Linking context and proximity through web corpus

Proximity, a vague spatial relationship between a reference object (RO) and a located object (LO), has been shown to be influenced by many contextual factors rather than determined solely by the Euclidean distance between RO and LO. In this paper, a list of potential contextual factors that may influence how human conceptualize proximity was reviewed. Focusing on RO, LO, and geographical context we adopted a web-based corpus method by crawling the Web to build a spatial language corpus on linguistically specified proximity relationships. Spatial named entity recognition and entity resolution were used to extract RO, LO, and contextual factors out from the corpus. By mapping and analyzing the correlation between distance and contextual factors, this study provides insights on how different contextual factors influence the conceptualization of spatial proximity.