Mark caption tags of Caption type as opposed to NarrativeText

unstructured/documents/elements.py

@@ -880,6 +880,11 @@ class Table(Text):
 
     category = "Table"
 
+class Caption(Text):
+    """An element for capturing captions."""
+
+    category = "Caption"
+
 
 class TableChunk(Table):
     """An element for capturing chunks of tables."""

unstructured/documents/html.py

@@ -25,6 +25,7 @@
     Table,
     Text,
     Title,
+    Caption
 )
 from unstructured.documents.xml import VALID_PARSERS, XMLDocument
 from unstructured.logger import logger
@@ -104,6 +105,8 @@ class HTMLListItem(TagsMixin, ListItem):
 class HTMLTable(TagsMixin, Table):
     """NarrativeText with tag information"""
 
+class HTMLCaption(TagsMixin, Caption):
+    """NarrativeText with tag information"""
 
 def has_table_ancestor(element: TagsMixin) -> bool:
     """Checks to see if an element has ancestors that are table elements. If so, we consider
@@ -466,6 +469,15 @@ def _text_to_element(
             links=links,
             emphasized_texts=emphasized_texts,
         )
+
+    if tag == CAPTION_TAG:
+        return HTMLCaption(
+            text,
+            tag=tag,
+            ancestortags=ancestortags,
+            links=links,
+            emphasized_texts=emphasized_texts,
+        )
 
     if len(text) < 2:
         return None