| 성과구성 |
ABSTRACT:
This paper aims to extract a predicate-argument structure from given building design require ment for automate building design rule-checking. Ruic-making that e<>ver.; the transl at ion of natural language into computer readable format is a ke y process of rule-checking. Previous studies for rule-making focused on
trai1sla tion mechanismor useful interfacesuch as visual programing language. However, a ta.sk of rule-making
•• been manually "°':dt'.cted and depends on pro얻a mmcrand rule e.xper.ts. Thcr':!' '.e,this pa er d,scribesan deep learning-ha,핵 design regulation analysis for automation in rule making. This paper cla.ssifies logical ele ments of dt>Sig n r여 uircmcnts and their semantic roles. Prcd1cate-argument structure for rule checking is
defined. lJsmg a bidirectional LSTM with condit10nal random field model is tram여 to rccog111ze o bj ect, pro perty and method that arc compone.nts of rule.-checkmg rulesel.
Key Word.: Automated rule checking, Building information m(저eli ng, Deep learning, Design requirements, L-0eic rule comoonc.nt |
