Clause extraction and the 41-category pack
Pull structured clause facts (type, verbatim text and typed attributes) under the same grounding rules as review.
Extraction runs over a controlled clause vocabulary (the 41 standard categories) and produces structured clause facts. It works like CitationGate: the model can only nominate a span. A span that can't be found verbatim in the source is dropped and counted, never stored.
- Every stored clause fact can be proven: clause type, verbatim text and character offsets.
- Typed attributes (amounts, currencies, percentages, dates, parties) are parsed deterministically by rule from the grounded span. The model does not guess them.
Related documentation
Compare (what changed) and the review grid
Compare shows what changed between the copy you sent and the copy that came back. The review grid, in Tabular Diligence, answers the same questions across many contracts and exports to Excel.
CitationGate: grounding every finding
A finding with no verifiable quote is rejected, and a quote that doesn't match the source is dropped. This is what makes LexCore's output traceable.
Authority-grounded legal research
Answers grounded in Nigerian authority, with every citation verified, superseded law flagged, and coverage limits stated plainly.