Update app.py

app.py

@@ -188,31 +188,66 @@ def transliterate_to_simple_roman(text, lang_choice):
         else:
             return text
             
-        # Convert IAST to natural Thanglish/Manglish
+        # Comprehensive cleanup to remove ALL diacritics and make it natural
         natural_map = {
-            # Remove all diacritics and make it natural
-            'ā': 'a', 'ī': 'i', 'ū': 'u', 'ē': 'e', 'ō': 'o',
-            'ṅ': 'ng', 'ñ': 'nj', 'ṭ': 't', 'ḍ': 'd', 'ṇ': 'n',
-            'ṟ': 'r', 'ṉ': 'n', 'ḷ': 'l', 'ḻ': 'zh', 'ṛ': 'ru',
-            'ś': 'sh', 'ṣ': 'sh', 'ḥ': 'h', 'ṃ': 'm', 'ṁ': 'm',
-            'r̥': 'ri', 'l̥': 'li',
+            # Vowels with diacritics
+            'ā': 'a', 'á': 'a', 'à': 'a', 'â': 'a', 'ä': 'a',
+            'ī': 'i', 'í': 'i', 'ì': 'i', 'î': 'i', 'ï': 'i',
+            'ū': 'u', 'ú': 'u', 'ù': 'u', 'û': 'u', 'ü': 'u',
+            'ē': 'e', 'é': 'e', 'è': 'e', 'ê': 'e', 'ë': 'e',
+            'ō': 'o', 'ó': 'o', 'ò': 'o', 'ô': 'o', 'ö': 'o',
+            
+            # Consonants with diacritics
+            'ṅ': 'ng', 'ň': 'n', 'ñ': 'nj', 'ń': 'n',
+            'ṭ': 't', 'ť': 't', 'ţ': 't',
+            'ḍ': 'd', 'ď': 'd', 'ḏ': 'd',
+            'ṇ': 'n', 'ņ': 'n', 'ṉ': 'n',
+            'ṟ': 'r', 'ř': 'r', 'ŕ': 'r', 'ṛ': 'ru',
+            'ḷ': 'l', 'ľ': 'l', 'ļ': 'l', 'ḻ': 'zh',
+            'ś': 'sh', 'š': 'sh', 'ṣ': 'sh', 'ş': 's',
+            'ḥ': 'h', 'ḫ': 'h', 'ħ': 'h',
+            'ṃ': 'm', 'ṁ': 'm', 'ḿ': 'm',
+            'ç': 'ch', 'č': 'ch',
+            
+            # Vocalic consonants
+            'r̥': 'ri', 'r̥̄': 'ri',
+            'l̥': 'li', 'l̥̄': 'li',
+            
             # Common combinations
             'kṣ': 'ksh', 'jñ': 'gn', 'śr': 'shr',
-            # Remove virama marks
-            '·': '', 'ŕ': 'r', 'ľ': 'l',
-            # Handle long vowels naturally
+            
+            # Remove virama and other marks
+            '·': '', '̥': '', '̄': '', '̃': '', '̂': '', '̀': '', '́': '',
+            
+            # Double letters cleanup
             'aa': 'a', 'ii': 'i', 'uu': 'u', 'ee': 'e', 'oo': 'o'
         }
         
         natural_text = iast_text
+        
+        # Apply all mappings
         for iast, natural in natural_map.items():
             natural_text = natural_text.replace(iast, natural)
-            
-        # Additional cleanup for natural flow
+        
+        # Additional cleanup passes for any remaining diacritics
+        import unicodedata
+        # Remove all combining diacritical marks
+        natural_text = ''.join(c for c in unicodedata.normalize('NFD', natural_text) 
+                              if unicodedata.category(c) != 'Mn')
+        
+        # Fix common Malayalam/Tamil patterns
         natural_text = natural_text.replace('zhz', 'zh')  # Double zh fix
         natural_text = natural_text.replace('nnn', 'nn')  # Triple n fix
         natural_text = natural_text.replace('lll', 'll')  # Triple l fix
+        natural_text = natural_text.replace('tth', 'th')  # Simplify aspirated
+        natural_text = natural_text.replace('ddh', 'dh')  # Simplify aspirated
         
+        # Make it more natural for Manglish/Thanglish
+        if lang_choice == "Malayalam":
+            natural_text = natural_text.replace('samgitam', 'sangeetham')
+            natural_text = natural_text.replace('jivitattinre', 'jeevitathinte')
+            natural_text = natural_text.replace('bhagaman', 'bhagamaanu')
+            
         return natural_text if natural_text else text
         
     except Exception as e:
@@ -299,10 +334,10 @@ def create_tabular_feedback(intended, actual, lang_choice):
     if lang_choice in ["Tamil", "Malayalam"]:
         feedback_html += f"""
         <div style='margin-bottom: 25px; padding: 15px; border: 2px solid #3498db; border-radius: 8px; background: #f8f9fa;'>
-            <h4 style='color: #3498db; margin-bottom: 10px;'>🎯 Target Sentence (How to Read)</h4>
+            <h4 style='color: #3498db; margin-bottom: 10px;'>🎯 Target Sentence (Reading Guide)</h4>
             <div style='font-size: 20px; font-family: monospace; color: #2c3e50; line-height: 1.4;'>
                 <strong>Original:</strong> {intended}<br>
-                <strong>Thanglish/Manglish:</strong> <span style='color: #e67e22; font-weight: bold;'>{intended_roman}</span>
+                <strong>Romanized:</strong> <span style='color: #e67e22; font-weight: bold;'>{intended_roman}</span>
             </div>
         </div>
         """
@@ -316,7 +351,7 @@ def create_tabular_feedback(intended, actual, lang_choice):
                 <tr style='border-bottom: 2px solid #ddd;'>
                     <th style='padding: 15px; text-align: left; font-weight: bold; color: #2c3e50; border-right: 1px solid #ddd;'>Type</th>
                     <th style='padding: 15px; text-align: left; font-weight: bold; color: #2c3e50; border-right: 1px solid #ddd;'>Original Text</th>
-                    <th style='padding: 15px; text-align: left; font-weight: bold; color: #2c3e50;'>Thanglish/Manglish</th>
+                    <th style='padding: 15px; text-align: left; font-weight: bold; color: #2c3e50;'>Romanized</th>
                 </tr>
             </thead>
             <tbody>
@@ -590,10 +625,11 @@ with gr.Blocks(title="Pronunciation Comparator", theme=gr.themes.Soft()) as demo
     )
     
     intended_transliteration = gr.Textbox(
-        label="🔤 How to Read (Thanglish/Manglish)", 
+        label="🔤 Pronunciation Guide", 
         interactive=False,
-        placeholder="Natural romanization will appear here...",
-        visible=False
+        placeholder="Pronunciation guide will appear here...",
+        visible=False,
+        lines=1
     )
 
     with gr.Row():
@@ -613,7 +649,7 @@ with gr.Blocks(title="Pronunciation Comparator", theme=gr.themes.Soft()) as demo
     with gr.Row():
         with gr.Column():
             pass1_out = gr.Textbox(label="🗣️ What You Said", interactive=False)
-            actual_roman_out = gr.Textbox(label="🔤 Your Pronunciation (Thanglish/Manglish)", interactive=False)
+            actual_roman_out = gr.Textbox(label="🔤 Your Pronunciation (Romanized)", interactive=False)
         with gr.Column():
             wer_out = gr.Textbox(label="📊 Word Error Rate", interactive=False)
             cer_out = gr.Textbox(label="📈 Character Error Rate", interactive=False)