Databases Reference
In-Depth Information
other users on new types of Web applications called Social Web platforms such as Delicious
11
, Flickr
12
,
Facebook
13
, and YouTube
14
.
Leveraging useful data from the large amount of user-contributed data available in the Social Web
represents a challenging topic which opens new opportunities for recommender system research [Guy
et al., 2010a; Freyne et al., 2011]. For example, one can think of a recommender that develops and
maintains a user profile by analyzing the blog entries the user has commented on recently [Turdakov,
2007]. Such an approach can help to alleviate the cold-start problem of recommender systems. Or think,
for example, of a recommender that analyzes the relationships in a social network such as Facebook to
compute a user's trust in other users. In the context of recommender systems, trust can serve as a new
measure of user-similarity [Golbeck, 2009]. Usually, social networks analysis (SNA) methods are used
to analyze the huge amount of data available in a social network [Scott, 2000]. SNA methods can, for
example, be used to analyze the role of a particular user in a social network and to determine the existing
user clusters. Different works exist in the literature that try to explore SNA methods for recommender
systems (see, for example, [Ting et al., 2012] and [He and Chu, 2010]).
Tags are today a popular means for users to organize and retrieve items of interest in the Social Web
[Vig et al., 2009]. As the application areas of tags are manifold, they play an increasingly important role
in the Social Web. They can be used to categorize items, express preferences about items, retrieve items
of interest, and so on. For a detailed description of the main purposes of tagging the reader is referred
to a textbook specifically focusing on tagging systems [Peters and Becker, 2009].
Collaborative tagging or social tagging describes the practice of collaboratively annotating items with
freely chosen tags [Golder and Huberman, 2006] which plays an important role in sharing content in the
Social Web [Ji et al., 2007]. In a social tagging system such as Delicious and Flickr users typically create
new content (items), assign tags to these items, and share them with other users [Cantador et al., 2010].
The result of social tagging is a complex network of interrelated users, items, and tags often referred to as
a community-created
folksonomy
[Mathes, 2004]. The term folksonomy is a neologism introduced by the
information architect Thomas Vander Wal and is composed of the terms
folk
as in people and
taxonomy
which stands for the practice and science of classification [Wal, 2007]. In contrast to typical taxonomies
such as formal Semantic Web ontologies, social tagging represents a more light-weight approach, which
does not rely on a pre-defined set of concepts and terms that can be used for annotation. For a detailed
comparison of both approaches the reader is referred to [Shirky, 2005] and [Ji et al., 2007]. A formal
definition of a folksonomy is provided in Chapter 3.
Tags also began to gain importance in the field of recommender systems. User-generated tags not only
convey additional information about the items, they also tell something about the user. For example,
if two users use the same set of tags to describe an item, we can assume a certain degree of similarity
between those. Therefore, tagging data can be used to augment the basic user-item rating matrix.
Recently, several works have been proposed in the literature concerning the topic of leveraging user
contributed data available in the Social Web for recommender systems [Guy et al., 2010a; Freyne et al.,
2011]. In this work, we pursue this line of research and present novel recommendation approaches. In
particular, we will show how tagging data can be utilized for recommender systems. In the following, we
will present a possible categorization of tag-based recommender systems in the literature.
2.3.1 Using tags as content
Maybe the easiest way to use tagging data for recommender systems is to consider tagging data as an
additional source of content. Several works exist that view tags as content descriptors for content-based
systems, see, for example, in [Firan et al., 2007; Li et al., 2008] or [Vatturi et al., 2008]. In Section 2.1.3,
we have seen that an item's content description can be represented by a set of keywords. Since tags can
be seen as user-provided keywords, the content-based recommendation scheme described in Section 2.1.3
can be applied without any modification.
Similarly, in [de Gemmis et al., 2008], tagging data is used for an existing content-based recommender
system in order to increase the overall pr
edictive accuracy of the system. Machine learning techniques
11
http://www.delicious.com
12
http://www.flickr.com
13
http://www.facebook.com
14
http://www.youtube.com
