A NOVEL APPROACH TO SOCIAL TAGGING: GROUPME!
Enhancing Social Tagging Systems with Groups
Fabian Abel, Nicola Henze and Daniel Krause
IVS – Semantic Web Group, Leibniz University Hannover, Appelstr. 4, 30167 Hannover, Germany
Keywords:
Social Media, Semantic Web, Tagging, Folksonomy, GroupMe!.
Abstract:
Common social tagging systems like Flickr, del.icio.us and others lately became very popular. The key benefits
of these systems include that users get involved in the content creation process, can easily — without overhead
– comment or annotate Web content, share resources with fellow users, and benefit from the comments /
annotations of other users with improved retrieval support. With GroupMe! we extend the idea of current
social tagging systems by enabling users to not only tag Web resources they are interested in, but also to create
collections (groups) of these Web resources by simple drag & drop operations. The grouping metaphor is
intuitive and easy for the users, and our evaluation shows that users appreciate the grouping facility, and use
this feature to organize and structure Web content. Technically, the grouping of resources carries valuable
information about Web resources and their relations, and can be exploited to improve the mining of Web
content, e.g. for search and retrieval.
1 INTRODUCTION
Popular systems like Flickr
1
, YouTube
2
, Blogger
3
or
others, which allow users to share photos, broadcast
own videos, or blog about topics they are interested
in, are obvious indicators for the success of Web 2.0.
These systems have shown that Web users are not ab-
solutely satisfied with their role of pure content con-
sumers. Instead, Web users want to contribute and
collaborate actively by providing their own content,
or by annotating content of other users. Web 2.0 web-
sites fulfill these needs perfectly: While the AJAX
technique (Garrett, 2005) enables websites to be in-
teractive, lots of Web 2.0 websites allow users to cre-
ate, upload or annotate content with free chosen key-
words (so-called tags) collaboratively, hence enable
intercreativity
4
. Tagging is one of the key factors
that makes these systems successful: It allows users
to annotate content with relevant keywords for future
retrieval. Especially multimedia content becomes in
this way searchable. With tagging, users create meta-
data collaboratively.
Another very well received feature of Web 2.0
1
http://www.flickr.com/
2
http://www.youtube.com/
3
http://www.blogger.com/
4
http://www.w3.org/Talks/9602seybold/slide6.htm
systems is the sharing of favorite resources with fel-
low users. E.g., in YouTube, users can store their
favorite videos, in BibSonomy
5
their favorite aca-
demic papers, and in del.icio.us
6
their favorite book-
marks. However, all of these systems are more or
less limited to a certain media type: Some systems
(like YouTube) support only one media type (videos),
while other systems, which can handle different kind
of media types, lack of an appropriate visualization
(in del.icio.us, e.g., all media types are displayed as
normal text links).
In this paper, we present the GroupMe! system
7
which offers a novel user interface to organize multi-
media Web resources. The core idea of the GroupMe!
approach is that users can group – via drag & drop
– the Web resources they are interested in. Appro-
priate media wrappers ensure that content of groups
is displayed in a concise manner. We report about
the evaluation of the systems which shows that a) the
GroupMe! group concept is very well accepted by the
users, b) that users like to combine resources of differ-
ent media types, and c) that these groups can be used
to improve search.
The paper is structured as follows: In Section 2
5
http://www.bibsonomy.org/
6
http://del.icio.us/
7
http://groupme.org/
42
Abel F., Henze N. and Krause D. (2008).
A NOVEL APPROACH TO SOCIAL TAGGING: GROUPME! - Enhancing Social Tagging Systems with Groups.
In Proceedings of the Fourth International Conference on Web Information Systems and Technologies, pages 42-49
DOI: 10.5220/0001522000420049
Copyright
c
 SciTePress

Figure 1: Screenshot of GroupMe! application: A user drags a photo from the left-hand side Flickr search bar into the
GroupMe! group on the right-hand side.
we describe our GroupMe! system which is evaluated
in Section 3. In Section 4 we compare the GroupMe!
system with other state-of-the-art tagging systems and
end with the Conclusion in Section 5.
