language: he
script: Hebrew

normalization:
  unicode_normal_form: NFC
  # Hebrew Unicode: use precomposed forms where available.
  # Full nikud (vowel points) are preserved in normalization.
  # Qere/Ketiv variants must be tracked separately in the lemma record.
  strip_nikud_for_indexing: false
  # Nikud is semantically relevant for disambiguation in many cases.
  # Do not strip it during normalization. Lemma lookup may index
  # both pointed and unpointed forms.
  strip_teamim: true
  # Cantillation marks (teamim) are stripped during normalization.
  # They carry liturgical meaning but are not part of semantic lift.
  final_forms: preserve
  # Final letter forms (ךםןףץ) are preserved exactly.

punctuation:
  strategy: preserve
  maqaf: preserve
  # The maqaf (־) is a Hebrew hyphen connecting words into
  # a single accentual unit. It is preserved and handled by frame rules.

spacing:
  strategy: whitespace-split

script_direction: rtl