Normalize spaced-letter obfuscation in text cleaning

app/utils/text_utils.py

@@ -2,6 +2,38 @@ import re
 import unicodedata
 
 
+def _merge_spaced_letter_chains(text: str) -> str:
+    tokens = text.split()
+    if not tokens:
+        return text
+
+    merged = []
+    i = 0
+    n = len(tokens)
+
+    while i < n:
+        tok = tokens[i]
+        if len(tok) == 1 and tok.isalpha():
+            letters = [tok]
+            j = i + 1
+            while j < n and len(tokens[j]) == 1 and tokens[j].isalpha():
+                letters.append(tokens[j])
+                j += 1
+
+            # Join only real obfuscation chains like "g e r i z e k a l i".
+            if len(letters) >= 2:
+                merged.append("".join(letters))
+            else:
+                merged.append(tok)
+            i = j
+            continue
+
+        merged.append(tok)
+        i += 1
+
+    return " ".join(merged)
+
+
 def clean_text_nfkc(text: str) -> str:
     text = unicodedata.normalize('NFKC', str(text))
     text = text.replace('İ', 'i').replace('I', 'ı').lower()
@@ -11,7 +43,8 @@ def clean_text_nfkc(text: str) -> str:
         text = text.replace(key, value)
     # Keep natural double letters (e.g., "kullanici") and only squash exaggerated repeats.
     text = re.sub(r'(.)\1{2,}', r'\1', text)
-    return " ".join(text.split())
+    text = " ".join(text.split())
+    return _merge_spaced_letter_chains(text)
 
 
 def check_blacklist(text: str, blacklist_set: set) -> bool: