Multi-Modal Knowledge Graph-Based Question Answering System for High-Rise Building Construction Technologies

Abstract

In the modern construction industry, the transmission of construction technologies still largely relies on traditional paper drawings and oral communications among workers, which fosters inaccuracies in information transfer, low efficiency in process guidance, and increased construction safety risks. For high-rise buildings particularly, the complexity of construction technologies, the difficulty of the tasks, and the low margin for error demand a higher need for construction technology processing and knowledge management. Therefore, this paper proposes a multi-modal knowledge graph-based Question Answering (QA) system for high-rise building construction technologies, aimed at enhancing the efficiency and accuracy of construction technology information retrieval during high-rise building construction. This system employs deep learning and natural language processing technologies, integrating various data types including texts, images, videos, and three-dimensional models, enabling it to comprehend and respond to complex queries and questions from engineering personnel regarding high-rise building construction. It also provides guidance supported by images, videos, and model parameters. Through testing the system, its efficacy and accuracy in handling real-world engineering problems were validated, particularly excelling in areas such as engineering process selection, construction methods, and safety standard guidance. This not only promotes the systematization of construction technology knowledge within the building field but also enhances the quality and efficiency of engineering construction.