Post correction.py

post-correction.py

@@ -17,16 +17,44 @@ if not arg_path.exists():
 
 import json
 import ftfy
+import re
+import urllib.request
+
+url = "https://raw.githubusercontent.com/hbenbel/French-Dictionary/master/dictionary/dictionary.txt"
+french_words = urllib.request.urlopen(url).read().decode('utf-8').splitlines()
+
+# 2. Pre-compute an O(1) lookup dictionary
+# We simulate the corruption by replacing accents with null bytes (\x00)
+lookup_map = {}
+for word in french_words:
+    # Replace all French accents with \x00 to create the "broken" key
+    broken_key = re.sub(r'[éèêëàâäîïôöùûüçœÉÈÊËÀÂÄÎÏÔÖÙÛÜÇŒ]', '\x00', word)
+    if '\x00' in broken_key:
+        lookup_map[broken_key] = word # e.g., "\x00cole" -> "école"
+
+# 3. Fast replace function
+def fast_fix(text):
+    # Find words containing regular letters and null bytes
+    def replacer(match):
+        broken_word = match.group(0)
+        # Return the fixed word from our map, or leave it if not found
+        # (Handles case-insensitivity by falling back to lowercase map)
+        return lookup_map.get(broken_word.lower(), broken_word)
+
+    return re.sub(r'[a-zA-Z\x00]+', replacer, text)
+
+
 
 INPUT_FILE = Path(INPUT_DIR) / "correction.json"
-OUPUT_FILE = Path(INPUT_DIR) / "fixed_correction.json"
-
+OUTPUT_FILE = Path(INPUT_DIR) / "fixed_correction.json"
 
 def clean_string(s: str) -> str:
-    # remove null bytes
-    s = s.replace("\x00", "")
     # fix encoding issues
     s = ftfy.fix_text(s)
+    s = re.sub(r'\x19', r'\x00', s)
+    s = re.sub(r'\x18', r'\x00', s)
+    s = fast_fix(s)
+    s = s.replace('\x00', '')
     return s