Information Technology Reference
In-Depth Information
person. For privacy and security concerns, Arvind et al. first addressed the user link-
age mining problem [
20
]. A two-step de-anonymizing framework is developed to
identify user accounts between Flickr and Twitter. In [
14
], the authors studied the
tagging activity of overlapped users on different social media networks, and proposed
to employ username and tag history for user account linkage. Liu et al. furthered the
previous work by differentiating users with the same usernames. Based on data obser-
vations, the user linkage mining task is casted as a pairwise classification problem,
which is solved by a novel unsupervised approach. Other user linkage mining work
include [
5
,
9
,
31
], where network topology as well as user activities are employed
with advanced machine learning and social network analysis algorithms. The sat-
isfied performance of state-of-the-art user linkage mining solutions have provided
large-scale overlapped user accounts and paved the way for micro cross-network
analysis.
Current micro cross-network analysis work is mainly devoted to examining the
overlapped users' activity and relation patterns among different social media net-
works. Abel et al. analyzed the same user's tagging activities on Flickr, Twitter, and
Delicious, and then addressed the cold-start recommendation problem by
comple-
mentary organization
-based user modeling [
1
]. In our previous work, we examined
the overlapped users' social relation patterns between Twitter and Flickr, based on
what a cross-network friend recommendation solution is developed [
29
]. In [
27
],
with concern on personal privacy protection, users' privacy settings across Face-
book, Twitter, and Foursquare are examined. In [
30
], an urban lifestyle spectrum is
built by analyzing large-scale cross-network user activities. Recently, an interesting
work is introduced to examine the novelty-seeking traits in check-in networks and
online shopping networks [
32
]. In this chapter, emphasizing the bridge role of the
overlapped users, we categorize micro cross-network analysis into two schemes of
organization and association. Novel application paradigms are presented with the
introduction of two practical problems.
Note that Zhong et al. have also conducted extensive work to employ overlapped
users to facilitate applications like link prediction and interest recommendation
[
34
,
35
]. However, in their work, instead of different social media networks, the
overlapped users actually serve as bridges between internal networks, e.g., Twitter
retweet and mention networks. Moreover, they focused the cross-network social
knowledge on network topologies, while we aim to explore the heterogeneity in both
social relation and social activities.
5.3 On User: Cross-Network User Modeling
Under social media circumstances, user data distribute among various social media
networks, which need to be jointly analyzed toward comprehensive user understand-
ing and personalized social media services. This calls for the necessity and demon-
strates the reasonability to organize the heterogeneous social multimedia data along
the overlapped users.
