Project-Based As-Needed Information Retrieval from Unstructured AEC Documents

Abstract

With the increasing complexity of architecture, engineering, and construction (AEC) projects and fast track execution of project works, written documents are becoming more and more important for project coordination, communication, and works control. Finding all the relevant information from unstructured construction documents is critical to various management tasks such as work planning, progress control, and claims. A framework is proposed in this research to retrieve project-wide as-needed information from AEC documents. Through this framework, improvement in the levels of precision and recall in the information retrieval process can be made effective through the use of a project-specific term dictionary and dependency grammar parsing information of textual documents. Their effectiveness is demonstrated through a series of experimental tests conducted on a real life building redevelopment project with different information retrieval and ranking strategies. The results and findings are presented in this paper along with discussion on the related issues on research and system development.