Update README.md

README.md

@@ -16,7 +16,7 @@ If you find this dataset helpful, please consider supporting me:
 </a>
 
 
-Dataset Summary
+## Dataset Summary
 
 DocBank is a large-scale dataset tailored for Document AI tasks, focusing on integrating textual and layout information. It comprises 500,000 document pages, divided into 400,000 for training, 50,000 for validation, and 50,000 for testing. The dataset is generated using a weak supervision approach, enabling efficient annotation of document structures without extensive manual labeling.
 Supported Tasks and Leaderboards
@@ -24,12 +24,12 @@ Supported Tasks and Leaderboards
     Document Layout Analysis: Identifying and classifying different layout elements within documents based on text and spatial information.
     Token Classification: Assigning layout type classes to individual tokens within the document.
 
-Languages
+## Languages
 
     English
 
-Dataset Structure
-Data Instances
+## Dataset Structure
+### Data Instances
 
 Each instance represents a single document page with the following fields:
 
@@ -40,7 +40,7 @@ Each instance represents a single document page with the following fields:
         bounding_box: Coordinates defining the position of the token on the page ([x1, y1, x2, y2]).
         label: Layout type class indicating the role of the token (e.g., title, paragraph, table).
 
-Data Fields
+## Data Fields
 
     filename (string): MD5 hash of the original filename, serving as a unique identifier.
     page_bounding_box (list of int): [min_x, min_y, max_x, max_y] coordinates of the entire document page.
@@ -49,8 +49,8 @@ Data Fields
         bounding_box (list of int): [x1, y1, x2, y2] coordinates of the token's position.
         label (string): Layout type class for the token.
 
-Additional Information
-Dataset Curators
+## Additional Information
+### Dataset Curators
 
 Minghao Li, Yiheng Xu, Lei Cui, Shaohan Huang, Furu Wei, Zhoujun Li, Ming Zhou
 Licensing Information