Information Technology Reference
In-Depth Information
Based on the overlapped users, one organization scheme is called
complementary
organization
, which directly aggregates the heterogeneous social multimedia data
around the same user. User understanding is devoted to obtaining various usermodels,
e.g., demographic model, SNS model, LBS model, and consuming model. User
data on different social media networks reflect user status, interest and preference
from different perspectives. Take health model for example, the heterogeneous social
multimedia data may include the number of steps you walk tracked by
F
itbit, how
often you check in to local gym using
Foursquare
, and what you eat based on the
pictures of your meals that you post on
Instagram
. Each piece of information, by
itself, may be inconsequential. However, organized and aggregated by the overlapped
users, the heterogeneous social health data will make up for the shortage of physical
health records, and significantly facilitate health insurance and smart healthcare.
Following this scheme, we have aggregated the user profiles from Google+ and
YouTube to facilitate user modeling and apply for video recommendation [
8
].
Another organization scheme is called
collaborative organization
, where collabo-
rative characteristics between heterogeneous social multimedia data of the same user
are explored to assist personalized services. Along this scheme, we have examined
the temporal characteristics of overlapped users' social activities between Twitter and
YouTube. With observation that user response on Twitter is faster than on YouTube,
we develop a real-time personalized YouTube video recommender by integrating the
auxiliary knowledge from Twitter, which will be introduced in the following two
subsections.
5.3.1 Data Analysis
Twitter has been recognized as an efficient platform for information sharing and
spread. Many breaking news is now first reported on Twitter even before mainstream
medias, e.g., Osama Bin Laden's death and the Hudson River plane crash [
13
]. At
the macro level, Twitter also responses and spreads faster than other types of social
media networks, such as wiki, blog, and media sharing web sites [
16
,
21
]. In this
work, we first extend this observation to the micro level and examine whether the
conclusion applies to the overlapped users.
We have collected 15,000 users from
About.me
2
who provide their user accounts
on both Twitter and YouTube. We further constrain that users have conducted more
than 10 social activities on both networks, resulting in the final dataset with 8,518
overlapped users, 8M Twitter social activities and 0.6M YouTube social activities.
3
To examine the collaborative characteristics between Twitter and YouTube, we
need first to identify some topics that widely spread over both networks in our dataset.
According to the trending topics of Google search in 2012, we retrieve the involved
2
A personal web hosting service that offers people a one-page profile to link multiple user accounts
from popular social media networks.
3
1M = 1 million.
