Afzal, Muneeb (2024) Improving request for information (RFI) processing in construction projects using natural language processing (NLP) techniques. PhD thesis, University of Technology Sydney, Australia.
Abstract
Request for information (RFI) document is an essential communication tool to seek clarifications across the project lifecycle. Preparing, evaluating, and responding to the RFIs consume resources such as time and cost. Moreover, a burst of RFIs, unresolved RFIs, or slow responses to RFIs may lead to project risks such as schedule delay and cost escalation. Finding effective ways of managing these risks associated with RFI documents can reduce their frequency and turnaround time, minimising their adverse impacts on projects. Previous research has tried to categorise the content of the RFIs to identify the root causes and develop best practices; however, their classification methods primarily depend on manual content analysis, which is inherently labour-intensive and time-consuming. Such processes lack effectiveness, and their outcomes tend to be error-prone and biased. Moreover, there is no existing automated framework in the body of knowledge to analyse the unstructured text data in the RFI document. Therefore, this research aims to develop a mechanism for automated, efficient, and unbiased text classification and entity extraction for efficient information extraction from the unstructured RFI statements.Accordingly, this research leverages natural language processing (NLP) techniques to decode unstructured information within RFIs into easily understandable and actionable insights. The developed work consists of three primary models: (1) an NLP-based multiclass text classification model employing deep learning-based recurrent neural networks (RNNs) to categorise RFIs according to their project phase, supplemented with a topic modelling approach to visualise key topics and themes from the RFIs. (2) an NLP-based multiclass text classification model designed to identify the predominant issue within RFIs, and (3) an information extraction —named entity recognition (NER) model aimed at obtaining critical entities from RFIs. This comprehensive research enables the early detection of design, execution, procurement issues, and specific issues presented by an RFI, such as coordination, constructability, specification/scope, design/drawing discrepancies, and review/approval. The NER model facilitates the automated identification of problematic components, their locations, and relevant drawing references in an automated manner.With this wealth of information readily available, project stakeholders can optimise the RFI process, potentially shortening review periods and streamlining the prioritisation of RFIs. The efficiency of the developed models was assessed through experimental investigation and evaluation on datasets from actual construction projects. This evaluation focused on their information extraction and text classification abilities. The results demonstrated that the automated processing of RFIs through the NLP models not only outperformed human performance in terms of accuracy and efficiency but also showed strong performance on independent test sets. Specifically, the models achieved higher precision, recall and F-1 scores in identifying key entities and classifying issue types during RFI analysis. This strong performance refers to the models' ability to generalize well across unseen data, achieving consistent evaluation metrics further validating their effectiveness and reliability for real-world application in construction document analysis.The significance of this research lies in the development of multiple text classification models powered by NLP and deep learning pipelines, enabling stakeholders to gain meaningful insights from RFIs and efficiently resolve issues. It enhances construction communication workflows and lays the foundation for intelligent, real-time RFI analysis. Furthermore, the research offers a detailed implementation roadmap, guiding researchers and practitioners on how to integrate these models into real-world construction settings. This roadmap serves as a practical tool to elevate current RFI management practices and improve the quality of future project documentation.
| Item Type: | Thesis (Doctoral) |
|---|---|
| Thesis advisor: | Wong, Johnny Kwok-Wai and Fini, Alireza Ahmadian Fard |
| Uncontrolled Keywords: | accuracy; communication; constructability; content analysis; coordination; document analysis; documentation; learning; lifecycle; reliability; specification |
| Index terms: | specification, approval, documentation, accuracy, content analysis, neural network, drawing, project documentation, practitioner, documents, project stakeholder, construction project, management practice, cost escalation, workflow, dataset, body of knowledge, pipeline, efficiency, schedule delay, deep learning, document analysis, streamlining, project lifecycle, investigation, lifecycle, effectiveness, human-performance, best practice, coordination, constructability, modelling, implementation |
| Subjects: | infrastructure and transport systems, data management, project controls, sociology, contractual condition, professional development, management, construction integration, performance management, economics, artificial intelligence, data science, data analysis and analytics, project delivery, technical documentation, data collection methods, business, production management, project planning, contractual arrangements, knowledge management, analytical methods, human factors and perception, practitioner, contractual role, project completion |
| Topics: | Roles and Professions, Stakeholder Management, Information Management, Research Practice, Business Strategy, Cost Management, Contract Administration, Time Control, Organizational Design, Digital Applications, Design Practice, Procurement, Engineering Principles, Project Management, Quality Management |
| Descriptive scope: | 3 PCA |
N.B. Descriptive scope is a count of how many of the five facets of empirical research are indicated by the words used in title, abstract and keywords. It is not intended as a judgement on the research; merely a count of the kind of word we would expect to indicate Phenomenon, Concepts, Theoretical framing, Empirical techniques, Analytical techniques. If all five are present, then a code of “5 PCTEA” will indicate this. If you feel the coding for this record is questionable, we welcome discussion around the terms we matched or the way we categorized them. The facet you would expect may not be coded, or a facet may be coded inappropriately. This can also bear on a larger question, of which facets should be treated as defining in construction management research. Please get in touch, and we will look at it. More details here