fix(pdf_parse): improve span removal logic for all content types

- Update remove_outside_spans function to handle all content types
- Add processing for text and equation spans
- Improve overlap calculation for better accuracy

magic_pdf/pdf_parse_union_core_v2.py

@@ -385,17 +385,17 @@ def revert_group_blocks(blocks):
 def remove_outside_spans(spans, all_bboxes):
     image_bboxes = []
     table_bboxes = []
-    all_block_bboxes = []
+    other_block_bboxes = []
     for block in all_bboxes:
         block_type = block[7]
         block_bbox = block[0:4]
-        all_block_bboxes.append(block_bbox)
+
         if block_type == BlockType.ImageBody:
             image_bboxes.append(block_bbox)
         elif block_type == BlockType.TableBody:
             table_bboxes.append(block_bbox)
         else:
-            continue
+            other_block_bboxes.append(block_bbox)
 
     new_spans = []
 
@@ -411,7 +411,7 @@ def remove_outside_spans(spans, all_bboxes):
                     new_spans.append(span)
                     break
         elif span['type'] in [ContentType.Text, ContentType.InlineEquation, ContentType.InterlineEquation]:
-            for block_bbox in all_block_bboxes:
+            for block_bbox in other_block_bboxes:
                 if calculate_overlap_area_in_bbox1_area_ratio(span['bbox'], block_bbox) > 0.5:
                     new_spans.append(span)
                     break