Detecting highlight boxes

[TRIAEIOU]

First off, great job devs!

A question, is there a way to tweak the settings or add a pipeline step to detect highlight boxes? Many PDF documents use these and lay them out in a manner that splits sentences, in my postprocessing I try to detect this based on node type and text content but it is unreliable (second half of split sentence beginning with a caps abbreviation resulting in "runaway boxes" etc).

It would be of great help if it would be possible to group box content, for instance in a GroupItem? Is that possible somehow? For instance by tweaking the table detection to detect these as single cell tables or other solution?

Thanks