2 GROUPME! SYSTEM
GroupMe! is a new kind of resource sharing sys-
tem. It is comparable to social bookmarking systems
– like del.icio.us – as it enables users to bookmark
Web resources and annotate them with free chosen
keywords. The core idea of GroupMe! is that users
build groups of arbitrary multimedia Web resources
on a specific topic and tag both, Web resources and
groups. Groups can be understood as lightweight wiki
pages. But instead of writing own content, users cre-
ate groups via simple drag & drop operations and via
visual arrangement of contained resources.
Figure 1 shows a screenshot of the GroupMe! sys-
tem. It illustrates a typical scenario. Let us assume
that user fabian plans a trip to the WEBIST 2008 con-
ference in Funchal, Portugal. Therefore, he wants to
build a GroupMe! group containing resources that are
relevant for the trip. Building such a group is simple
and requires just three steps. At first fabian specifies
the group’s name (WEBIST 2008), then he utilizes in-
tegrated search engines – like Google or Flickr search
engine – in order to search for adequate resources, and
finally he adds resources, which are from his point of
view relevant, via drag & drop into the group. Figure
1 depicts such a drag & drop operation with a photo
gathered from Flickr. Furthermore, it shows the en-
tire group fabian has designed. This group contains
images (like a photo of Madeira Island), the official
website of the WEBIST 2008, and a shockwave flash
movie, which presents several photos of Tivoli Ocean
Park Hotel – where the conference will be held – and
gives the opportunity to book rooms. All elements
are visualized according to their media types so that
fabian and other users can see relevant information at
a glance. For example the RSS news feed that informs
about cheap flights to Madeira (see bottom left) di-
rectly presents the latest flight offers to the user. And
the video showing a Portuguese language course (see
top right) can be played back immediately.
Altogether the arranged group in Figure 2 appears
A NOVEL APPROACH TO SOCIAL TAGGING: GROUPME! - Enhancing Social Tagging Systems with Groups
43

like a collage of information artifacts about WEBIST
2008 trip, which is comprehensible for users. Impor-
tantly, content of this group is also accessible and un-
derstandable for machines. Because, when users cre-
ate groups, GroupMe! produces RDF. This is done in
two different ways:
1. Each user interaction – grouping and tagging – is
captured as RDF using several vocabularies, e.g.
FOAF
8
and a GroupMe!-specific vocabulary
9
that
defines new GroupMe! concepts.
2. Whenever a user drops a new Web resource into
a group, domain dependent content extractors
gather useful metadata so that resources can be
enriched with semantically well defined descrip-
tions. When e.g. adding a Flickr photo into a
group, GroupMe! translates Flickr-specific de-
scriptions into a well defined RDF description us-
ing DCMI element set
10
.
RDF created in GroupMe! is made available to
other Web applications and can be accessed via RSS
feeds or RESTful API. Hence, other applications can
benefit from the feature of grouping and enriching re-
sources with machine understandable semantics.
Regarding the above scenario, GroupMe! can be
utilized by a Web service that searches for photos ac-
cording to a location specified via geographic coor-
dinates. With the semantic descriptions captured in
groups like the one about WEBIST 2008 trip, which
contains photos of Funchal as well as a link to Google
maps (see bottom in Figure 1) equipped with the co-
ordinates of Funchal, such a service is now able to
retrieve photos by locations even if these photos are
not directly annotated with geographic coordinates.
2.1 GROUPME! Architecture
In technical terms, GroupMe! is a modular Web ap-
plication that adheres to the Model-View-Controller
pattern. It is implemented using the J2EE application
framework Spring
11
. Figure 2 illustrates the underly-
ing architecture, which consists of four basic layers:
Aggregation. The aggregation layer provides func-
tionality to search for resources a user wants
to add into GroupMe! groups. Currently,
GroupMe! supports Google, Flickr, and of course
a GroupMe!-internal search, as well as adding re-
sources by specifying their URL manually. Con-
tent Extractors allow us to process gathered re-
sources in order to extract useful data and meta-
8
http://xmlns.com/foaf/spec/
9
http://groupme.org/rdf/groupme.owl
10
http://dublincore.org/documents/dces/
11
http://springframework.org/
Data Store
Aggregation
Web
Model
Presentation
Application
Logic
Client API
Java ...
Search Engines
Flickr
Google
Technorati
Content Extractors
Photos Websites
Feeds
...
...
Dublin Core
FOAF
RSS
...
GroupMe!
Core Extended
User
Tag
Group
Resource
Visualizations
Ontology Bridge
Data Access
SQL
...
