core/packs/he/orthography.yaml

27 lines
827 B
YAML

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