Scalability to very wide schemas

Investigate the behavior of O-Funnel on very wide schemas containing hundreds of candidate fields, particularly with respect to increasing ambiguity and the growth of its O(N · R) resolution cost.

Background

O-Funnel resolves each requirement by walking the captured document tree, giving a resolution cost of O(N * R), where N is the number of document nodes and R is the number of requirements. The evaluation uses small documents containing at most approximately fifty elements and does not assess substantially wider schemas.

The paper explicitly identifies very wide schemas as an unstudied setting in which both candidate ambiguity and computational cost may increase. Characterizing accuracy, ambiguity handling, and practical scalability in that regime remains unresolved.

References

Our documents are small (a handful to about fifty elements); we have not studied very wide schemas with hundreds of candidate fields, where ambiguity and the O(N \cdot R) cost both grow.

— O-Funnel: Lossless Structural Capture and Requirement-Driven Extraction from Drifting, Heterogeneous Documents  (2609.39209 - Mustafa, 30 Sep 2026) in Section 7, Discussion, Limitations