Abstract

Building codes are composed of a set of requirements that govern the design, construction, and maintenance of buildings and structures. Currently, most building codes are stored in unstructured text-based documents. However, with the advance of artificial intelligence, these unstructured building codes are no longer meet the requirements toward regulation intelligence scenarios such as design compliance review, relation analysis, and so on. To address this problem, this research proposes a method to automatically collect and formalize building codes and transform them into a knowledge graph representation. The method mainly consists of three steps: 1) data collection, which automatically collects building codes by crawling data in the web; 2) data structuring, which automatically transforms text-based building codes (e.g., HTML and PDF) into XML structure; and 3) graph generation, which transforms XML-based building codes into Neo4j graph database. The proposed method is implemented and tested in a case study. The result demonstrates the feasibility of the method and shows that the generated knowledge graph can support multiple regulation intelligence scenarios such as regulation relation retrieval, regulation conflict analysis, design compliance review, and so forth.

First author of this paper, Mr. Yu-Cheng Zhou shared our research in the ICCREM 2021 on October 16, 2021, and we also won the Best Paper Award of ICCREM 2021.

Download preprint here

Download paper here

The authors are grateful for the financial support received from the National Natural Science Foundation of China (No. 51908323, No. 72091512), and the Tsinghua University Initiative Scientific Research Program (No. 2019Z02UOT).

Financial Sources: