orbitkit 0.8.61__tar.gz → 0.8.63__tar.gz

{orbitkit-0.8.61/orbitkit.egg-info → orbitkit-0.8.63}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: orbitkit
-Version: 0.8.61
+Version: 0.8.63
 Summary: This project is only for Orbit Tech internal use.
 Home-page: https://github.com/clown-0726/orbitkit
 Author: Lilu Cao

orbitkit-0.8.63/orbitkit/VERSION

@@ -0,0 +1 @@
+0.8.63

{orbitkit-0.8.61 → orbitkit-0.8.63}/orbitkit/pdf_extractor/pdf_extractor_minerU_v1.py

@@ -429,36 +429,60 @@ class MinerUExtract:
         return self.level_two_txt_com(block_seq, pages_block, pages_body, 'sentence')
 
     def level_two_table(self, block_seq, pages_block, pages_body):
-        """处理 table 类型 block，sentence 直接存储 HTML 格式"""
+        """处理 table 类型 block，sentence 直接存储 HTML 格式，脚注等文本单独生成 block（各自保留 bbox）"""
         table_blocks = pages_block.get('blocks', [])
         table_html = ""
+        table_bbox = pages_block.get('bbox', [0, 0, 0, 0])
         _image_detail_arr = []
-        bbox = pages_block.get('bbox', [0, 0, 0, 0])
+        # 收集非 table 类型的子 block（如 table_footnote, table_caption 等），保留各自的 bbox
+        text_blocks = []
 
-        for table in table_blocks:
-            pages_lines = table.get("lines", [])
+        for sub_block in table_blocks:
+            sub_bbox = sub_block.get('bbox', table_bbox)
+            pages_lines = sub_block.get("lines", [])
             for _line in pages_lines:
                 for _span in _line.get('spans', []):
                     span_type = _span.get('type', '')
                     if span_type == 'table' and 'html' in _span:
                         table_html = _span['html']
+                    elif 'content' in _span:  # table_footnote, table_caption 等文本类型
+                        text_blocks.append({
+                            'content': _span['content'],
+                            'bbox': sub_bbox
+                        })
                     if _span.get('image_path'):
                         _image_detail_arr.append({"path": f'images/{_span["image_path"]}', "desc": ""})
 
-        _block_str = table_html
-        if _block_str or _image_detail_arr:
+        # 表格 HTML 单独一个 block
+        if table_html or _image_detail_arr:
             block_seq = block_seq + 1
-            text_location = self.get_location(bbox)
+            text_location = self.get_location(table_bbox)
             self.blocks.append({
                 "id": id_srv.get_random_short_id(),
                 "page": self.page,
                 "seq_no": block_seq,
-                "sentence": _block_str,
+                "sentence": table_html,
                 "type": 'table',
                 "image_detail": _image_detail_arr,
                 "text_location": text_location
             })
-            pages_body.append(_block_str)
+            pages_body.append(table_html)
+
+        # 每个脚注/标题等文本单独一个 block，使用各自的 bbox
+        for text_block in text_blocks:
+            block_seq = block_seq + 1
+            text_location = self.get_location(text_block['bbox'])
+            self.blocks.append({
+                "id": id_srv.get_random_short_id(),
+                "page": self.page,
+                "seq_no": block_seq,
+                "sentence": text_block['content'],
+                "type": 'sentence',
+                "image_detail": [],
+                "text_location": text_location
+            })
+            pages_body.append(text_block['content'])
+
         return block_seq, pages_body
 
     def level_two_image(self, block_seq, pages_block, pages_body):

{orbitkit-0.8.61 → orbitkit-0.8.63/orbitkit.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: orbitkit
-Version: 0.8.61
+Version: 0.8.63
 Summary: This project is only for Orbit Tech internal use.
 Home-page: https://github.com/clown-0726/orbitkit
 Author: Lilu Cao

orbitkit-0.8.61/orbitkit/VERSION

@@ -1 +0,0 @@
-0.8.61