TY - GEN
T1 - A Machine-Learning Approach for Semantic Matching of Building Codes and Building Information Models (BIMs) for Supporting Automated Code Checking
AU - Zhang, Ruichuan
AU - El-Gohary, Nora
N1 - Funding Information:
Acknowledgments. The authors would like to thank the National Science Foundation (NSF). This material is based on work supported by the NSF under Grant No. 1827733. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the NSF.
Publisher Copyright:
© Springer Nature Switzerland AG 2020.
PY - 2020
Y1 - 2020
N2 - Various automated code compliance checking (ACC) systems have been developed and used to check the compliance of building information models (BIMs) with building codes, to reduce the time, cost, and errors of the code compliance checking process. All these systems require some form of code-BIM matching – matching of the concept representations in the codes to those in the BIMs – which is a difficult task. Traditionally, semantic matching was conducted in a highly-manual manner. To address this problem, more recently, a limited number of efforts have proposed fully automated semantic matching methods, which mostly rely on matching annotations and/or rules developed by domain experts. Despite their relatively good performance, these methods are by nature difficult to generalize or scale up (e.g., the matching rules need to be updated, modified, or extended when switching from one type of code to another). There is, thus, a need for semantic matching approaches that are more generalizable and scalable. To address this need, this paper proposes a new, machine learning-based approach to automatically match the building-code concepts and relations to their equivalent concepts and relations in the Industry Foundation Classes (IFC). The proposed approach consists of five primary tasks: (1) prepare and process the training and testing data; (2) automatically identify the domain word embeddings by learning from a large corpus of building-code text and generate the final semantic representations by combining the domain and general word embeddings; (3) match the building-code concepts to the IFC elements; (4) match the building-code relations to the IFC relations; and (5) evaluate the performance of the proposed approach using accuracy. The proposed approach was implemented and tested on a number of chapters from the 2009 International Building Code (IBC) and the Champaign 2015 IBC Amendments. The preliminary results show that the proposed approach achieved an accuracy of 77% for matching building-code concepts to IFC elements, and 78% for matching building-code relations to IFC relations, indicating promising semantic matching performance.
AB - Various automated code compliance checking (ACC) systems have been developed and used to check the compliance of building information models (BIMs) with building codes, to reduce the time, cost, and errors of the code compliance checking process. All these systems require some form of code-BIM matching – matching of the concept representations in the codes to those in the BIMs – which is a difficult task. Traditionally, semantic matching was conducted in a highly-manual manner. To address this problem, more recently, a limited number of efforts have proposed fully automated semantic matching methods, which mostly rely on matching annotations and/or rules developed by domain experts. Despite their relatively good performance, these methods are by nature difficult to generalize or scale up (e.g., the matching rules need to be updated, modified, or extended when switching from one type of code to another). There is, thus, a need for semantic matching approaches that are more generalizable and scalable. To address this need, this paper proposes a new, machine learning-based approach to automatically match the building-code concepts and relations to their equivalent concepts and relations in the Industry Foundation Classes (IFC). The proposed approach consists of five primary tasks: (1) prepare and process the training and testing data; (2) automatically identify the domain word embeddings by learning from a large corpus of building-code text and generate the final semantic representations by combining the domain and general word embeddings; (3) match the building-code concepts to the IFC elements; (4) match the building-code relations to the IFC relations; and (5) evaluate the performance of the proposed approach using accuracy. The proposed approach was implemented and tested on a number of chapters from the 2009 International Building Code (IBC) and the Champaign 2015 IBC Amendments. The preliminary results show that the proposed approach achieved an accuracy of 77% for matching building-code concepts to IFC elements, and 78% for matching building-code relations to IFC relations, indicating promising semantic matching performance.
UR - http://www.scopus.com/inward/record.url?scp=85092049869&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85092049869&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-34216-6_5
DO - 10.1007/978-3-030-34216-6_5
M3 - Conference contribution
AN - SCOPUS:85092049869
SN - 9783030342159
T3 - Sustainable Civil Infrastructures
SP - 64
EP - 73
BT - Recent Research in Sustainable Structures - Proceedings of the 3rd GeoMEast International Congress and Exhibition, Egypt 2019 on Sustainable Civil Infrastructures – The Official International Congress of the Soil-Structure Interaction Group in Egypt SSIGE
A2 - Rodrigues, Hugo
A2 - Morcous, George
A2 - Shehata, Mohamed
PB - Springer
T2 - 3rd GeoMEast International Congress and Exhibition on Sustainable Civil Infrastructures, GeoMEast 2019
Y2 - 10 November 2019 through 15 November 2019
ER -