Currently, the reading order parser merges overlapping boxes, but implicitly assumes that only two boxes can be overlapping, rather than potential clusters of boxes. This needs to be addressed to merge clusters of boxes if necessary, but we also need to check how we're ending up with so many clusters of boxes.
Currently, the reading order parser merges overlapping boxes, but implicitly assumes that only two boxes can be overlapping, rather than potential clusters of boxes. This needs to be addressed to merge clusters of boxes if necessary, but we also need to check how we're ending up with so many clusters of boxes.