Tamil visual normalization rules added for flipped two-part vowel sig…

…ns. Example: SIGN EE + SIGN AA -> SIGN OO PiperOrigin-RevId: 641977604
google-research · Jun 10, 2024 · 21a5c87 · 21a5c87
1 parent 9300462
commit 21a5c87
Show file tree

Hide file tree

Showing 2 changed files with 19 additions and 0 deletions.
diff --git a/nisaba/scripts/brahmic/data/Taml/visual_rewrite.textproto b/nisaba/scripts/brahmic/data/Taml/visual_rewrite.textproto
@@ -106,3 +106,18 @@ item {
   uname: ["SIGN AU", "SIGN I"] raw: "ௌி"
   to_uname: ["SIGN E", "LLA", "SIGN I"] to_raw: "ெளி"
 }
+
+# Flipped two-part vowel signs.
+# The non-flipped sequence is covered by NFC.
+item {
+  uname: ["SIGN AA", "SIGN E"] raw: "ாெ"
+  to_uname: ["SIGN O"] to_raw: "ொ"
+}
+item {
+  uname: ["SIGN AA", "SIGN EE"] raw: "ாே"
+  to_uname: ["SIGN OO"] to_raw: "ோ"
+}
+item {
+  uname: ["AU LENGTH MARK", "SIGN E"] raw: "ௗெ"
+  to_uname: ["SIGN AU"] to_raw: "ௌ"
+}
diff --git a/nisaba/scripts/brahmic/testdata/visual_norm.textproto b/nisaba/scripts/brahmic/testdata/visual_norm.textproto
@@ -44,6 +44,10 @@ rewrite { rule: "SINH" input: "අපේ‍්‍රල්" output: "අප්
 # rewrite { rule: "TAML" input: "தமிழர்‌கள்‍" output: "தமிழர்கள்" }
 rewrite { rule: "TAML" input: "ஆக்‌ஷன்" output: "ஆக்‌ஷன்" }
 
+rewrite { rule: "TAML" input: "காெள்" output: "கொள்" }
+rewrite { rule: "TAML" input: "ப்ராேஷன்" output: "ப்ரோஷன்" }
+rewrite { rule: "TAML" input: "சௗெந்தர்யம்" output: "சௌந்தர்யம்" }
+
 rewrite { rule: "DEVA" input: "श्रीमान्‌को" output: "श्रीमान्‌को" }
 rewrite { rule: "DEVA" input: "गोल्‍डबर्ग" output: "गोल्डबर्ग" }