Integrating semantic NLP and logic reasoning into a unified system for fully-automated code checking

• Published in: Automation in construction Vol. 73; pp. 45 - 57
• Main Authors: Zhang, Jiansong, El-Gohary, Nora M.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.01.2017; Elsevier BV
• Subjects: Automated code checking; Automated construction management systems; Automated information extraction; Automated reasoning; Automation; Building codes; Building information modeling (BIM); Building management systems; Compliance; Logic; Natural language processing; Reasoning; Semantic systems; Semantics; Automated reasoning; Automated information extraction; Semantic systems; Natural language processing; Logic; Automated construction management systems; Building information modeling (BIM); Automated code checking
• ISSN: 0926-5805; 1872-7891
• DOI: 10.1016/j.autcon.2016.08.027

Existing automated compliance checking (ACC) systems are limited in their automation; they rely on the use of hard-coded, proprietary rules for representing regulatory requirements, which requires major manual effort in extracting regulatory information from textual regulatory documents and coding these information into a rule format. To address this limitation, this paper proposes a new unified ACC system that integrates: (1) semantic natural language processing techniques and EXPRESS data-based techniques to automatically extract and transform both regulatory information (in regulatory documents) and design information [in building information models (BIMs)] for automated compliance reasoning, and (2) semantic logic-based information representation so that the reasoning could be fully automated. To test the proposed system, a BIM test case was checked for compliance with Chapter 19 of the International Building Code 2009. Comparing to a manually-developed gold standard, 98.7% recall and 87.6% precision in noncompliance detection were achieved. •A fully-automated approach to code compliance checking in construction is proposed.•Requirements are extracted from the code and formalized into rules, automatically.•Natural language processing & a logic representation enable such full automation.•Semantic transformation aligns design information with regulatory information.•A prototype achieved 98.7% recall & 87.6% precision in noncompliance detection.