期刊
JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT
卷 142, 期 12, 页码 -出版社
ASCE-AMER SOC CIVIL ENGINEERS
DOI: 10.1061/(ASCE)CO.1943-7862.0001199
关键词
Underground utility; Compliance checking; Natural language processing (NLP); Spatial reasoning; Geographical information system (GIS); Information technologies
资金
- National Science Foundation (NSF) [CMMI-1265895, CMMI-1462638]
- Directorate For Engineering
- Div Of Civil, Mechanical, & Manufact Inn [1462638] Funding Source: National Science Foundation
Underground utility incidents, such as utility conflicts and utility strikes, result in time and cost overruns in construction projects, property damages, environmental pollution, personnel injuries, and fatalities. A main cause of recurrent utility incidents is the noncompliance with the spatial configurations between utilities and their surroundings. Utility specifications usually contain textual descriptions of the spatial configurations. However, detection of spatial defects, according to the textual descriptions, is difficult and time consuming. This deficiency is because of the lack of spatial cognition in many rule-checking systems to process massive amounts of data. This study aims to automate utility compliance checking by integrating natural language processing (NLP) and spatial reasoning. NLP algorithm translates the textual descriptions of spatial configurations into computer-processable spatial rules. Spatial reasoning executes the extracted spatial rules following a logical order in a geographical information system (GIS) to identify noncompliance. The intellectual contribution of this study is twofold. First, complex spatial rules are retrieved automatically from textual data with their hierarchies classified, which provides the inputs and indicates the sequence of rule execution in spatial reasoning. Second, semantic spatial relations are modeled on the basis of their metric and topological implications, enabling the automatic execution of multiple spatial rules. Experiments were conducted to test this framework. The average precision, recall, and combination of the two (F-measure) achieved by the NLP algorithm for extracting spatial rules are 87.88%, 79.09%, and 83.25%, respectively. In addition, the spatial reasoning mechanism also was found to be a powerful tool for compliance checking under various scenarios.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据