fix: Detect RTL text and put according HTML tags

Signed-off-by: Christoph Auer <[email protected]>
DS4SD · Feb 5, 2025 · 7a00100 · 7a00100
1 parent 327f902
commit 7a00100
Show file tree

Hide file tree

Showing 9 changed files with 221 additions and 193 deletions.
diff --git a/docling_core/types/doc/document.py b/docling_core/types/doc/document.py
@@ -44,7 +44,7 @@
 from docling_core.types.doc.base import ImageRefMode
 from docling_core.types.doc.labels import CodeLanguageLabel, DocItemLabel, GroupLabel
 from docling_core.types.doc.tokens import DocumentToken, TableToken
-from docling_core.types.doc.utils import relative_path
+from docling_core.types.doc.utils import get_text_direction, relative_path
 
 _logger = logging.getLogger(__name__)
 
@@ -866,7 +866,8 @@ def export_to_html(
 
         caption_text = ""
         if len(text) > 0:
-            caption_text = f"<figcaption>{text}</figcaption>"
+            dir = get_text_direction(text)
+            caption_text = f'<figcaption dir="{dir}">{text}</figcaption>'
 
         default_response = f"<figure>{caption_text}</figure>"
 
@@ -1090,15 +1091,23 @@ def export_to_html(
                 if colspan > 1:
                     opening_tag += f' colspan="{colspan}"'
 
+                dir = get_text_direction(content)
+                opening_tag += f' dir="{dir}"'
+
                 body += f"<{opening_tag}>{content}</{celltag}>"
             body += "</tr>"
 
+        dir = get_text_direction(text)
+
         if len(text) > 0 and len(body) > 0:
-            body = f"<table><caption>{text}</caption><tbody>{body}</tbody></table>"
+            body = (
+                f'<table><caption dir="{dir}">{text}</caption>'
+                f"<tbody>{body}</tbody></table>"
+            )
         elif len(text) == 0 and len(body) > 0:
             body = f"<table><tbody>{body}</tbody></table>"
         elif len(text) > 0 and len(body) == 0:
-            body = f"<table><caption>{text}</caption></table>"
+            body = f'<table><caption dir="{dir}">{text}</caption></table>'
         else:
             body = "<table></table>"
 
@@ -2470,16 +2479,18 @@ def _prepare_tag_content(
                 continue
 
             elif isinstance(item, TextItem) and item.label in [DocItemLabel.TITLE]:
-
-                text = f"<h1>{_prepare_tag_content(item.text)}</h1>"
+                text_inner = _prepare_tag_content(item.text)
+                dir = get_text_direction(item.text)
+                text = f'<h1 dir="{dir}">{text_inner}</h1>'
                 html_texts.append(text)
 
             elif isinstance(item, SectionHeaderItem):
 
                 section_level: int = min(item.level + 1, 6)
+                dir = get_text_direction(item.text)
 
                 text = (
-                    f"<h{(section_level)}>"
+                    f'<h{(section_level)} dir="{dir}">'
                     f"{_prepare_tag_content(item.text)}</h{(section_level)}>"
                 )
                 html_texts.append(text)
@@ -2544,13 +2555,13 @@ def _image_fallback(item: TextItem):
                     )
 
             elif isinstance(item, ListItem):
-
-                text = f"<li>{_prepare_tag_content(item.text)}</li>"
+                dir = get_text_direction(item.text)
+                text = f'<li dir="{dir}">{_prepare_tag_content(item.text)}</li>'
                 html_texts.append(text)
 
             elif isinstance(item, TextItem) and item.label in [DocItemLabel.LIST_ITEM]:
-
-                text = f"<li>{_prepare_tag_content(item.text)}</li>"
+                dir = get_text_direction(item.text)
+                text = f'<li dir="{dir}">{_prepare_tag_content(item.text)}</li>'
                 html_texts.append(text)
 
             elif isinstance(item, CodeItem):
@@ -2561,8 +2572,8 @@ def _image_fallback(item: TextItem):
                 html_texts.append(text)
 
             elif isinstance(item, TextItem):
-
-                text = f"<p>{_prepare_tag_content(item.text)}</p>"
+                dir = get_text_direction(item.text)
+                text = f'<p dir="{dir}">{_prepare_tag_content(item.text)}</p>'
                 html_texts.append(text)
             elif isinstance(item, TableItem):
 

diff --git a/docling_core/types/doc/utils.py b/docling_core/types/doc/utils.py
@@ -5,6 +5,7 @@
 
 """Utils for document types."""
 
