A proposed framework for ambiguous, conflicting and contradicting clause detection in construction regulations

Tools

Balakrishnan, Govardhini and Kumar, Bimal and Robinson, Melanie and West, Graeme; (2026) A proposed framework for ambiguous, conflicting and contradicting clause detection in construction regulations. In: Proceedings of the TUM GNI International Symposium 2026. TUM Georg Nemetschek Institute, Munich. (In Press)

[thumbnail of Balakrishnan-etal-TUM-GNI-2026-A-proposed-framework-for-ambiguous-conflicting-and-contradicting-clause-detection-in-construction-regulations] Text. Filename: Balakrishnan-etal-TUM-GNI-2026-A-proposed-framework-for-ambiguous-conflicting-and-contradicting-clause-detection-in-construction-regulations.pdf
Accepted Author Manuscript
Restricted to Repository staff only until 1 January 2099.
Download (1MB) | Request a copy

Abstract

Automated Compliance Checking (ACC) has traditionally focused on translating regulatory provisions into computable rules. However, a persistent and underexplored challenge is that regulatory texts themselves often contain ambiguity, internal conflict, and logical contradiction, which inhibit reliable digitisation at the outset. This paper addresses this foundational pre-processing gap by proposing DeCCAR (Detection of Conflict, Contradiction, and Ambiguity in Regulations), a hybrid interpretive framework that combines controlled Large Language Model (LLM) hypothesis probing, ontology-based semantic graph modelling, and expert adjudication to identify clauses that are unstable for automation prior to rule formalisation. Using the Scottish Building Standards Technical Handbooks (Domestic and Non-Domestic), with a focus on fire and life safety provisions, a controlled pilot study is conducted on a curated subset of clauses. LLM outputs are evaluated against reference labels using precision, recall, F1-score, and error tendency pattern analysis. Results demonstrate that while LLMs are unreliable as autonomous interpreters, their interpretive instability provides a robust diagnostic signal for identifying problematic regulatory constructs. The paper contributes a model-agnostic, ambiguity-aware interpretive framework that supports more reliable Compliance checking process by explicitly addressing regulatory instability before computational reasoning is applied.

ORCID iDs

Balakrishnan, Govardhini, Kumar, Bimal ORCID logoORCID: https://orcid.org/0000-0002-2539-4902, Robinson, Melanie and West, Graeme ORCID logoORCID: https://orcid.org/0000-0003-0884-6070;
CORE (COnnecting REpositories)