Hybrid NLP-based extraction method to develop a knowledge graph for rock tunnel support design

In the realm of drill and blast (D&B) tunneling, the design of tunnel support emerges as a pivotal concern. As a typical knowledge-intensive task, the practical application of tunnel support design often relies extensively on experienced engineers or experts who formulate designs based on their expertise. Compared with data-driven approaches, knowledge-driven tunnel support design allows for a nuanced understanding of the intricate geological and engineering factors influencing support design, and yield results that are more transparent and explainable. However, in practice, the substantial knowledge sources and complex interrelationships between factors which have direct or indirect influences on support design hinder the implementation of knowledge-based support design methods. To solve such knowledge gap, this study proposed a hybrid natural language processing (NLP)-based method to develop a knowledge graph for rock tunnel design. Specifically, rule-based methods are used to extract entities and classification relationships from standards and specifications, and statistical and artificial intelligence (AI)-based methods are used for extracting entities and non-classification relationships from a large amount of scientific literature, academic papers, and support design schemes. A total of 947 entities, 3 kinds of classification relationships and 11 kinds of non-classification relationships were extracted. On the basis of extracted entities, relations, and attributes, a knowledge graph was developed using ontology-based method, providing functionalities such as knowledge element retrieval, semantic retrieval, and query expansion. The findings of this study are expected to provide practical recommendations for the design of the tunnel support and advance the existing knowledge about the tunnel design from a knowledge-driven perspective.