Geography Reference
In-Depth Information
Fig. 4 Chart showing the number of placenames found in MOOC forum posts
their lab assignments for the course. Urban geography was of particular interest, as
were topics centered on change of all types. One of the first discussion prompts in
the course asked students to consider spatial privacy concerns, which resulted in a
few related topics shown here in this category.
Geo-Parsing Forum Posts
We continued our exploration of text generated in the Maps MOOC by applying
text mining techniques to identify and geocode the place names mentioned by
students in forum posts. This general process is frequently called geo-parsing in
contemporary literature (Gelernter and Zhang
2013
). To analyze our course forum
dataset of over 95,000 posts, we used the GeoTXT.org service. GeoTXT combines
named-entity recognition methods with geocoding capabilities in order to extract
and locate placenames found in text (Karimzadeh et al.
2013
) (Fig.
4
).
Processing the forum post dataset with GeoTXT resulted in the extraction of
38,317 places from 95,958 posts. There were 21,357 posts that included at least one
place mention. Some posts referenced multiple placenames; an overview of which
can be found in Fig.
3
.
After extracting and geocoding the placenames mentioned in class discussion
forum posts, we sought ways to map these data in order to understand the frequency
and distribution of placename mentions. First, we aggregated all placenames found
into hexagons (2
wide at the Equator), in order to explore the overall global density
of place-oriented discussion in the class. Figure
5
shows the resulting map, which
looks much like a population density map of the Earth, with some notable excep-
tions in China and parts of Africa.
