Curious about how you handle clause level parsing when clauses are highly context dependent (e.g., indemnity linked to limitation of liability, or IP ownership

Curious about how you handle clause level parsing when clauses are highly context dependent (e.g., indemnity linked to limitation of liability, or IP ownership tied to subcontractor terms). Also how do you approach cross referencing between clauses? A lot of “AI contract analysis” tools I’ve seen tend to miss dependencies like “subject to Section 10(b)” or misinterpret defined terms. Is your model resolving definitions and references in a structured graph, or relying purely on token level embeddings?