverage = false
|
||||||
|
|
||||||
|
[notes]
|
||||||
|
role = """
|
||||||
|
Default articulation base. English is chosen as the default for practical
|
||||||
|
reasons (current primary working language) but the contract does not privilege
|
||||||
|
it semantically. Any language could serve this role in a custom CORE instance.
|
||||||
|
The default flag controls surface output, not field-layer priority.
|
||||||
|
"""
|
||||||
15
packs/he/corpora/manifest.yaml
Normal file
15
packs/he/corpora/manifest.yaml
Normal file
|
|
@ -0,0 +1,15 @@
|
||||||
|
corpus_id: he-tier0
|
||||||
|
status: draft
|
||||||
|
|
||||||
|
sources: []
|
||||||
|
|
||||||
|
license_review: pending
|
||||||
|
hashes: []
|
||||||
|
|
||||||
|
notes: |
|
||||||
|
Populate with vetted Hebrew source manifests.
|
||||||
|
Priority tier-0 sources: Biblia Hebraica Stuttgartensia (BHS) or
|
||||||
|
Westminster Leningrad Codex (WLC) — both have well-established structural
|
||||||
|
segmentation (book/chapter/verse = D0 boundary markers).
|
||||||
|
Each source must include: title, edition, license, SHA-256, and
|
||||||
|
segmentation strategy. No LLM-extracted content.
|
||||||
27
packs/he/orthography.yaml
Normal file
27
packs/he/orthography.yaml
Normal file
|
|
@ -0,0 +1,27 @@
|
||||||
|
language: he
|
||||||
|
script: Hebrew
|
||||||
|
direction: rtl
|
||||||
|
|
||||||
|
normalization:
|
||||||
|
unicode_normal_form: NFC
|
||||||
|
lowercase_for_lookup: false # Hebrew has no case
|
||||||
|
strip_diacritics: configurable
|
||||||
|
# Nikkud (vowel pointing) is stripped by default for lemma lookup
|
||||||
|
# but preserved when semantically load-bearing (e.g. Qere/Ketiv distinction)
|
||||||
|
nikkud_policy: strip_for_lookup
|
||||||
|
cantillation_marks: strip
|
||||||
|
|
||||||
|
punctuation:
|
||||||
|
strategy: preserve
|
||||||
|
paragraph_marker: "\u05c0" # paseq
|
||||||
|
section_marker: "\u05c3" # sof pasuq
|
||||||
|
|
||||||
|
spacing:
|
||||||
|
strategy: whitespace-delimited
|
||||||
|
maqaf_handling: preserve
|
||||||
|
# maqaf (\u05be) connects words syntactically; preserve as a unit marker
|
||||||
|
|
||||||
|
shoresh:
|
||||||
|
# Root-based morphology. Lemma lookup resolves to 3- or 4-letter root.
|
||||||
|
# Morphological analysis must identify shoresh before feature extraction.
|
||||||
|
extraction_required: true
|
||||||
32
packs/he/pack.toml
Normal file
32
packs/he/pack.toml
Normal file
|
|
@ -0,0 +1,32 @@
|
||||||
|
pack_id = "he"
|
||||||
|
name = "Hebrew Depth Pack"
|
||||||
|
version = "0.1.0"
|
||||||
|
status = "draft"
|
||||||
|
script = "Hebrew"
|
||||||
|
default = false
|
||||||
|
|
||||||
|
[contracts]
|
||||||
|
lemma_first = true
|
||||||
|
shared_field_target = true
|
||||||
|
local_readback = true
|
||||||
|
alignment_required = true
|
||||||
|
|
||||||
|
[gates]
|
||||||
|
schema = false
|
||||||
|
lexical = false
|
||||||
|
morphology = false
|
||||||
|
lift = false
|
||||||
|
readback = false
|
||||||
|
determinism = false
|
||||||
|
alignment = false
|
||||||
|
coverage = false
|
||||||
|
|
||||||
|
[notes]
|
||||||
|
role = """
|
||||||
|
Depth language pack 1. Hebrew was chosen because of the range of depth it
|
||||||
|
brings to the vocabulary manifold: root-based morphology (shoresh), semantic
|
||||||
|
families that carry multiple layers of meaning simultaneously, and the
|
||||||
|
language in which the universe was spoken into existence (Genesis 1).
|
||||||
|
Hebrew morphological stems (Qal, Niphal, Piel, Pual, Hiphil, Hophal,
|
||||||
|
Hithpael) are semantically load-bearing and must be represented natively.
|
||||||
|
"""
|
||||||
Loading…
Reference in a new issue