Semantic Rule-Based Construction Procedural Information Extraction to Guide Jobsite Sensing and Monitoring

Abstract

Existing construction monitoring systems rely on designated personnel to compare a planned procedure in the instructional documents with an execution procedure on the jobsite. It requires major human efforts and is time-consuming, costly, and human error–prone. To reduce manual efforts in collecting information from construction procedural documents, selecting appropriate sensing techniques to collect data on the jobsite, and giving in-time feedback for progress monitoring and compliance checking, the authors: (1) proposed a semantic rule-based information extraction (IE) method to extract construction execution steps from construction procedural documents automatically; (2) developed a construction procedure and data collection (CPDC) ontology to classify construction site information and provide guidance on selecting sensing techniques for collecting jobsite data based on the extracted information; and (3) proposed a construction procedural data integration (CPDI) framework, which could integrate textual data and sensing data to conduct construction execution steps compliance checking automatically. The proposed IE method offers a novel way to retrieve construction activities (e.g., install door and prepare surface) from the construction procedural documents that are an important source of information but were seldom analyzed in previous research. It focuses on extracting construction execution steps, which supports the incorporation of execution sequence information into construction jobsite management. A novel information classification application based on a newly developed CPDC ontology is provided to support IE. The IE results provide a new instrument for selecting sensing techniques based on different construction site data categories. The proposed method is developed to support a targeted unified CPDI framework that integrates the following: natural language processing (NLP) and sensing techniques to automatically extract, analyze, and process both planned procedures from textual data from construction procedural documents and executed procedures from sensing data of construction jobsites. It also can be extended to any domain applications that contain textual information and site activity information that needs to be matched or compared. In this paper, the authors focus on the NLP-based textual information extraction of the construction procedural documents, introduced the detailed steps involved, and proposed a CPDI framework on top of and as an application of the IE method. An experiment was conducted to evaluate the IE method with a set of open-source specifications. Comparing it to a manually developed gold standard, 97.08% precision and 93.23% recall were achieved using the proposed IE method for the extraction of construction execution steps. A qualitative analysis on sensing technique selection based on the IE results was also performed. The proposed IE method can be applied to the automation of construction site management tasks (e.g., construction monitoring) that integrates both textual information and sensing data to support construction decision-makings. It enables the transformation of construction monitoring/control from a human-intensive process to an automated one.