GroupMe! Search Engine
Group Context Strategy
Tag Relation Strategy
GroupMe! Controller
Group Builder
User Management
RESTful API
searchResources
searchGroups
...
...
HTTP
Request
RDF
Dynamic UI
Interaction Observer
Client User
Resource Rendering
Figure 2: Technical overview of the GroupMe! application.
data, which are converted to RDF using well-
known vocabularies. As mentioned in Section 2,
when e.g. adding a Flickr image into a group,
a Photo content extractor converts Flickr-specific
descriptions into RDF descriptions using Dublin
Core vocabulary. At the moment of writing this
paper, we are extending content extraction func-
tionality by utilizing services like DBpedia
12
or
Sindice
13
, and frameworks like Aperture
14
.
Model. The core GroupMe! model is composed of
four main concepts: User, Tag, Group, and Re-
source. These concepts constitute the base for the
GroupMe! folksonomy (cf. section 2.2). In ad-
dition, the model covers concepts concerning the
users’ arrangements of groups, etc. The Data Ac-
cess layer cares about storing model objects. The
actual data store backend is arbitrarily exchange-
able. At the moment we are using a MySQL
database.
Application Logic. The logic layer provides various
controllers for modifying the model, exporting
RDF, etc. The internal GroupMe! search func-
tionality, which is implemented according to the
strategy pattern in order to switch between differ-
ent search and ranking strategies, is made avail-
able via a RESTful API. It enables third parties
to benefit from the improved search capabilities
(cf. Section 3.1), and to retrieve RDF descrip-
tions about resources – even such resources that
were not equipped with RDF descriptions before
they were integrated into GroupMe!. To simplify
12
http://dbpedia.org/
13
http://sindice.com/
14
http://aperture.sourceforge.net/
WEBIST 2008 - International Conference on Web Information Systems and Technologies
44

usage of exported RDF data, we further provide
a lightweight Java Client API, which transforms
RDF into GroupMe! model objects.
Presentation. The GUI of the GroupMe! applica-
tion is based on AJAX principles. Therefore,
we applied Ajax and JavaScript frameworks like
script.aculo.us
15
, DWR
16
, or Prototype
17
. Such
frameworks provide already functionality to drag
& drop elements, resize elements, etc. Visualiza-
tion of groups and resources is highly modular
and extensible. Switching between components
that render a specific resource or type of resource
can be done dynamically, e.g. visualization of
group elements is adapted to their media type (see
Fig. 1). In the future, users should also be en-
abled to choose an appropriate resource visualiza-
tion from different applicable options.
When creating or modifying groups, each user in-
teraction (e.g. moving and resizing resources) is
monitored and immediately communicated to the
responsible GroupMe! controller so that e.g. the
actual size or position of a resource within a group
is stored in the database.
2.2 GROUPME! Folksonomy
In social tagging systems data is created by users (the
folks), who assign freely chosen tags to resources
(→ tag assignment). The evolving collection of such
tag assignments is called folksonomy
18
. In general,
a folksonomy is formally defined using finite sets
of users, tags and resources, and a finite set of tag
assignments, whereas a tag assignment constitutes a
triple of a certain user, tag and resource (cf. (Mika,
2007)). With GroupMe! we introduce a new concept
to social tagging systems, namely groups.
Definition 1 (Group). A group is a set of resources.
A group is a resource as well. Hence, groups can
contain groups, and groups can be tagged by users.
With definition 1 we extend the formal definition of
a folksonomy introduced in (Hotho et al., 2006a) as
follows.
Definition 2 (GroupMe! Folksonomy). A GroupMe!
folksonomy is a tuple F := (U, T,
˘
R, G,
˘
Y ), where:
– U, T, R, G are finite sets that contain instances of
users, tags, resources, and groups.
15
http://script.aculo.us/
16
http://getahead.org/dwr/
17
http://prototypejs.org/
18
http://vanderwal.net/folksonomy.html
–
˘
R = R ∪ G is the union of the set of resources and
the set of groups.
–
˘
Y defines a relation between these sets (tag as-
signment):
˘
Y ⊆ U × T ×
˘
R × G.
Thus, tagging of resources within the GroupMe! sys-
tem is always done in context of a group, which it-
self may have tags. In comparison to traditional folk-
sonomies, in which relations between tags mainly rely
on their co-occurrences (i.e. two tags are assigned to
the same resource), we obtain new relations between
tags:
1. A relation between tags assigned to different re-
sources that are contained in the same group.
2. A relation between tags assigned to a group and
tags assigned to resources that are contained in the
group.
Similarly, we gain relations between resources that
are contained in the same group, and a part-of-
relation between resources and groups. These new
relations can be exploited by search and ranking
algorithms. For example, when searching for re-
sources with a given tag, an algorithm could also rank
resources that are not directly tagged with the given
query string but which are member of a group that is
tagged with that query string.
A more detailed discussion of such an algorithm and
the GroupMe! folksonomy model can be found in
(Abel et al., 2007).
3 EVALUATION
This section gives an analysis of the GroupMe! sys-
tem, in particular on usage and tagging characteris-
tics, and evaluates the effects of the structure given
by the groups to search and retrieve resources. The
data underlying the analysis was collected during the
first three month after the system’s launch on July 14,
2007. During the observed period, GroupMe! had
a total of 502 resources of which 428 were normal
resources and 74 (14.74%) were groups. Altogether,
929 tag assignments were monitored, with 1.85 tags
per resource in average. The overall evolution of re-
sources and groups is given in Figure 3.
Interestingly, groups were tagged more exten-
sively than ordinary resources: In average, 2.53 tags
were assigned to groups, whereas only 1.73 tags were
attached to other resources. Thus, groups were tagged
1.5 times more often than traditional resources. This
effect was present over time, as depicted in Fig. 4.
Furthermore, at the end of the observed period only
A NOVEL APPROACH TO SOCIAL TAGGING: GROUPME! - Enhancing Social Tagging Systems with Groups
45

0
100
200
300
400
500
Jul 21 Aug 04 Aug 18 Sep 01 Sep 15 Sep 29
number of groups/resources
date
resources and groups
resources
groups
Figure 3: Evolution of number of resources/groups.
17.57% of the groups were not annotated with any tag
in contrast to 32.71% of the resources. These initial
observations give support for the hypothesis that users
adopt the group idea to organize Web resources, and
that they also invest in groups by annotating them.
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
Jul 21 Aug 04 Aug 18 Sep 01 Sep 15 Sep 29
average number of tag assignments
date
tag assignments per resource and group
tag assignments per resource
tag assignments per group
Figure 4: Average number of tags assigned to re-
sources/groups.
A typical group in GroupMe! consists of 4 – 8
resources. That we do not observe groups with sig-
nificantly more members can be explained from the
user interface, which gives the users a canvas to place
and arrange the Web resources. As the size of this
canvas is limited, the on-screen display of the group
becomes impractical with too many Web resources.
Users collect resources with different media types in
their group, as can be seen in table 1. Most popu-
lar among the media types are images, followed by
videos and RSS feeds. Web sites, academic papers,
presentation slides, etc. are denoted as other Web re-
sources and are not mentioned separately, because to
users they appear as simple bookmarks, i.e. their visu-
alization is not yet adapted to their media type partic-
ularly. The possibility to include groups into a group
Table 1: Percentage of resources’ media types that are part
of GroupMe! groups.
Type of Resource AVG Occurrences
images 41.01%
videos 8.57%
rss feeds 4.55%
groups 1.87%
other Web resources 43.96%
was only seldomly used, we explain this by the small
number of available groups during the observation pe-
riod.
3.1 Search Analysis
To examine the effect of groups on search in social
tagging systems, we analyzed the search behavior of
the users. A search operation in our experiment was
either a search performed via the search interface, or
a search which was initiated whenever a user clicked
on a tag in the tag cloud. Obviously, not all clicks
on a tag were intended to perform a search, but more
often were used to explore the tags, and in particu-
lar the popular tags. Hence, we restricted our data
set to only those search operations, which were fol-
lowed by at least one click on a resource. Figure 5
shows how a search result looks like. Images are dis-
played directly, whereas groups, which are denoted
by “(Group)”, and other resources are listed textual.
Figure 5: Screenshot: typical search result list.
We observed altogether 1747 search operations, to
which in average 7.1 results were returned. 11.3%
of the results delivered to a search were groups, and
53.11% of the returned results were not tagged with
WEBIST 2008 - International Conference on Web Information Systems and Technologies
46

the original search string or the tag on which the user
had clicked (query string).
