Developing a hybrid approach to extract constraints related information for constraint management.

Construction projects face various constraints (e.g., materials and equipment). Constraint management approaches such as advanced working packaging (AWP) can remove constraints and ensure smooth work. However, due to inefficient information extraction, the prerequisite of AWP, i.e., identifying and modelling constraints, are performed manually. Efforts that integrate constraint information into project knowledge bases are also limited. This paper proposes a hybrid approach to automatically extract and integrate constraint information from texts. The approach combines a deep learning model with pre-defined rules. The model extracts constraint entities whereas rules created based on domain knowledge are used to establish relations between these entities. Extracted information is encoded into the original ontologies. The approach can extract both entities and relations with over 90% accuracy. The original ontologies can be successfully enriched and support semantic queries. The approach improves AWP by partially automating constraint identification and modelling as well as ontology development for information integration. • A deep-learning model is trained to extract constraints related entities from project documents • Domain knowledge is used to create rules to setup relations among extracted entities • Extracted entities and relations are encoded into domain ontologies to enrich its contents • The approach partially automates identification/modelling for advanced working packaging • The approach also partially automates ontology development [ABSTRACT FROM AUTHOR]