Information Technology Reference
In-Depth Information
by integrating the discovered short-term interest from Twitter,
T-Y Recommender
achieves slightly better performance than the other methods. Note that the shortage
in obtaining users' complete video-related activity, e.g., the video viewing data,
makes the F-score very low. Investigating into alternative user information toward
more promising evaluation strategy remains an open issue for the future personalized
services.
5.4 From User: Cross-Network Knowledge
Association Mining
In the above heterogeneous social multimedia data organization, for a target user,
his/her accounts on multiple social media networks are needed when conducting
complementary or collaborative user modeling. While, in practical applications, it is
difficult to get access to all the overlapped users' accounts. Moreover, with explicit
user account requirements, only those who participate into certain network can ben-
efit from the social multimedia data in this network. This requirements limits the
potentials of exploring heterogeneous social multimedia data.
Social multimedia data association aims to discover correlations between het-
erogeneous social knowledge,
4
which can then be applied to unseen users without
the multiple-account dependence. Analogous to cross-media applications whose key
problem is to mine the association between different modalities, social multime-
dia data association will contribute much to the exploration of heterogeneous social
multimedia data and further applications. Our basic idea is to investigate the over-
lapped users' collaborative involvement into heterogeneous social multimedia data
for association mining. The premise is related to crowded intelligence that,
if a set
of users collaboratively involve in social knowledge A on one network, and social
knowledge B on another network, we are confident to associate knowledge A and B
for the unseen users
.
When realizing the idea, the selection of social knowledge is expected to reflect
the characteristics or meet the demands of the examined social media networks.
With focuses still on Twitter and YouTube, two facts are: (1) YouTube has an obvious
demands for video promotion
5
; and (2) Twitter is efficient in information propagation
and has grown as the top referrer for web video discovery.
6
Therefore, we select
the social knowledge as the video-related activity on YouTube and the following
relation on Twitter. Association between the heterogeneous social knowledge will
address a cross-network collaborative video promotion problem, i.e., identifying the
4
Social knowledge indicates a typical pattern in users' social relation or social activity data, e.g.,
the SNS patterns in Facebook, the video watching patterns in YouTube, and the consuming patterns
in Amazon
5
YouTube has started to let video content providers be partners to cash in on the videos posted by
sharing ad revenue and charging rental fees to viewers.
