Intervening Construction Workers’ Unsafe Behaviour with a Chatbot

As an effective method to reduce the unsafe behavior of construction workers, safety training has always been the hotspot of safety management research. In recent years, while there is an ever-growing research interest on developing effective training techniques and methods, few studies have improved safety training with the targeted interactions with construction workers. Therefore, based on natural language processing technology, this paper introduced the chatbot into construction safety training and designed a framework for personalized construction worker safety training on mobile phones. In particular, the single-round question and answer technique with the chatbot was introduced with an illustrative example. Through word segmentation, part-of-speech tagging, similarity calculation, and threshold comparison, questions and sentences from regulations could be compared to determine which sentence should be chosen as the most matching answer, and to improve workers’ ability to work safely. In this way, this research provided an innovative, adaptive, convenient and knowledge-rich personalized safety training approach, in the hope of reducing cognitive difficulty and increasing learning interests of construction workers.