Update src/gradio_app.py

src/gradio_app.py

@@ -25,6 +25,9 @@ DICT_DIR = os.path.join(os.path.dirname(__file__), "dicts")
 AFF_PATH = os.path.join(DICT_DIR, "kab.aff")
 DIC_PATH = os.path.join(DICT_DIR, "kab.dic")
 
+# Caractères de ponctuation à stripper (définis avec des escapes Unicode pour éviter les problèmes d'encodage)
+PUNCTUATION_CHARS = '.,!?;:"\'()[]{}«»—–-'
+
 _hunspell_dict = None
 
 def get_hunspell():
@@ -45,7 +48,7 @@ def correct_word(word: str) -> str:
     dic = get_hunspell()
 
     # Nettoyage: séparer ponctuation
-    stripped = word.strip(".,!?;:\"'()[]{}«»—–-").lower()
+    stripped = word.strip(PUNCTUATION_CHARS).lower()
     if not stripped:
         return word
 
@@ -71,8 +74,8 @@ def correct_word(word: str) -> str:
         best = best[0].upper() + best[1:]
 
     # Restaurer la ponctuation attachée
-    prefix_len = len(word) - len(word.lstrip(".,!?;:"'()[]{}«»—–-"))
-    suffix_len = len(word) - len(word.rstrip(".,!?;:"'()[]{}«»—–-"))
+    prefix_len = len(word) - len(word.lstrip(PUNCTUATION_CHARS))
+    suffix_len = len(word) - len(word.rstrip(PUNCTUATION_CHARS))
     prefix = word[:prefix_len]
     suffix = word[-suffix_len:] if suffix_len > 0 else ""
 
@@ -97,7 +100,7 @@ def spellcheck_transcript(text: str, auto_correct: bool = True) -> tuple[str, li
         else:
             # Mode suggestion seule: on ne corrige pas, on signale juste
             dic = get_hunspell()
-            stripped = word.strip(".,!?;:\"'()[]{}«»—–-").lower()
+            stripped = word.strip(PUNCTUATION_CHARS).lower()
             corrected = word if (not stripped or dic.lookup(stripped)) else word + " [?]"
 
         corrected_words.append(corrected)