The Networked Semantic Desktop
,
Stefan Decker
DERI, NUIG, Ireland
stefan@deri.ie
Martin Frank
ISI, USC, USA
frank@isi.edu
,


Abstract

We present our vision of a new group collaboration
infrastructure, the Networked Semantic Desktop, drawing
from co-evolving research in the Semantic Web, Peer-to-
Peer (P2P) Networks, and Online Social Networking.

1. Introduction

The Internet, electronic mail, and the Web have
revolutionized the way we communicate and collaborate -
their mass adoption is one of the major technological
success stories of the 20th century. We now face a
qualitatively different problem, information overload, that
necessitates smarter and more fine-grained computer
support for networked information, and that has to blend
the boundaries between personal and group data, while
simultaneously safeguarding privacy and establishing
trust. In other words, the current computing infrastructure
does not really support knowledge workers all that well:
for example, sending a single file to a mailing list
multiplies the cognitive processing effort of filtering and
organizing this file times the number of recipients –
leading to more and more of peoples’ time going into
information filtering and organization activities.
Centralized collaborative infrastructures (like BSCW or
Sharepoint) help to a certain extent, but the current
application infrastructure does not let you interconnect
separate data items, like the author of a document and her
corresponding entry in your address book - much less let
you share that interconnection with others.

Several new technology thrusts have now emerged which
could dramatically impact how people interact and
collaborate: The Semantic Web, P2P Computing, and
Online Social Networking. This paper presents a vision of
how the different thrusts will evolve to produce the
Networked Semantic Desktop, which enables people and
communities to directly collaborate with their peers while
dramatically reducing the amount of time they spend
filtering and filing information.

2. Usage Scenarios

We exemplify the impact of a Networked Semantic
Desktop with two usage scenarios.

2.1. Surviving the Information Flood and
Creating Knowledge in the Process

In our daily life, many of us get hundreds of emails,
often with documents attached from the various different
projects and communities we are involved in. These
documents are always created within a context on the
author’s machine – but they are send out as if they had no
context – they arrive without trusted metadata that would
allow automatic processing and filing on the recipient’s
machine. The process has the following shortcomings:
• Apart from the folder structure the current Windows-
style desktop and file system provides no support for
organizing the information in the documents. This
means the recipient has to cope with an insufficient
support of current desktops systems for organizing
the information – you may put them under the 2004-
04 folder, or the Proposals folders, or the Semantic
Web Research folder – but not all three.
• Since the metadata of the document has been lost
when sending the email every recipient has to
reinvent and recreate their own metadata, re-
categorize the document and create the possible
connections to other information. This is only of
marginal societal value because most metadata has
been created before.
• Even if metadata was integrated into the email, the
author and the recipient of the document usually have
different, personal classification schemes, and there
is no way to selectively “open-source” them and
align them with others’.
It is clearly possible to share and replicate documents
as well as metadata via direct P2P connections. It should
be possible for sub-communities to derive metadata in a
distributed fashion via an implicit or explicit consensus
process. Connections and relationships with other pieces
of information could then be accessed by all members of
the community – in this sense a distributed knowledge
base is constructed around the work topics, documents,
and information contained in these document. The
information can then be viewed in multiple dimensions

(year, type of document, content of document, and so on).
Also, the community itself can be queried, similar to
current Online Social Networks: members of the
community can query to whom they are connected to and
can share information with “friends of a friend” or can
query for information – thus existing trust relationships
between individuals can be leveraged to compute one’s
trust in metadata.

2.2 Connecting to Trusted Colleagues

Many of the same ideas are applicable to the audience
of this workshop - research is performed in scientific
communities of interest in which members develop,
review, publish and discuss each other work. Membership
in a scientific community is based on interest and
abilities: promising applicants are included by accepting
their conference papers or inviting them on editorial
boards. As a research field progresses these communities
define their own language, which is often
incomprehensible to outsiders and difficult to learn, yet
necessary for efficiently classifying and communicating
the topics of interest and for gaining an overview of the
different types of research. This language evolves in a
sluggish community process today - usually by members
reading each others’ papers, by attending the same
scientific events (e.g., workshops such as this one) and by
then adopting each others’ terminology over time; the
appearance of text books and dedicated overview papers
consolidates a field’s vocabulary in the later stages. This
evolutionary structuring of new scientific fields is largely
invisible to non-scientists, both because journal
publications can often not be accessed without paying
hefty fees and because in any case it would take too much
effort to derive a taxonomy of a scientific field by reading
scholarly publications.
Paradoxically, being an “insider scientist” can
nevertheless be a lonely experience because
communication with other community members by e.g.
ping-pong journal publications is a slow and faceless
process, lacking the spontaneity and friendship-building
opportunities of face-to-face communication.
The Networked Semantic Desktop has the potential to
accelerate scientific collaboration via a peer-to-peer end-
user application for maintaining shared views of scientific
fields, as well as to make these evolving views explicit
and available to the public at no cost. Such a collaborative
application is peer-based both in the scientific sense
(“peer review”) and technical sense (“peer-to-peer
technology”). Participants automatically become part of a
global peer-to-peer network for scientific meta-data, and
takes responsibility for a (proportional) fraction of the
disk storage, bandwidth, and computing cycles to support
it, probably based on a structured P2P network. This
single global network could then support a large number
of small scientific sub-communities, each of which
revolves around jointly maintaining a shared view of a
small sub-field. Maintaining this shared view is, of
course, not typically an end to itself, but serves as the
focal point that enables a scientific community to
effectively exchange research papers, data sets, and ideas.
Such close-knit communities may work on bottom-up
taxonomies for a tiny new sub-field of Science, such as
say Pteroylglutamic Acids, or may work on a top down
categorization of say Liberal Arts as a whole, and each
community can refer to concepts in other communities.
Think of this type of Networked Semantic Desktop as
collectively harnessing the power of millions of currently
hand-scribbled categorizations of scientific sub-fields into
an inter-linked, grass-roots, and world-wide view of
Science.
For individual scientists to participate in this global
scientific meta-data construction effort, there must be
immediate benefits to joining, as well as assurances that
certain things participants will naturally fear will not
occur. The benefits are: By joining the global scientific
meta-data network, (b1) one can access and query
existing scientific terminology, (b2) one can mine queries
by others to find out what the “hot” research topics are,
and (b3) one can view others’ structuring of their fields -
thus providing instant gratification to joining the global
meta-data network. By joining a specific scientific sub-
community as an active participant, over time, (b4) one
can get to know others in one’s research area for
friendship and scientific collaboration, and (b5) one can
make a name for oneself by contributing to community
taxonomies and the research itself. The assurances are:
The peer-to-peer application must re-assure prospective
participants that (a1) others will not claim their
intellectual contributions as their own, (a2) they will
always be in control of their personal view of the field,
and they can always choose to cease collaborating with
any individual or sub-community, (a3) they will not be
politically dependent on any individual in some
“gatekeeper” position in the global scientific meta-data
network (a4) they will not be technically dependent on
the people who own the technical community
infrastructure, and (a5) their machine will not be
intolerably slowed down by participating. While these
requirements seem daunting, we have hope that with
recent work on structured P2P networks for meta-data
exchange as well as advances in the Resource Description
Framework-based Semantic Web languages, such a
global scientific collaboration network is within reach in
the next five years - whereas just five years ago it would
have seemed like a fantasy.

3. Components of the Networked Semantic
Desktop

These information items can be treated as Semantic Web
resources, which enable the usage of Semantic Web
technologies to manage desktop data. Similar to Web
resources, files, emails and other items can be identified
using a Uniform Resource Identifier (URI). The metadata
of these desktop items are represented as RDF graphs,
which can then be used for browsing or searching by
faceted metadata browsing techniques [17] and can again
be shared with others.

This enables a co-evolution with current desktop
technology, enabling the development of methods to
create metadata and to interlink information on a on a
local desktop computer (e.g. the author of a text
document with the entry in the address database).
Approaches like MITs Haystack [1], or European
approaches like Gnowsis [2], or Fenfire [3] show that
there is activity going on in this direction. However, these
approaches concentrate on the higher levels of the
desktop system like the user interface, or realize “add-on”
components to handle metadata rather than being built
upon meta-data from the ground up. For example, the file
system in most operating systems does not have the
capability to manage larger amount of heterogeneous
metadata because of minimum file size issues. File system
approaches like ReiserFS (see [4]) or Microsofts
Longhorn raise the hope that current operating system can
evolve to effectively support management of metadata on
all levels – from the disk storage to the user interface.

3.2 P2P Systems

Starting initially with file sharing application like
Gnutella, P2P systems have achieved an enormous
amount of research attention, especially in the database
community [14] [15]. [13] suggested to add metadata
management to P2P networks. This idea has been taken
up by approaches like Edutella [5][6][7][11], RDFPeers
[8], P-Grid [121] and other projects [9][10]. Initial ideas
of a collaboration infrastructure based on P2P technology
has been realized in Groove which allows small groups to
share a calendar, discussions, and files without a central
server and see who is off- and on-line and to sent instant-
message with them (but replicates all content to all group
members which does not scale well with group size).

However, the field is still in his infancy –there is an
distributed efficient query model for distributed RDF
sources, but no agreement how “semantics” (as in,
inferencing within the network) should be incorporated
into P2P networks. Topics like scalable infrastructures for
service discovery in P2P networks, interoperation
Ontology driven
distributed
Social Networking
Ontology driven Social Networking
Semantic Desktop Networked
Semantic Desktop
P2P networks
Semantic Web
Desktop
Semantic P2P
Social Networking
Phase 1 Phase 2 Phase 3
Figure 2 Phases towards the Networked Semantic Desktop

infrastructure for heterogeneous peers to translate
information, emergent semantics and incremental
learning, evolution of ontologies in an P2P environment,
semantics-based routing, and semantics-based topologies
for P2P networks are currently all open topics and
actively discussed in the scientific community (e.g.,
[18][19]).

3.3 Online Social Networking

A relatively new field is Online Social Networking,
which recently got a lot of attention by venture capitalists
and internet users. Sites like linkedin.com, Friendster, and
Orkut.com were able to attract millions of users and
provide the infrastructure to
a) make relationships explicit, so persons can
explore their personal network, and
b) make new connections and establish new
relationships.

People are using social networking sites for personal
and professional use, communications, new business
developments and contacts, dating and virtual meetings.
New communities can be established (like in Orkut.com
or Tribes.net).
Individuals are highly motivated to sign up (for
example, Orkut invitations have been auctioned off on
Ebay) and to increase their visibility within a network,
and to get as many people to join their network, driven by
vanity (much like much of the original Web sites were).
While Online Social Networking (OSN) itself – as the
current success of the existing sites shows – provides
motivation for individuals to sign up, the current use
possibilities of online Social Networking Sites are rather
limited. The sheer availability of social connection
information, however, opens up new collaboration
possibilities, such as “Link Routing”: the routing of
information based on the social connections between
people. To exploit this kind of information the
relationship information has to be made explicit and
accessible. Approaches like the “Friend of a Friend”
(FOAF) project [21] – actually predating the current
Online Social Networking boom – provide this
information in a machine accessible exploitable way.
Furthermore current Online Social Networking sites
have another disadvantage:
Maintaining an Online Social Networking site requires
a major investment. The main exploitable capital of the
OSN sites is the user profile and relationship information
– information that can be used e.g., for providing targeted
advertisement (see e.g., the privacy statement of
Orkut.com). Thus these sites are extremely unlikely to
share this information openly or even with each other,
which will seriously hamper the development of a next
generation collaborative infrastructure. The obvious
general solution is to build an Online Social Networking
Infrastructure on top of a P2P system, with the following
advantages:
• New applications can be easily added to the network
• Profile and user information remains the property of
individual users – multiple OSN sites can crawl it,
and the user is protected against losing here data if an
OSN site shuts down.
• No large investment in a centralized site is necessary,
since the cost for maintaining the overall network is
shared among all users.

As a result it seems desirable to a) base Online Social
Networks on Semantic Web technology, and b) exchange
and deploy the social information in P2P infrastructures
to resolve the control and ownership issue, resulting in an
infrastructure similar to the one described in [20].

4. Development Phases for the Networked
Semantic Desktop

We envision three research, development and
deployment phases for the Networked Semantic Desktop
(see figure 2). In the first phase, Semantic Web, P2P, and
Social Networking technologies are developed,
researched and partially deployed. In the second stage, we
will see convergence: Semantic Web technology is
deployed on the Desktop, resulting in the Semantic
Desktop. Similarly, Semantic Web technology is
incorporated in P2P networks and Social Networking.
Once there is a reliable technology available for the
technology convergences, the next phase can be tackled:
the combination of the three fields Semantic Desktop,
Semantic P2P and ontology-driven Social Networking
into the Networked Semantic Desktop.

5. Conclusion

In this paper, we described our vision of how different,
currently very active research fields will interact and co-
evolve, resulting in a new internet-based group
collaboration infrastructure we called the “Networked
Semantic Desktop”, which will let individuals collaborate
at a much finer-grained level as is possible with the
Windows-style desktop today, and will result in dramatic
time savings in filtering out marginal information and
discovering vital information.

6 . References

[1] Dennis Quan, David Huynh, and David R. Karger:
Haystack: A Platform for Authoring End User Semantic
Web Applications. International Semantic Web
Conference 2003: 738-753

[2] The Gnowsis project (visited April 2004): URL:
http://www.gnowsis.com/wiki

[3] The FenFire project (visited April 2004): URL:
http://www.nongnu.org/fenfire/

[4] Hans Reiser: Future Vision (visited April 2004):
URL: http://www.namesys.com/whitepaper.html

[5] Mario T. Schlosser, Michael Sintek, Stefan Decker,
and Wolfgang Nejdl: A Scalable and Ontology-Based
P2P Infrastructure for Semantic Web Services. Peer-to-
Peer Computing 2002: 104-111

[6] Mario Schlosser, Michael Sintek, Stefan Decker, and
Wolfgang Nejdl: HyperCuP - Hypercubes, Ontologies
and P2P Networks.Springer Lecture Notes on Computer
Science, Vol. 2530 – Agents and Peer-to-Peer Systems

[7] Wolfgang Nejdl, Boris Wolf, Changtao Qu, Stefan
Decker, Michael Sintek, Ambjörn Naeve, Mikael Nilsson,
Matthias Palmér, and Tore Risch: EDUTELLA: a P2P
networking infrastructure based on RDF. In 11th World
Wide Web Conference, May 2002: 604-615

[8] Min Cai and Martin Frank. RDFPeers: A Scalable
Distributed RDF Repository based on A Structured Peer-
to- Peer Network, University of Southern California -
Computer Science Technical Report 03-807, December
2003 (also accepted for the Semantic Web track of the
2004 World Wide Web conference -- this is the version
of the paper as it was submitted).

[9] A. Castano, S. Ferrara, S. Montanelli, E. Pagani, G.P.
Rossi (University of Milan): Ontology-Addressable
Contents in P2P Networks. In: 1st Workshop on 1st
Workshop on Semantics in Peer-to-Peer and Grid
Computing at the Twelfth International World Wide Web
Conference 20 May 2003, Budapest, Hungary. See:
http://www.isi.edu/~stefan/SemPGRID/

[10] Arturo Crespo and Hector Garcia-Molina. Semantic
Overlay Networks, Submitted for Publication, 2002. See:
http://www-db.stanford.edu/peers/

[11] W. Nejdl, M. Wolpers, W. Siberski, C. Schmitz, M.
Schlosser, I. Brunkhorst, and A. Lser. Super-peer-based
routing and clustering strategies for RDF-based peer-to-
peer networks. In 12th World Wide Web Conference,
May 2003.

[12] Karl Aberer, Philippe Cudré-Mauroux, and Manfred
Hauswirth: Start making sense: The Chatty Web approach
for global semantic agreements. In: Journal of Web
Semantics (Elsevier) 1(1), 2004.
[13] Rael Dornfest and Dan Brickley: The Power of
Metadata. http://www.openp2p.com/pub/a/
p2p/2001/01/18/metadata.html, excerpted from the book
"Peer-to-Peer: Harnessing the Power of Disruptive
Technologies (Jan. 2001).

[14] Mayank Bawa, Brian Cooper, Arturo Crespo, Neil
Daswani, Prasanna Ganesan, Hector Garcia-Molina,
Sepandar D. Kamvar, Sergio Marti, Mario T. Schlosser,
Qi Sun, Patrick Vinograd, and Beverly Yang: Peer-to-
peer research at Stanford. SIGMOD Record 32(3): 23-28
(2003)

[15] Groove: URL: http://www.groove.net/ (Visited April
2004).

[16] Alon Y. Halevy, Zachary G. Ives, Peter Mork, and
Igor Tatarinov: Piazza: data management infrastructure
for semantic web applications. WWW 2003: 556-567

[17] Yee, K-P., Swearingen, K., Li, K., and Hearst, M.,
Faceted Metadata for Image Search and Browsing, in the
CHI 2003 Conference on Human Factors in Computing
Systems.

[18] Karl Aberer, Stefan Decker, David De Roure, Carole
Goble (eds.): 1st Workshop on Semantics in Peer-to-Peer
and Grid Computing at the Twelfth International World
Wide Web Conference 20 May 2003, Budapest, Hungay

[19] Karl Aberer, Stefan Decker, David De Roure, Carole
Goble, Hongsuda Tangmunarunkit (eds.): The Second
Workshop on Semantics in Peer-to-Peer and Grid
Computing at the Thirteenth International World Wide
Web Conference 18 May 2004, New York, USA.

[20] Ken Jordan, Jan Hauser, and Steven Foster: The
Augmented Social Network: Building identity and trust
into the next-generation Internet. In: First Monday,
volume 8, number 8 (August 2003), URL:
http://firstmonday.org/issues/issue8_8/jordan/index

[21] The Friend of a Friend (FOAF) project. URL:
http://www.foaf-project.org/. Visited: April 2004.