viewers

doocus/extractors/office.py

@@ -11,9 +11,29 @@ from .base import ExtractionResult
 
 MAX_SHEET_ROWS = 200
 
+# Generic core-property titles Office writes by default — worse than useless as a
+# display title (they hide the real one), so we drop them and derive from content.
+PLACEHOLDER_TITLES = {
+    "word document", "powerpoint presentation", "excel workbook",
+    "microsoft word document", "microsoft powerpoint presentation",
+    "microsoft excel worksheet", "presentation", "workbook", "document",
+}
+
+
+def _good_title(t) -> bool:
+    return bool(t) and str(t).strip().lower() not in PLACEHOLDER_TITLES
+
+
+def _first_line(text: str, limit: int = 120) -> str:
+    for ln in text.splitlines():
+        s = ln.lstrip("#").strip()
+        if s:
+            return s[:limit]
+    return ""
+
 
 def _core_props(props) -> dict:
-    """Common OOXML core properties → metadata (only non-empty ones).
+    """Common OOXML core properties → metadata (only non-empty, non-placeholder).
 
     python-docx/pptx expose `.author`/`.last_modified_by`; openpyxl uses
     `.creator`/`.lastModifiedBy`. Each target checks both spellings.
@@ -31,6 +51,8 @@ def _core_props(props) -> dict:
         for src in sources:
             val = getattr(props, src, None)
             if val:
+                if dst == "title" and not _good_title(val):
+                    continue  # skip "Word Document" & friends; derive from content
                 out[dst] = str(val)
                 break
     return out
@@ -43,6 +65,10 @@ def _extract_docx(source: Path) -> ExtractionResult:
     content = "\n\n".join(p for p in paras if p.strip()) + "\n"
     metadata = _core_props(doc.core_properties)
     metadata["paragraphs"] = len([p for p in paras if p.strip()])
+    if "title" not in metadata:
+        t = _first_line(content)
+        if t:
+            metadata["title"] = t
     return ExtractionResult(content=content, metadata=metadata, extractor="office/python-docx@1")
 
 
@@ -62,6 +88,12 @@ def _extract_pptx(source: Path) -> ExtractionResult:
     content = "\n\n".join(blocks) + "\n"
     metadata = _core_props(prs.core_properties)
     metadata["slides"] = len(prs.slides)
+    if "title" not in metadata and prs.slides:
+        # first slide's title placeholder, if any
+        for shape in prs.slides[0].shapes:
+            if shape.has_text_frame and shape.text_frame.text.strip():
+                metadata["title"] = shape.text_frame.text.strip()[:120]
+                break
     return ExtractionResult(content=content, metadata=metadata, extractor="office/python-pptx@1")