NLP-based regulatory compliance – using GPT 4.0 to decode regulatory documents

Tools

Kumar, Bimal and Roussinov, Dmitri (2024) NLP-based regulatory compliance – using GPT 4.0 to decode regulatory documents. In: Georg Nemetschek Institute Symposium & Expo on Artificial Intelligence for the Built World, 2024-09-10 - 2024-09-12.

[thumbnail of Kumar-Roussinov-GNIS-2024-NLP-based-regulatory-compliance-using-GPT-4-0-to-decode] Text. Filename: Kumar-Roussinov-GNIS-2024-NLP-based-regulatory-compliance-using-GPT-4-0-to-decode.pdf
Accepted Author Manuscript
Restricted to Repository staff only until 1 January 2099.
Download (753kB) | Request a copy

Abstract

The well-publicised Hackitt Review (Hackitt, 2018) into the Grenfell Tower disaster points to some of the main reasons behind the ill-fated fire in the London blocks of flats in 2017. To avert similar disasters in future, Dame Hackitt recommended that the building regulations and associated guidance, including the Approved Documents in the UK, need to be authored, applied and enforced in a fundamentally different way: “[...] the current regulatory system for ensuring fire safety in high-rise and complex buildings is not fit for purpose.” In the UK, currently there are some 485 standards, 85 other government guidance, 176 industry guidance, and 79 other government legislation documents (MHCLG, 2020) that a building design must comply with. The corpus suffers from various challenges that include inconsistent use of terms, ambiguities, contradictions and lack of clarity necessitating interpretations of requirements.

ORCID iDs

Kumar, Bimal and Roussinov, Dmitri ORCID logoORCID: https://orcid.org/0000-0002-9313-2234;
CORE (COnnecting REpositories)