Automated Prioritization of Requirements to Support Risk-Based Construction Inspection of Highway Projects Using LSTM Neural Network

Construction inspection is a crucial stage that verifies if a construction project has met all contractual requirements. However, the lack of resources due to the budget reduction for maintaining the large aging transportation network has greatly affected the construction inspection capabilities. There is an emerging need for risk-based construction inspection practices that can help highway agencies optimize the use of limited resources without compromising inspection quality. Automated prioritization of requirements according to their criticality would be extremely helpful since contractual requirements are typically presented in an unstructured natural language in voluminous text documents. The current study introduces a novel model for predicting the risk level of contractual requirements using the Long Short-Term Memory (LSTM) neural network. The train data include sequences of requirement texts which were manually rated with severity level by industry professionals. For feature extraction, the study used word embeddings. The developed requirement risk prediction model was evaluated using root mean squared error. The proposed model is expected to provide construction inspectors with a means for the automated classification of voluminous requirements by their importance, thus help to maximize the effectiveness of inspection activities under resource constraints.