Table 2 lists the results of our experiments on
search operations. We analyzed the top k search re-
sults of each search operation and users’ first click
on a particular resource within the top k. The average
percentage of groups that were part of the top k results
was between 10.38% and 15.39%. However, the per-
centage of search operations, in which users clicked
on groups was between 19.40% and 24.51%. For ex-
ample, when considering the top 10 search results,
we observed that 12.15% of the results were groups
but the percentage of group clicks was 1.84 times
higher, namely 22.34%. Overall, normalization of re-
source and group clicks according to the number of
resources and groups respectively reveals that groups
were selected between 1.39 (Top 3) and 2.76% (Top
50) more frequent than ordinary resources. These ob-
servations support the hypothesis that groups itself
constitute content users are interested in. The demand
for groups is even higher than illustrated in Table 2
because groups can also be accessed using an explo-
rative user interface different from the search inter-
face. However, statistics of the explorative user inter-
face, which has been utilized 271times, are not con-
sidered within our search analysis.
An important benefit of the GroupMe! system
is that it provides the ability to increase the recall
of queries. For our experiments we implemented
search and ranking algorithms that take advantage of
the GroupMe! folksonomy model in order to return
also resources and groups that are not directly tagged
with the given query string (cf. Section 2.2). In gen-
eral, such untagged resources are ranked lower than
resources which are directly tagged with the given
query string. This explains the big increase of the per-
centage of untagged resources when increasing k (see
Untagged in Table 2). Table 2 also points out that
untagged resources are also well-accepted by users.
Considering the top 10, 38.14% of the search results
were not tagged with the query string and in 21.28%
of the search operations users first clicked on an un-
tagged resource, thus on a resource that would not
have been found in a traditional tagging system which
just considers direct resource annotations.
Consequently, the analysis of users’ search opera-
tions validates two main hypotheses:
1. Users are interested in the new group concept:
Groups are selected about two times more often
than they occur within the search result list.
2. GroupMe!’s search and ranking strategies in-
crease recall without reducing proportion of rele-
vant resources remarkably: More than 20% of the
clicks in the top 10 are performed on untagged re-
sources and groups.
4 RELATED WORK
GroupMe! is a social tagging system and com-
petes with systems like BibSonomy, del.icio.us or
Flickr. Table 3 summarizes some characteristics of
GroupMe! according to the dimensions in the tagging
system design taxonomy developed in (Marlow et al.,
2006a), and compares them with related tagging sys-
tems.
Tagging Rights. GroupMe! allows every user to
tag everything (free-for-all) as this enables us to
gather more tags about a resource and alsoa higher
variety of keywords than in constrainted systems.
However, Flickr restricts tagging e.g. to the re-
source owner, friends, or contacts.
Tagging Support. When users annotate resources
they are not supported with tag suggestions as this
would limit the variety of tags. However, they
have the ability to list tags that have already been
assigned to a resource in context of the actual
group. Tags, that have been assigned in context of
other groups – and hence are possibly not appro-
priate in the actual group centext – are not visible
to the user when tagging (blind/viewable).
Aggregation Model. In comparison to Flickr,
which does not allow for duplicated tags (set),
GroupMe! allows different users to assign the
same tag to a certain resource (bag). This may
enable a better evaluation of the importance of
the tags.
Object Type. GroupMe! is the only system listed
in Table 3 that supports tagging of resources dis-
played in a multimedia fashion. Although sys-
tems like del.icio.us enable users to bookmark and
tag arbitrary Web resources, they just visualize re-
sources in a textual way. Hence, while tagging
e.g. an image in del.icio.us, users usually do not
see the image they tag.
Source of Material. Resources that can be annotated
and grouped in GroupMe! are globally distributed
over the Web, and referenced by their URL. This
enables GroupMe! to handle often changing re-
sources like RSS feeds appropriately: Whenever
a group is accessed, the most recent versions of
the contained resources are displayed.
Social Connectivity. All systems listed in Table 3 al-
low users to be linked together. GroupMe! does
not provide integrated features, but utilizes users’
FOAF descriptions in order to identify links be-
tween users.
A NOVEL APPROACH TO SOCIAL TAGGING: GROUPME! - Enhancing Social Tagging Systems with Groups
47

Table 2: Analysis of the Top k search results: (1) percentage of groups contained in the top k search results, (2) percentage
of users who clicked on groups in the top k search results, (3) percentage of resources and groups in the top k that were not
tagged with the given query string, (4) percentage of clicks on untagged resources and untagged groups.