+import unicodedata
 from pathlib import Path
 
 
@@ -46,3 +47,19 @@ def relative_path(src: Path, target: Path) -> Path:
 
     # Combine and return the result
     return Path(*up_segments, *down_segments)
+
+
+def get_text_direction(text):
+    """Determine the text direction of a given string as LTR or RTL script."""
+    if not text:
+        return "ltr"  # Default for empty input
+
+    rtl_scripts = {"R", "AL"}
+    rtl_chars = sum(unicodedata.bidirectional(c) in rtl_scripts for c in text)
+
+    return (
+        "rtl"
+        if unicodedata.bidirectional(text[0]) in rtl_scripts
+        or rtl_chars > len(text) / 2
+        else "ltr"
+    )
diff --git a/test/data/doc/2206.01062.yaml.html b/test/data/doc/2206.01062.yaml.html
diff --git a/test/data/doc/bad_doc.yaml.html b/test/data/doc/bad_doc.yaml.html
@@ -69,6 +69,6 @@
     }
     </style>
     </head>
-<h1>This is the title</h1>
-<h2>This is the first section</h2>
+<h1 dir="ltr">This is the title</h1>
+<h2 dir="ltr">This is the first section</h2>
 </html>
diff --git a/test/data/doc/constructed_doc.embedded.html.gt b/test/data/doc/constructed_doc.embedded.html.gt
@@ -69,23 +69,23 @@
     }
     </style>
     </head>
-<h1>Title of the Document</h1>
-<p>Author 1<br>Affiliation 1</p>
-<p>Author 2<br>Affiliation 2</p>
-<h2>1. Introduction</h2>
-<p>This paper introduces the biggest invention ever made. ...</p>
+<h1 dir="ltr">Title of the Document</h1>
+<p dir="ltr">Author 1<br>Affiliation 1</p>
+<p dir="ltr">Author 2<br>Affiliation 2</p>
+<h2 dir="ltr">1. Introduction</h2>
+<p dir="ltr">This paper introduces the biggest invention ever made. ...</p>
 <ul>
-<li>list item 1</li>
-<li>list item 2</li>
-<li>list item 3</li>
+<li dir="ltr">list item 1</li>
+<li dir="ltr">list item 2</li>
+<li dir="ltr">list item 3</li>
 <ol>
-<li>list item 3.a</li>
-<li>list item 3.b</li>
-<li>list item 3.c</li>
+<li dir="ltr">list item 3.a</li>
+<li dir="ltr">list item 3.b</li>
+<li dir="ltr">list item 3.c</li>
 </ol>
-<li>list item 4</li>
+<li dir="ltr">list item 4</li>
 </ul>
-<table><caption>This is the caption of table 1.</caption><tbody><tr><td rowspan="2">Product</td><td colspan="2">Years</td></tr><tr><td>2016</td><td>2017</td></tr><tr><td>Apple</td><td>49823</td><td>695944</td></tr></tbody></table>
-<figure><figcaption>This is the caption of figure 1.</figcaption></figure>
-<figure><figcaption>This is the caption of figure 2.</figcaption><img src="data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAEAAAABACAIAAAAlC+aJAAABNElEQVR4nO2ayw7DIAwEIf//z+nBErKAEBMe9rrMmYRZNr60xDCVW7AmTt1xwtsk0uu2H3rDiPosiY/PzlLnfFPpfmqFOqdXqGP9anWOXOsSrttp37WdKMBm+65NX7pSUc9oK7YasGAf3jQeAxixJxoy0iE2Sz2AqeMnnpQqAQzaE1WxPIBZe6LU8zUDxo+fyCQdNQBx/ARX9dIA0PETSdhFA3DHT5C2iwaggQ8QQQcgAd/ACaDNCaDNCaDNCaDNCaDNNfc/w81EDw1oC4ziIgDoGJC2iwYCYAlJ2EsDAaoEruqogQBSQibpq4FgvoRSr9KA2QxVsfonZDDDk5K7GUiYKqEh02rASIa2hkhS6xdsiZxoBlSqEG4qHeLNGeTb/dO1Sw7wxVcO8NVjDvDl75L91+9/ESIkdDQ3IX0AAAAASUVORK5CYII="></figure>
+<table><caption dir="ltr">This is the caption of table 1.</caption><tbody><tr><td rowspan="2" dir="ltr">Product</td><td colspan="2" dir="ltr">Years</td></tr><tr><td dir="ltr">2016</td><td dir="ltr">2017</td></tr><tr><td dir="ltr">Apple</td><td dir="ltr">49823</td><td dir="ltr">695944</td></tr></tbody></table>
+<figure><figcaption dir="ltr">This is the caption of figure 1.</figcaption></figure>
