docling
docling copied to clipboard
Multicolumn text recognition breaks in image
Bug
...
"The original file is shown in Figure 1, and the Docling parsing result in HTML is shown in Figure 2. The two parts marked in red in Figure 2 belong to two columns on the page, but the layout analysis results classify these two parts as the same type of block, so they should be merged together."
Figure 1:
Figure 2:
Steps to reproduce
...
Docling version
...
Python version
...