Schematization for the analysis of geolocated microblog messages

Abstract

An ample amount of research has been conducted on geolocated information in social media. Many researchers use more or less sophisticated analysis tools, from descriptive statistics to machine learning. While the decidedly geometric analyses have been relatively sparse, many papers field some form of map showing geographic distributions of social media messages from their respective studies. As one of the attractive points of social media analysis lies in the potentially very large number of individual messages, many of these distribution maps fail to convey more than a very rough impression of the geographic patterns involved. Furthermore, in the case of portraying links, visual clutter and overload is a common problem. We start with a short overview of currently employed visualiza-tion techniques. This is followed up by our proposal of using techniques from cartographic generalization and schematization to portray patterns and analysis results for geolocated microblogging research. Specifically, we show how network schematization and territorial outline schematization can be used to portray and analyse social links and land-use information generated from microblogging data respectively.

Full text

Schematization for the analysis of geolocated
microblog messages
A. Reimer and R. Westerholt ∗
March 15, 2014
Abstract
An ample amount of research has been conducted on geolocated infor-
mation in social media. Many researchers use more or less sophisticated
analysis tools, from descriptive statistics to machine learning. While the
decidedly geometric analyses have been relatively sparse, many papers
field some form of map showing geographic distributions of social media
messages from their respective studies. As one of the attractive points of
social media analysis lies in the potentially very large number of individ-
ual messages, many of these distribution maps fail to convey more than a
very rough impression of the geographic patterns involved. Furthermore,
in the case of portraying links, visual clutter and overload is a common
problem. We start with a short overview of currently employed visualiza-
tion techniques. This is followed up by our proposal of using techniques
from cartographic generalization and schematization to portray patterns
and analysis results for geolocated microblogging research. Specifically, we
show how network schematization and territorial outline schematization
can be used to portray and analyse social links and land-use information
generated from microblogging data respectively.
Keywords: Social Media, Data Mining, Schematization, General-
ization
∗Heidelberg University, Institute of Geography Dept. for GIScience
1