+<figure><figcaption dir="ltr">This is the caption of figure 2.</figcaption><img src="data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAEAAAABACAIAAAAlC+aJAAABNElEQVR4nO2ayw7DIAwEIf//z+nBErKAEBMe9rrMmYRZNr60xDCVW7AmTt1xwtsk0uu2H3rDiPosiY/PzlLnfFPpfmqFOqdXqGP9anWOXOsSrttp37WdKMBm+65NX7pSUc9oK7YasGAf3jQeAxixJxoy0iE2Sz2AqeMnnpQqAQzaE1WxPIBZe6LU8zUDxo+fyCQdNQBx/ARX9dIA0PETSdhFA3DHT5C2iwaggQ8QQQcgAd/ACaDNCaDNCaDNCaDNCaDNNfc/w81EDw1oC4ziIgDoGJC2iwYCYAlJ2EsDAaoEruqogQBSQibpq4FgvoRSr9KA2QxVsfonZDDDk5K7GUiYKqEh02rASIa2hkhS6xdsiZxoBlSqEG4qHeLNGeTb/dO1Sw7wxVcO8NVjDvDl75L91+9/ESIkdDQ3IX0AAAAASUVORK5CYII="></figure>
 </html>
diff --git a/test/data/doc/constructed_doc.placeholder.html.gt b/test/data/doc/constructed_doc.placeholder.html.gt
@@ -69,23 +69,23 @@
     }
     </style>
     </head>
-<h1>Title of the Document</h1>
-<p>Author 1<br>Affiliation 1</p>
-<p>Author 2<br>Affiliation 2</p>
-<h2>1. Introduction</h2>
-<p>This paper introduces the biggest invention ever made. ...</p>
+<h1 dir="ltr">Title of the Document</h1>
+<p dir="ltr">Author 1<br>Affiliation 1</p>
+<p dir="ltr">Author 2<br>Affiliation 2</p>
+<h2 dir="ltr">1. Introduction</h2>
+<p dir="ltr">This paper introduces the biggest invention ever made. ...</p>
 <ul>
-<li>list item 1</li>
-<li>list item 2</li>
-<li>list item 3</li>
+<li dir="ltr">list item 1</li>
+<li dir="ltr">list item 2</li>
+<li dir="ltr">list item 3</li>
 <ol>
-<li>list item 3.a</li>
-<li>list item 3.b</li>
-<li>list item 3.c</li>
+<li dir="ltr">list item 3.a</li>
+<li dir="ltr">list item 3.b</li>
+<li dir="ltr">list item 3.c</li>
 </ol>
-<li>list item 4</li>
+<li dir="ltr">list item 4</li>
 </ul>
-<table><caption>This is the caption of table 1.</caption><tbody><tr><td rowspan="2">Product</td><td colspan="2">Years</td></tr><tr><td>2016</td><td>2017</td></tr><tr><td>Apple</td><td>49823</td><td>695944</td></tr></tbody></table>
-<figure><figcaption>This is the caption of figure 1.</figcaption></figure>
-<figure><figcaption>This is the caption of figure 2.</figcaption></figure>
+<table><caption dir="ltr">This is the caption of table 1.</caption><tbody><tr><td rowspan="2" dir="ltr">Product</td><td colspan="2" dir="ltr">Years</td></tr><tr><td dir="ltr">2016</td><td dir="ltr">2017</td></tr><tr><td dir="ltr">Apple</td><td dir="ltr">49823</td><td dir="ltr">695944</td></tr></tbody></table>
+<figure><figcaption dir="ltr">This is the caption of figure 1.</figcaption></figure>
+<figure><figcaption dir="ltr">This is the caption of figure 2.</figcaption></figure>
 </html>
diff --git a/test/data/doc/constructed_doc.referenced.html.gt b/test/data/doc/constructed_doc.referenced.html.gt
@@ -69,23 +69,23 @@
     }
     </style>
     </head>
-<h1>Title of the Document</h1>
-<p>Author 1<br>Affiliation 1</p>
-<p>Author 2<br>Affiliation 2</p>
-<h2>1. Introduction</h2>
-<p>This paper introduces the biggest invention ever made. ...</p>
+<h1 dir="ltr">Title of the Document</h1>
+<p dir="ltr">Author 1<br>Affiliation 1</p>
+<p dir="ltr">Author 2<br>Affiliation 2</p>
+<h2 dir="ltr">1. Introduction</h2>
+<p dir="ltr">This paper introduces the biggest invention ever made. ...</p>
 <ul>
-<li>list item 1</li>
-<li>list item 2</li>
-<li>list item 3</li>
+<li dir="ltr">list item 1</li>
+<li dir="ltr">list item 2</li>
+<li dir="ltr">list item 3</li>
 <ol>
-<li>list item 3.a</li>
-<li>list item 3.b</li>
-<li>list item 3.c</li>
+<li dir="ltr">list item 3.a</li>
+<li dir="ltr">list item 3.b</li>
+<li dir="ltr">list item 3.c</li>
 </ol>
-<li>list item 4</li>
+<li dir="ltr">list item 4</li>
 </ul>