Top k Groups (1) Group Clicks (2) Untagged Resources (3) Untagged Resource Clicks (4)
3 14.70% 19.40% 4.95% 14.93%
5 15.39% 21.69% 17.23% 15.66%
10 12.15% 22.34% 38.14% 21.28%
15 11.89% 22.45% 43.59% 23.47%
20 11.53% 23.00% 46.80% 25.00%
30 10.91% 24.51% 50.73% 25.49%
50 10.38% 24.27% 53.11% 26.21%
Table 3: GroupMe! tagging design in comparison to other social tagging systems. And user incentives in terms of tagging.
Dimension/System GroupMe! BibSonomy del.icio.us Flickr
Tagging Rights free-for-all free-for-all free-for-all permission-based
Tagging Support blind/viewable suggested suggested viewable
Aggregation Model bag bag bag set
Object Type multimedia textual textual images
Source of Material global global global user-contributed
Social Connectivity links links, groups links links
Resource Connectivity groups none none groups
5*User Incentives - future retrieval - future retrieval - future retrieval - future retrieval
- contribution - contribution - contribution - contribution
- sharing - sharing - sharing - sharing
- attract attention - attract attention - attract attention
- self presentation - self presentation
Resource Connectivity. Independent of the users’
tags, a few resource sharing systems provide other
features to connect resources. There are some sys-
tems that allow users to organize themselves into
groups, and that provide functionality to retrieve
resources, which are related to these groups – e.g.
BibSonomy or Connotea
19
. However, to the best
of our knowledge, Flickr and GroupMe! are at the
moment the only notable tagging systems that en-
able users to assign resources to groups explicitly.
Such hand-selected groups are highly valued by
the users as indicated in our analysis (see Section
3.1).
User Incentives. GroupMe! users have several moti-
vations to annotate resource ranging from simpli-
fication of future retrieval to self presentation (e.g.
some users tag resources with holiday in order to
express which locations they have visited).
What makes GroupMe! unique is that groups can be
tagged and resources are always tagged in context of
a specific group. Thereby, GroupMe! extends the tra-
ditional folksonomy model, which has been theorized
in (Marlow et al., 2006b) or (Mika, 2007), and for-
malized in (Hotho et al., 2006a). With the GroupMe!
19
http://www.connotea.org/
folksonomy model (see Section 2.2) new relations
between resources, groups and tags emerge that can
be exploited by search and ranking algorithms (Abel
et al., 2007). Search and ranking algorithms that op-
erate on traditional folksonomies have already been
successfully applied in order to improve Web search.
In (Bao et al., 2007) the authors introduced Social-
SimRank, which adapts SimRank (Jeh and Widom,
2002) and computes similarity between tags and re-
sources respectively. Furthermore, Bao et al. pre-
sented the SocialPageRank algorithm, which ranks
Web resources according to how popular they are an-
notated. FolkRank (Hotho et al., 2006c) is another
folksonomy-based search algorithm, which adapts the
famous PageRank (Page et al., 1998) algorithm and
involves user preferences. In our future work we will
compare our algorithms, which exploit the GroupMe!
folksonomy model, with the mentioned algorithms.
Learning relations between tags is another chal-
lenge in social tagging systems that can be utilized to
improve retrieval of resources additionally. Hotho et
al. presented an approach to mine association rules
in folksonomies that point to subtag-supertag rela-
tions (Hotho et al., 2006b). The GroupMe! folk-
sonomy model provides a foundation to deduce such
relations more precisely, e.g. by analyzing tags that
WEBIST 2008 - International Conference on Web Information Systems and Technologies
48

have been assigned to a group and tags of group mem-
bers. In (Rattenbury et al., 2007) the authors inves-
tigated how to learn more concrete semantics from
folksonomies. In particular, they presented an ap-
proach to distinguish between event tags and place
tags. Mentioned approaches for learning semantics
can also be applied to GroupMe!. At the moment, in-
stead of learning vague semantics, GroupMe! extracts
semantic descriptions explicitly when new resources
are added to a group. Hence, these descriptions can
be utilized by machines offhand in order to search for
certain type of resources (cf. example in Section 2).
