Unlocking Insights: Analysing Construction Issues in Request for Information (RFI) Documents with Text Mining and Visualisation

Request for Information (RFI) is an essential communication and decision support tool that assists project teams in identifying and resolving construction queries. RFI occurrences are common throughout the project lifecycle and predominantly comprise issues related to conflicting drawings and specifications, unclear requirements, vague contract documents or unexpected site conditions that inhibit project progress. RFIs are typically unstructured textual documents, and their manual content analysis for knowledge extraction is arduous and time-consuming. Previous research has successfully harnessed the potential of Natural Language Processing (NLP) and text mining to process unstructured text documents and extract useful information. While NLP and text mining approaches have been applied in different domains, their application in the construction industry is limited, particularly for analysing RFI datasets. Hence, the present research analyses RFIs and their query subjects through an unsupervised learning approach. Key contributions include the implementation of Latent Dirichlet Allocation (LDA), an unsupervised text-mining algorithm to identify predominant topics and themes to classify the issues discussed within RFIs. This analysis successfully identified and highlighted issues related to structural discrepancies, construction coordination, building fixtures, building specifications and construction drawings as prominent problems mentioned in the project RFIs. As exploratory research, the findings of this study aim to enhance the understanding of RFI issues and to inspire future investigations that can delve deeper into specific aspects of the RFI review process and motivate future studies, which efficiently dissect, and address issues related to RFIs.