Extending Building Information Models Semi-Automatically Using Semantic Natural Language Processing Techniques

Abstract

Automated compliance checking (ACC) of building designs requires automated extraction of information from building information models (BIMs). However, current Industry Foundation Classes (IFC)-based BIM models provide limited support for ACC, because they lack the necessary information that is needed to perform compliance checking (CC). In this paper, we are proposing a new approach for extending the IFC schema to incorporate CC-related information, in a semi-automated and objective manner. Our method utilizes semantic natural language processing (NLP) techniques to extract concepts and relations from documents that are related to CC (e.g. building codes). We utilize pattern-matching-based rules for the extraction. We use three types of features in the matching patterns: part-of-speech tags, dependency relations, and term sequence numbers in a sentence. The selected concepts and relations are then automatically encoded into the EXPRESS-language-represented IFC schema. The automated encoding in EXPRESS is enabled using a set of mapping rules. To evaluate our proposed approach, we compared the concepts and relations that we automatically extracted from the International Building Code 2006 to extend the IFC schema with a manually-developed gold-standard, and evaluated the results in terms of precision and recall. We achieved higher than 90% precision and recall, which shows that our approach is promising.