Building Dynamic Ontological Models for Place using Social Media Data from Twitter and Sina Weibo

Place holds human thoughts and experiences. Space is defined with geometric measurement and coordinate systems. Social media served as the connection between place and space. In this study, we use social media data (Twitter, Weibo) to build a dynamic ontological model in two separate areas: Beijing, China and San Diego, the U.S.A. Three spatial analytics methods are utilized to generate the place name ontology: 1) Kernel Density Estimation (KDE); 2) Dynamic Method Density-based spatial clustering of applications with noise (DBSCAN); 3) hierarchal clustering. We identified feature types of place name ontologies from geotagged social media data and classified them by comparing their default search radius of KDE of geo-tagged points. By tracing the seasonal changes of highly dynamic non-administrative places, seasonal variation patterns were observed, which illustrates the dynamic changes in place ontology caused by the change in human activities and conversation over time and space. We also investigate the semantic meaning of each place name by examining Pointwise Mutual Information (PMI) scores and word clouds. The major contribution of this research is to link and analyze the associations between place, space, and their attributes in the field of geography. Researchers can use crowd-sourced data to study the ontology of places rather than relying on traditional gazetteers. The dynamic ontology in this research can provide bright insight into urban planning and re-zoning and other related industries.