Location: PSH (Professor Stuart Hall Building) - 302,
Goldsmiths, University of London, Building 2
Campus Map
Abstract: Spatiotemporal analysis is essential for social media analytic applications, such as disease outbreak monitoring, marketing analysis, and business analytics. To understand the spatial and temporal distribution patterns of social media messages, researchers need to use geocode engines and space-time analysis methods to enhance their research models and analytic frameworks. This workshop will provide a good overview of geocoding methods for Twitter data by Dr. Ming-Hsiang Tsou and space-time analysis methods by Dr. Tao Cheng and Juntao Lai.
The geocoding method section will include various mapping approaches using geo-tagged tweets, user profile locations, place name extraction from texts, and the analysis of historical locations of individual users. We will also introduce several geocoding engines, including Google Map Geocoding API, Yahoo BOSS PlaceFinder, and OpenStreetMap Nominatim. Popular digital gazetters, such as GeoNames.org and the gazetteer of the Library of Congress (http://loc.gazetteer.us/) will also be discussed.
The section in space-time analysis methods will demonstrate how user interests and place profiles can be inferred from text harvested from geo-tagged Tweets. We will introduce an unsupervised topic modelling method, Latent Dirichlet Allocation (LDA) to extract meaningful topics from Tweets, and the clustering methods to generate the profile of places based upon the space-time patterns of these topics.
Pre-Workshop Prep:
- Participants should bring their own laptop computers and power cords for conducting web- based tutorials.
- Participants will need to access their own Twitter accounts and install R (preferable R studio) in their computers.
Workshop ContactMing-Hsiang Tsou - mtsou@mail.sdsu.edu
