Group Recommendation Using External Follwee for Social Tv

Only available on StudyMode
  • Download(s) : 71
  • Published : December 16, 2012
Open Document
Text Preview
GROUP RECOMMENDATION
USING EXTERNAL FOLLOWEE FOR SOCIAL TV
XiaoyanWang1, Lifeng Sun1, ZhiWang1 and Da Meng2
1

Department of Computer Science and Technology, Tsinghua University, Beijing, China Department of Computer Science and Technology, Beijing University of Posts and Telecommunications, Beijing, China 1

muyushiok@gmail.com, 1sunlf@tsinghua.edu.cn, 1wangzhi04@mails.tsinghua.edu.cn, 2mengda0710@126.com

2

Abstract—Group recommendation plays a significant role in
Social TV systems, where online friends form into temporary
groups to enjoy watching video together and interact with each other. Online microblogging systems introduce the "following" relationship that reflects the common interests between users in a group and external representative followees outside the group. Traditional group recommendation only considers

internal group members’ preferences and their relationship. In our study, we measure the external followees’ impact on group interest and establish group preference model based on
external experts’ guidance for group recommendation. In
addition, we take advantage of the current watching video to improve context-aware recommendations. Experimental
results show that our solution works much better in situations of high group dynamic and inactive group members than
traditional approaches.
Keywords- Social Media, Group Recommendation, External
Expert, Context Filtering

I.

INTRODUCTION

The rapid development of user generated content (UGC) and
social network results in a large number of social media
content. Media content related with social connections
differs from traditional ones concerning that sources of social media tend to be generated from friends’ sharing and
recommender systems in social networks (SNS). User
interaction in SNS continues to strengthen, followed by
relationship expansion. Social TV brings a new trend of
combining video streaming service with social network
service, which has deeply influenced the way people produce
and consume video content [1]. Some examples include the
integration of Twitter updates during live video streaming [2] and Facebook applications that allow commenting while
watching video content. Users tend to watch videos in
groups instead of doing that alone [1]; as a result, the concept of group interactions in such systems has been boosted. Such group viewing provides great potential for users to find
videos that interest members in the group, namely, group
recommendation.
The composition of a group in such Social TV system
varies from relatives, friends, classmates, and colleagues to even online concept users (e.g., public home page). And
sizes of groups also vary from 3, 5 to 8 or even larger.
Determining a group decision among alternatives with
multiple attributes is a well-known problem that has been
investigated for decades. Generally, there are two proposed

solutions. One is to combine group members into a “pseudouser”, which can represent overall characteristics of the group [3]. The other is to merge recommendation results of
individuals [4]. Existing work made use of different group
decision strategies among which Average satisfaction, Least
Misery and Most Pleasure are the most popular ones [5]. All
methods are evaluated from the perspective of the content
within group members.
Relationship within group members is also taken into
account in group recommender system. Some measure user
status within a group by evaluating relation strength with
others. Superordinate users are even named “expert” and his characteristics can largely influence the preference of the
whole group [5]. Others study the dissimilarity among group
members [6], which also concern the content level. There are also some more studies on how each member contributes to
reaching group consensus [9]. However, the existing
approaches are facing the following problems. Firstly, group characteristics are affected not only by inside group...
tracking img