Zhang, Zijing; (2024) Exploring ambiguity in healthcare design requirements to enable automated code compliance checking. Doctoral thesis (Ph.D), UCL (University College London).

Text PhD thesis-Zijing-final submission.pdf - Accepted Version Access restricted to UCL open access staff until 1 May 2026. Download (5MB)

Abstract

In the architectural, engineering and construction (AEC) industry, Automated Compliance Checking (ACC) has been studied to address the inefficiencies of manual compliance checking. Although many ACC systems have been developed, a bottleneck still exists in developing a suitable machine-readable representation that can capture the full meaning and cover all types of building requirements, especially those with ambiguity. Using healthcare design requirements as an example, this research aims to identify the capabilities required and desired for a sufficiently expressive rule representation method to cover requirements’ full meaning, and systematically explore how ambiguity can be classified, automatically detected and managed. A literature review was conducted to understand the recent research, practice and current processes of ACC, ambiguity, text classification in the AEC industry, and research gaps. Building requirements of different constraint levels and scopes were analysed inductively to identify the required and desirable capabilities. An ambiguity taxonomy was developed based on qualitative, inductive analysis of healthcare design requirements that were reported to have ambiguity. A supervised machine learning approach was proposed to automatically detect ambiguity. Strategies were proposed to manage different types of ambiguity based on experts’ inputs from a focus group. The results show that 17 capabilities are required and two are desired to represent the full meaning of healthcare design requirements. The ambiguity taxonomy includes intentional and unintentional ambiguity, where three main categories of unintentional ambiguity was identified, including ambiguity related to the use of language, tacit knowledge, and the ACC domain. The automatic ambiguity type detection method based on Bidirectional Encoder Representations from Transformers (BERT) achieved satisfying performance on the validation dataset, showing the feasibility of such a method. Based on experts’ comments and suggestions, strategies were proposed to manage the main types of ambiguity in regulation drafting, rule interpretation and design modelling stages. This research contributes to the body of knowledge by identifying the most comprehensive set of capabilities required and desired for rule representations for ACC, developing the first taxonomy of ambiguity in healthcare building requirements, proposing a method to automatically detect four main types of ambiguity, and suggesting strategies to manage ambiguity and enable better ACC. Regulators, designers and researchers can benefit from the findings and achieve better ACC performance and compliance outcomes.

Download activity - last month

Export as XMLCSVJSON

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as JSONCSVXML

Archive Staff Only

View Item