feat: IPA-Lautschrift in Cell-Texte einfuegen (fuer Overlay-Modus)

fix_cell_phonetics() ersetzt fehlerhafte IPA-Klammern UND fuegt fehlende
Lautschrift fuer englische Woerter ein (z.B. badge, film, challenge, profit).
Wird auf alle Zellen mit col_type column_en/column_text angewandt.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

klausur-service/backend/cv_ocr_engines.py

@@ -885,6 +885,105 @@ def _replace_phonetics_in_text(text: str, pronunciation: str = 'british') -> str
     return text
 
 
+def _insert_missing_ipa(text: str, pronunciation: str = 'british') -> str:
+    """Insert IPA pronunciation for English words that have no brackets at all.
+
+    OCR sometimes drops the phonetic transcription entirely (e.g. "challenge"
+    instead of "challenge [ˈtʃælɪndʒ]").  This scans the text for lone English
+    words that have a dictionary IPA entry and appends [ipa] after them.
+
+    Only inserts for words that:
+    - are standalone (not already followed by a bracket)
+    - have an IPA entry in the dictionary
+    - appear to be English headwords (at the start of text or after common
+      separators like ",", ";", "•")
+
+    This is intentionally conservative: it only inserts at the END of each
+    whitespace-separated token group to avoid breaking phrases.
+    """
+    if not IPA_AVAILABLE:
+        return text
+    if not text or not text.strip():
+        return text
+
+    # Skip if already has brackets (IPA replacement handles those)
+    if any(ch in text for ch in '[{('):
+        return text
+
+    # Only process short text fragments (typical vocab cells).
+    # Long sentences / paragraphs should not get IPA insertions.
+    words = text.strip().split()
+    if len(words) > 6:
+        return text
+
+    # Try to insert IPA for the first alphanumeric word
+    # Typical patterns: "challenge", "profit", "film", "badge"
+    for i, w in enumerate(words):
+        # Clean punctuation for lookup
+        clean = re.sub(r'[^a-zA-ZäöüÄÖÜß\'-]', '', w)
+        if not clean or len(clean) < 2:
+            continue
+        # Skip German/grammar words
+        if clean.lower() in _GRAMMAR_BRACKET_WORDS:
+            continue
+        ipa = _lookup_ipa(clean, pronunciation)
+        if ipa:
+            words[i] = f"{w} [{ipa}]"
+            # Only insert for the FIRST word that has IPA
+            # (headword in English column)
+            break
+
+    return ' '.join(words)
+
+
+def fix_cell_phonetics(
+    cells: List[Dict[str, Any]],
+    pronunciation: str = 'british',
+) -> List[Dict[str, Any]]:
+    """Apply IPA phonetic fixes to cell texts for overlay mode.
+
+    In the normal pipeline, _fix_phonetic_brackets operates on vocab entries
+    (entry['english']).  But the overlay reads cell['text'] directly, so
+    phonetic fixes must be applied to cells too.
+
+    This function:
+    1. Replaces garbled IPA brackets with correct dictionary IPA
+    2. Inserts missing IPA for English headwords that have no brackets
+
+    Only processes cells in English-like columns (column_en, column_text).
+    German columns are never processed (they contain meaningful parentheses).
+    """
+    if not IPA_AVAILABLE:
+        return cells
+
+    # Column types where IPA processing makes sense
+    ipa_col_types = {'column_en', 'column_text'}
+    replaced = 0
+
+    for cell in cells:
+        col_type = cell.get('col_type', '')
+        if col_type not in ipa_col_types:
+            continue
+        text = cell.get('text', '') or ''
+        if not text.strip():
+            continue
+
+        # Step 1: replace garbled IPA brackets
+        new_text = _replace_phonetics_in_text(text, pronunciation)
+        # Step 2: insert missing IPA if no brackets were present
+        if new_text == text:
+            new_text = _insert_missing_ipa(text, pronunciation)
+
+        if new_text != text:
+            logger.debug(f"fix_cell_phonetics: '{text}' → '{new_text}'")
+            cell['text'] = new_text
+            replaced += 1
+
+    if replaced:
+        logger.info(f"fix_cell_phonetics: {replaced} IPA fixes in {len(cells)} cells")
+    return cells
+
+
 def _assign_row_words_to_columns(
     row: RowGeometry,
     columns: List[PageRegion],