Interprets visual layout, hierarchical structures, and spatial cues to preserve document intent and meaning.
Recovers what the layout means rather than what it says: reading order across columns, section hierarchy, which caption belongs to which figure, where a table continues onto the next page. Underpins most of the other solutions in this group rather than being deployed alone.
Long structured documents where the hierarchy carries meaning — contracts, regulatory filings, technical manuals, annual reports.
Short flat documents. A one-page invoice has no structure worth recovering.
Where reading order affects meaning — a clause's scope depending on which section contains it — the recovered structure should be verifiable rather than assumed.
Reading order errors in multi-column layouts, which reorder sentences into text that is grammatical, plausible and not what the document said.
Parsing a lease to determine which obligations sit under which clause, where a term recovered from the wrong section changes the accounting treatment.