From 6bdaddb366fa400f194b4cc5e21f084a2bd9e8bc Mon Sep 17 00:00:00 2001
From: Research Assistant <research@example.com>
Date: Sun, 7 Jun 2026 11:45:11 +0800
Subject: [PATCH] fix: prevent author bylines from becoming section headings

---
 paperforge/worker/ocr_roles.py | 21 +++++++++++++++++++++
 tests/test_ocr_roles.py        | 26 ++++++++++++++++++++++++++
 2 files changed, 47 insertions(+)

diff --git a/paperforge/worker/ocr_roles.py b/paperforge/worker/ocr_roles.py
index bbcf58e6..069cf24c 100644
--- a/paperforge/worker/ocr_roles.py
+++ b/paperforge/worker/ocr_roles.py
@@ -482,6 +482,27 @@ def assign_block_role(
                     f"backmatter {'boundary heading' if role_name == 'backmatter_boundary_heading' else 'boundary candidate'}: {text[:60]}"
                 ],
             )
+        # Author-like heading guard: prevent bylines from becoming headings
+        _is_author_byline = (
+            re.search(r"&|,.*,", text)
+            and len(text.split()) <= 15
+            and not _has_heading_numbering(text)
+            and not any(text.lower().startswith(w) for w in ["abstract", "introduction", "methods", "results", "discussion", "conclusion", "references"])
+        )
+        if _is_author_byline and page_num == 1:
+            return RoleAssignment(
+                role="frontmatter_noise",
+                confidence=0.5,
+                evidence=[f"author byline on page 1, not heading: {text[:60]}"],
+            )
+
+        if re.search(r"(?:correspondence|orcid|@)", text.lower()) and len(text.split()) <= 20 and page_num == 1:
+            return RoleAssignment(
+                role="frontmatter_noise",
+                confidence=0.5,
+                evidence=[f"correspondence/noise on page 1: {text[:60]}"],
+            )
+
         if _has_heading_numbering(text):
             depth = _heading_number_depth(text)
             return RoleAssignment(
diff --git a/tests/test_ocr_roles.py b/tests/test_ocr_roles.py
index 86277a06..006f4f5b 100644
--- a/tests/test_ocr_roles.py
+++ b/tests/test_ocr_roles.py
@@ -490,6 +490,32 @@ def test_weak_backmatter_boundary_signal_emits_candidate():
     )
 
 
+def test_author_byline_not_section_heading():
+    from paperforge.worker.ocr_roles import assign_block_role
+
+    # Author byline with & on page 1
+    block = {"block_label": "paragraph_title", "block_content": "John Smith & Jane Doe", "block_bbox": [100, 50, 500, 80], "page": 1}
+    page_blocks = [block,
+        {"block_label": "text", "block_content": "Some frontmatter", "block_bbox": [100, 100, 500, 130], "page": 1},
+        {"block_label": "text", "block_content": "Some abstract content", "block_bbox": [100, 200, 500, 230], "page": 1},
+    ]
+    result = assign_block_role(block, page_blocks=page_blocks, page_width=600, page_height=800)
+    assert result.role not in ("section_heading", "subsection_heading", "sub_subsection_heading"), (
+        f"Author byline got heading role: {result.role}"
+    )
+
+
+def test_correspondence_marker_not_heading():
+    from paperforge.worker.ocr_roles import assign_block_role
+
+    block = {"block_label": "paragraph_title", "block_content": "*Correspondence: john@example.com", "block_bbox": [100, 500, 500, 530], "page": 1}
+    page_blocks = [block]
+    result = assign_block_role(block, page_blocks=page_blocks, page_width=600, page_height=800)
+    assert result.role in ("frontmatter_noise", "unknown_structural"), (
+        f"Correspondence got role: {result.role}"
+    )
+
+
 def test_backmatter_boundary_detects_on_early_page() -> None:
     """Backmatter boundary should be detectable on papers with fewer
     than 8 pages, without a hard page gate."""