Therefore, all RDF produced in GroupMe! is feeded
back to the Web via RDF feeds and RESTful API.
Other systems like CiteULike
20
or BibSonomy just of-
fer RSS export.
5 CONCLUSIONS
GroupMe! gives users the possibility to group Web
resources in an easy way – by simple drag & drop
operations – and combines this idea with features of
social tagging systems. The evaluation of GroupMe!
shows that users appreciate the grouping facility to or-
ganize Web resources they are interested in. Groups
can be seen as hand selected collections of Web con-
tent for a certain topic or domain. As such, they are
also valuable results to search queries, and our inves-
tigations have shown that users recognize this and se-
lect groups among the search results often.
The structure inherently given by the groups can also
be used to infer information about the content of Web
resources. This is interesting for non-tagged Web
resources, and particularly for multimedia Web re-
sources whose content is - without tags - hardly deter-
minable (like videos etc.). The analysis of the search
behavior of users has revealed that this exploitation
of grouping information uncovers relevant content,
which – with tagging alone – would not have been
found.
REFERENCES
Abel, F., Frank, M., Henze, N., Krause, D., Plappert, D.,
and Siehndel, P. (2007). GroupMe! – Where Semantic
Web meets Web 2.0. In International Semantic Web
Conference (ISWC 2007).
Bao, S., Xue, G., Wu, X., Yu, Y., Fei, B., and Su, Z.
(2007). Optimizing web search using social annota-
tions. In WWW ’07: Proceedings of the 16th interna-
20
http://www.citeulike.org/
tional conference on World Wide Web, pages 501–510,
New York, NY, USA. ACM Press.
Garrett, J. J. (2005). AJAX: A new approach to web
applications. Adaptive Path. http://www.adaptive-
path.com/publications/essays/archives/000385.php.
Hotho, A., J
¨
aschke, R., Schmitz, C., and Stumme, G.
(2006a). BibSonomy: A social bookmark and pub-
lication sharing system. In de Moor, A., Polovina,
S., and Delugach, H., editors, Proc. First Conceptual
Structures Tool Interoperability Workshop, 14th Int.
Conf. on Conceptual Structures, pages 87–102, Aal-
borg.
Hotho, A., J
¨
aschke, R., Schmitz, C., and Stumme, G.
(2006b). Emergent semantics in BibSonomy. In
Hochberger, C. and Liskowsky, R., editors, Informatik
2006 - Informatik f
¨
ur Menschen, volume 94(2) of LNI,
Bonn. GI.
Hotho, A., J
¨
aschke, R., Schmitz, C., and Stumme, G.
(2006c). FolkRank: A ranking algorithm for folk-
sonomies. In Proc. FGIR 2006.
Jeh, G. and Widom, J. (2002). SimRank: A Measure
of Structural-Context Similarity. In Proc. of Inter-
national Conference on Knowledge Discovery and
Data Mining (SIGKDD), Edmonton, Alberta, Canada.
ACM.
Marlow, C., Naaman, M., Boyd, D., and Davis, M. (2006a).
HT06, tagging paper, taxonomy, flickr, academic arti-
cle, to read. In HYPERTEXT ’06: Proceedings of the
seventeenth conference on Hypertext and hypermedia,
pages 31–40, New York, NY, USA. ACM Press.
Marlow, C., Naaman, M., Boyd, D., and Davis, M. (2006b).
Position Paper, Tagging, Taxonomy, Flickr, Article,
ToRead. In Collaborative Web Tagging Workshop at
WWW2006.
Mika, P. (2007). Ontologies are us: A unified model of so-
cial networks and semantics. Web Semantics: Science,
Services and Agents on the World Wide Web, 5(1):5–
15.
Page, L., Brin, S., Motwani, R., and Winograd, T. (1998).
The Pagerank citation ranking: Bringing order to the
web. Technical report, Stanford Digital Library Tech-
nologies Project.
Rattenbury, T., Good, N., and Naaman, M. (2007). To-
wards automatic extraction of event and place seman-
tics from flickr tags. In SIRIR ’07: Proceedings of the
30th annual international ACM SIGIR conference on
Research and development in information retrieval,
pages 103–110, New York, NY, USA. ACM Press.
A NOVEL APPROACH TO SOCIAL TAGGING: GROUPME! - Enhancing Social Tagging Systems with Groups
49