fix: deduplicate overlapping OCR words and use per-word Y positions in overlay

Backend: Add spatial overlap check (>=50% horizontal IoU) to Kombi merge
so words at the same position are deduplicated even when OCR text differs.

Frontend: Add yPct/hPct to WordPosition so each word renders at its actual
vertical position instead of all words collapsing to the cell center Y.

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

klausur-service/backend/ocr_pipeline_api.py

@@ -2704,6 +2704,19 @@ def _merge_row_sequences(paddle_row: list, tess_row: list) -> list:
         # Same text or one contains the other
         is_same = (pt == tt) or (len(pt) > 1 and len(tt) > 1 and (pt in tt or tt in pt))
 
+        # Spatial overlap check: if words overlap >= 50% horizontally,
+        # they're the same physical word regardless of OCR text differences
+        if not is_same:
+            overlap_left = max(pw["left"], tw["left"])
+            overlap_right = min(
+                pw["left"] + pw.get("width", 0),
+                tw["left"] + tw.get("width", 0),
+            )
+            overlap_w = max(0, overlap_right - overlap_left)
+            min_w = min(pw.get("width", 1), tw.get("width", 1))
+            if min_w > 0 and overlap_w / min_w >= 0.5:
+                is_same = True
+
         if is_same:
             # Matched — average coordinates weighted by confidence
             pc = pw.get("conf", 80)

klausur-service/backend/tests/test_paddle_kombi.py

@@ -410,6 +410,45 @@ class TestMergeRealWorldRegression:
         assert abs(be_word["top"] - take_word["top"]) > 30, "Rows should stay separate"
 
 
+class TestSpatialOverlapDedup:
+    """Test that words at the same position are deduplicated even if text differs."""
+
+    def test_same_position_different_text_deduplicated(self):
+        """Both engines find same physical word but OCR text differs slightly.
+        Spatial overlap should catch this as a duplicate."""
+        paddle = [_word("hello", 100, 50, 80, 20, conf=90)]
+        tess = [_word("helo", 102, 52, 76, 18, conf=70)]
+        merged = _merge_row_sequences(paddle, tess)
+        assert len(merged) == 1, (
+            f"Expected 1 word (deduped by overlap), got {len(merged)}: "
+            f"{[w['text'] for w in merged]}"
+        )
+        # Paddle text preferred (higher confidence)
+        assert merged[0]["text"] == "hello"
+
+    def test_same_position_single_char_deduplicated(self):
+        """Single-char words at same position should be deduplicated via overlap."""
+        paddle = [_word("a", 100, 50, 20, 20, conf=90)]
+        tess = [_word("a!", 101, 51, 22, 19, conf=60)]
+        merged = _merge_row_sequences(paddle, tess)
+        assert len(merged) == 1
+
+    def test_no_overlap_different_words_kept(self):
+        """Different words at different positions should both be kept."""
+        paddle = [_word("cat", 100, 50, 50, 20, conf=90)]
+        tess = [_word("dog", 300, 50, 50, 20, conf=70)]
+        merged = _merge_row_sequences(paddle, tess)
+        assert len(merged) == 2
+
+    def test_partial_overlap_below_threshold_kept(self):
+        """Words with < 50% overlap are different words and both kept."""
+        paddle = [_word("take", 100, 50, 60, 20, conf=90)]
+        tess = [_word("part", 145, 50, 60, 20, conf=70)]
+        merged = _merge_row_sequences(paddle, tess)
+        # 15px overlap / 60px min width = 25% < 50% → kept as separate
+        assert len(merged) == 2
+
+
 class TestSplitThenMerge:
     """Test the full pipeline: split multi-word Paddle boxes, then merge."""