feat(ar): add Arabic normalization

README.md

@@ -143,6 +143,7 @@ Pipelines are defined declaratively in **YAML presets**. Each preset lists the s
 
 | Code | Language |
 | ---- | -------- |
+| `ar` | Arabic   |
 | `da` | Danish   |
 | `de` | German   |
 | `en` | English  |

docs/reference/steps.md

@@ -4,6 +4,16 @@ Auto-generated — do not edit by hand. Run `scripts/generate_step_docs.py` to u
 
 ## Text Steps
 
+### `apply_character_replacements`
+
+**Base class:** `TextStep`
+
+Replace or delete single characters using ``character_replacements`` from the language config.
+
+Runs first in the pipeline so that script-level variants (letter forms, native
+digits, vowel marks) are unified before protection, numeric, and symbol steps.
+Skipped when the language does not define ``character_replacements``.
+
 ### `apply_sentence_level_replacements`
 
 **Base class:** `TextStep`

normalization/languages/__init__.py

@@ -1,4 +1,5 @@
 from . import (
+    arabic,
     danish,
     dutch,
     english,
@@ -18,6 +19,7 @@
 register_language(LanguageOperators)
 
 __all__ = [
+    "arabic",
     "danish",
     "dutch",
     "english",

normalization/languages/arabic/__init__.py

@@ -0,0 +1,7 @@
+from .operators import ArabicOperators
+from .replacements import ARABIC_REPLACEMENTS
+
+__all__ = [
+    "ArabicOperators",
+    "ARABIC_REPLACEMENTS",
+]