-<table><caption>This is the caption of table 1.</caption><tbody><tr><td rowspan="2">Product</td><td colspan="2">Years</td></tr><tr><td>2016</td><td>2017</td></tr><tr><td>Apple</td><td>49823</td><td>695944</td></tr></tbody></table>
-<figure><figcaption>This is the caption of figure 1.</figcaption></figure>
-<figure><figcaption>This is the caption of figure 2.</figcaption><img src="constructed_images/image_000001_ccb4cbe7039fe17892f3d611cfb71eafff1d4d230b19b10779334cc4b63c98bc.png"></figure>
+<table><caption dir="ltr">This is the caption of table 1.</caption><tbody><tr><td rowspan="2" dir="ltr">Product</td><td colspan="2" dir="ltr">Years</td></tr><tr><td dir="ltr">2016</td><td dir="ltr">2017</td></tr><tr><td dir="ltr">Apple</td><td dir="ltr">49823</td><td dir="ltr">695944</td></tr></tbody></table>
+<figure><figcaption dir="ltr">This is the caption of figure 1.</figcaption></figure>
+<figure><figcaption dir="ltr">This is the caption of figure 2.</figcaption><img src="constructed_images/image_000001_ccb4cbe7039fe17892f3d611cfb71eafff1d4d230b19b10779334cc4b63c98bc.png"></figure>
 </html>
diff --git a/test/data/doc/constructed_document.yaml.html b/test/data/doc/constructed_document.yaml.html
@@ -69,23 +69,23 @@
     }
     </style>
     </head>
-<h1>Title of the Document</h1>
-<p>Author 1<br>Affiliation 1</p>
-<p>Author 2<br>Affiliation 2</p>
-<h2>1. Introduction</h2>
-<p>This paper introduces the biggest invention ever made. ...</p>
+<h1 dir="ltr">Title of the Document</h1>
+<p dir="ltr">Author 1<br>Affiliation 1</p>
+<p dir="ltr">Author 2<br>Affiliation 2</p>
+<h2 dir="ltr">1. Introduction</h2>
+<p dir="ltr">This paper introduces the biggest invention ever made. ...</p>
 <ul>
-<li>list item 1</li>
-<li>list item 2</li>
-<li>list item 3</li>
+<li dir="ltr">list item 1</li>
+<li dir="ltr">list item 2</li>
+<li dir="ltr">list item 3</li>
 <ol>
-<li>list item 3.a</li>
-<li>list item 3.b</li>
-<li>list item 3.c</li>
+<li dir="ltr">list item 3.a</li>
+<li dir="ltr">list item 3.b</li>
+<li dir="ltr">list item 3.c</li>
 </ol>
-<li>list item 4</li>
+<li dir="ltr">list item 4</li>
 </ul>
-<table><caption>This is the caption of table 1.</caption><tbody><tr><td rowspan="2">Product</td><td colspan="2">Years</td></tr><tr><td>2016</td><td>2017</td></tr><tr><td>Apple</td><td>49823</td><td>695944</td></tr></tbody></table>
-<figure><figcaption>This is the caption of figure 1.</figcaption></figure>
-<figure><figcaption>This is the caption of figure 2.</figcaption></figure>
+<table><caption dir="ltr">This is the caption of table 1.</caption><tbody><tr><td rowspan="2" dir="ltr">Product</td><td colspan="2" dir="ltr">Years</td></tr><tr><td dir="ltr">2016</td><td dir="ltr">2017</td></tr><tr><td dir="ltr">Apple</td><td dir="ltr">49823</td><td dir="ltr">695944</td></tr></tbody></table>
+<figure><figcaption dir="ltr">This is the caption of figure 1.</figcaption></figure>
+<figure><figcaption dir="ltr">This is the caption of figure 2.</figcaption></figure>
 </html>
diff --git a/test/data/doc/dummy_doc.yaml.html b/test/data/doc/dummy_doc.yaml.html
@@ -69,7 +69,7 @@
     }
     </style>
     </head>
-<h1>DocLayNet: A Large Human-Annotated Dataset for Document-Layout Analysis</h1>
-<figure><figcaption>Figure 1: Four examples of complex page layouts across different document categories</figcaption></figure>
+<h1 dir="ltr">DocLayNet: A Large Human-Annotated Dataset for Document-Layout Analysis</h1>
+<figure><figcaption dir="ltr">Figure 1: Four examples of complex page layouts across different document categories</figcaption></figure>
 
 </html>