Database Reference
In-Depth Information
similarities between geo-social data and mobility phone data, as explained in
Section
16.4.2
, the conceptual framework and the characteristics of geo-social
data lead to a real new branch of research. The research about this new domain
is far from being exhaustive. As described in Section
16.3
, trajectories resulting
from geo-social data are built from a collection of sparse data points. This ends
up in different groups of applications, as described here.
We can distinguish a first group of applications that use only the location
from geo-social data, generally to filter the contents (message, photo, video,
news, tweets, and so on) from a zone they want to analyze or about which
they want to receive alerts (newsfeed mechanism). Some examples from the
natural disaster field include wild fires in the United States and France, hur-
ricanes in the United States, the 2010 earthquake in Haiti, and floods in the
United Kingdom, while an example from social-political field is the Arabic
revolutions started in late 2010. In all these cases, messages were filtered using
the related location such as coordinates, user location settings, or place names
in text or tags. The impact of (geo)social media during crisis events has been
shown to have high value for relief workers or coordinators and the affected
population.
Another group of applications uses the set of places to discover patterns.
An example is the tourism knowledge scenario. In Web 2.0 communities, peo-
ple share their traveling experience in blogs and forums. These articles, named
travelogues, contain various tourism-related information, including text depic-
tion of landmarks, photos of attractions, and so on. Travelogue provides an
abundant data source to extract tourism-related knowledge. Travelogues can be
exploited to generate location overviews in the form of both visual and textual
descriptions. The method consists first in mining a set of location-representative
keywords from travelogues, and then in retrieving web images using the learned
keywords. The model learns the word-topic (local and global tourism topic,
such as an attraction sight) distribution of travelogue documents and identifies
representative keywords within a given location. Complementing travelogues,
geo-referenced photos also tell a great deal about tourism knowledge. The pho-
tos, together with their time- and geo-references, implicitly document the pho-
tographer's spatio/temporal movement paths. The tourist-visited points can be
grouped, mined to distinguish patterns, and used to rank places of interest and
generate recommendations. In most of these cases, applications use location
extracted from human trajectories in the real world, but they are not really using
the trajectories itself.
A third group of applications also considers the users' interactions and rela-
tionships. In fact, geo-social networks provide not only the location, but also
the explicit social links, and in some cases explicit declaration of kinships
and partnerships, giving the possibility to overcome the shortcomings of tech-
niques to infer tie strength. They also give high-resolution location data, as one
