[WIP] kab-Latn: add missing

epitran/data/map/kab-Latn.csv

@@ -1,9 +1,37 @@
 Orth,Phon
+čč,t͡ʃː
+ǧǧ,d͡ʒː
+qq,qː
+jj,ʒː
+cc,ʃː
+hh,hː
+ww,wː
+yy,jː
+ḥḥ,ħː
+bb,bː
+dd,dː
+ff,fː
+gg,ɡː
+kk,kː
+ll,lː
+mm,mː
+nn,nː
+rr,rː
+ss,sː
+tt,tː
+xx,χː
+zz,zː
+ɣɣ,ʁː
+ṣṣ,sˤː
+ṭṭ,tˤː
+ḍḍ,dˤː
+ẓẓ,zˤː
+ṛṛ,rˤː
+θ,θ
 a,a
 i,i
 u,u
 e,ə
-o,o
 y,j
 w,w
 b,b
@@ -19,10 +47,9 @@ q,q
 r,r
 s,s
 t,t
-v,v
 x,χ
 z,z
-ɣ,ɣ
+ɣ,ʁ
 ḥ,ħ
 ṣ,sˤ
 ṭ,tˤ
@@ -33,22 +60,4 @@ z,z
 ǧ,d͡ʒ
 j,ʒ
 c,ʃ
-gh,ɣ
-ch,ʃ
-dj,d͡ʒ
-tch,t͡ʃ
-ts,t͡s
-dz,d͡z
-bb,bː
-dd,dː
-ff,fː
-gg,ɡː
-kk,kː
-ll,lː
-mm,mː
-nn,nː
-rr,rː
-ss,sː
-zz,zː
-xx,χː
-ɣɣ,ɣː
+v,v

epitran/data/post/kab-Latn.txt

@@ -0,0 +1,24 @@
+% Kabyle (kab-Latn) postprocessor for Epitran
+% Author: Athmane MOKRAOUI
+%
+% This postprocessor cleans up phonological output after mapping.
+% 
+% Current version is minimal - most Kabyle phonology is handled
+% in the preprocessor. Future versions may add:
+% - Schwa deletion in specific cluster environments
+% - Emphatic spread effects on vowels
+% - Vowel reduction rules
+
+% ============================================================
+% 1. SCHWA CLEANUP (optional)
+% ============================================================
+% Delete schwa between consonants in certain environments
+% This is a conservative rule - uncomment if needed
+% ə -> 0 / (b|d|f|g|ɣ|h|j|k|l|m|n|q|r|s|t|w|x|z|ʃ|ʒ|χ|ħ|θ) _ (b|d|f|g|ɣ|h|j|k|l|m|n|q|r|s|t|w|x|z|ʃ|ʒ|χ|ħ|θ)
+
+% ============================================================
+% 2. VOWEL LENGTH REMOVAL
+% ============================================================
+% Kabyle does not have phonemic vowel length.
+% If length markers appear (from geminate mapping), they are
+% consonantal length, not vocalic. No action needed here.

epitran/data/pre/kab-Latn.txt

@@ -0,0 +1,30 @@
+% Kabyle (kab-Latn) preprocessor for Epitran
+% Author: ButterflyOfFire
+
+% ============================================================
+% 0. RECOMPOSITION (fix Unicode decomposition issue)
+% ============================================================
+% Epitran decomposes ṭ to t+̣ internally. Recompose before other rules.
+ṭ -> ṭ / _
+
+% ============================================================
+% 1. CONSONANT ASSIMILATION & ADAPTATIONS
+% ============================================================
+n -> m / _ (b|m|p)
+p -> b / _
+
+% ============================================================
+% 2. GEMINATE SHIELDING
+% ============================================================
+ṭṭ -> W / _
+
+% ============================================================
+% 3. SPIRANTIZATION
+% ============================================================
+t -> θ / _ (a|e|i|u)
+t -> θ / (a|e|i|u) _
+
+% ============================================================
+% 4. UNSHIELD
+% ============================================================
+W -> ṭṭ / _

epitran/test/test_kabyle.py

@@ -7,11 +7,12 @@ def setUp(self):
 
     def test_basic_words(self):
         for i, o in [
-            ("taqcict", "taqʃiʃt"),
-            ("axxam", "aχːam"),
-            ("Nniɣ", "nːiɣ"),
-            ("abrid", "abrid"),
-            ("tuwiḍ", "tuwidˤ"),
+            ("taqcict", "θæqʃiʃθ"),
+            ("tamellalt", "θæməlːælt"),
+            ("axxam", "æχːæm"),
+            ("Nniɣ", "nːiʁ"),
+            ("abrid", "æβrid"),
+            ("tewwiḍ", "θəwːidˤ"),
         ]:
             tr = self.epi.transliterate(i)
             self.assertEqual(tr, o)