The Semantic Web Revisited - The Semantic Web
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
The Semantic Web
Editor: Steffen Staab
University of Koblenz-Landau
staab@uni-koblenz.de
The Semantic Web Revisited
Nigel Shadbolt and Wendy Hall, University of Southampton
Tim Berners-Lee, Massachusetts Institute of Technology
entific American article assumed that this would be straight-
I n the 50 years since the term AI was coined at the Dart-
mouth Conference, the digital world has evolved at a
prodigious rate. It has produced an information infrastruc-
forward, but it’s still difficult to achieve in today’s Web.
The article included many scenarios in which intelli-
gent agents and bots undertook tasks on behalf of their
ture that few would have anticipated—with the possible human or corporate owners. Of course, shopbots and auc-
tion bots abound on the Web, but these are essentially
exception of Vannevar Bush,1 although even he might have handcrafted for particular tasks; they have little ability to
thought the scale of achievement extraordinary. Today, the interact with heterogeneous data and information types.
World Wide Web links 10 billion pages, and search engines Because we haven’t yet delivered large-scale, agent-based
can divine themes embodied in the links to serve useful and mediation, some commentators argue that the Semantic
relevant content almost instantaneously. Web has failed to deliver. We argue that agents can only
Fifty years ago it might have appeared audacious to build flourish when standards are well established and that the
a global web of information, to deploy semantics on such Web standards for expressing shared meaning have pro-
a scale, and to attempt inference over the resulting compo- gressed steadily over the past five years. Furthermore, we
nents. Fifty years ago, even if you could have explained see the use of ontologies in the e-science community pre-
it, a Semantic Web would have seemed as remote as general saging ultimate success for the Semantic Web—just as the
AI. Yet today we believe that the Semantic Web is attain- use of HTTP within the CERN particle physics commu-
able. We are seeing its first stirrings, and it will draw on nity led to the revolutionary success of the original Web.
some key insights, tools, and techniques derived from 50
years of AI research. A growing need for data integration
Meanwhile, the need has increased for shared seman-
From documents to data and information tics and a web of data and information derived from it.
The original Scientific American article on the Seman- One major driver—one that this magazine has reported on
tic Web appeared in 2001.2 It described the evolution of a extensively—has been e-science (IEEE Intelligent Sys-
Web that consisted largely of documents for humans to tems, special issue on e-science, Jan. 2004). For example,
read to one that included data and information for comput- life sciences research demands the integration of diverse
ers to manipulate. The Semantic Web is a Web of action- and heterogeneous data sets that originate from distinct
able information—information derived from data through communities of scientists in separate subfields. Scientists,
a semantic theory for interpreting the symbols. The seman- researchers, and regulatory authorities in genomics, pro-
tic theory provides an account of “meaning” in which the teomics, clinical drug trials, and epidemiology all need a
logical connection of terms establishes interoperability way to integrate these components. This is being achieved
between systems. This was not a new vision. Tim Berners- in large part through the adoption of common conceptual-
Lee articulated it at the very first World Wide Web Confer- izations referred to as ontologies. In the past five years,
ence in 1994. This simple idea, however, remains largely the argument in favor of using ontologies has been won—
unrealized. numerous initiatives are developing ontologies for biology
A Web of data and information would look very differ- (for example, see http://obo.sourceforge.net), medicine,
ent from the Web we experience today. It would routinely genomics, and related fields. These communities are devel-
let us recruit the right data to a particular use context—for oping language standards that can be deployed on the Web.
example, opening a calendar and seeing business meet- Many other disciplines are adopting what began in the
ings, travel arrangements, photographs, and financial life sciences. Environmental science is looking to integrate
transactions appropriately placed on a time line. The Sci- data from hydrology, climatology, ecology, and oceanogra-
96 1541-1672/06/$20.00 © 2006 IEEE IEEE INTELLIGENT SYSTEMS
Published by the IEEE Computer SocietyResource Description Framework
RDF assigns specific Universal Re-
source Identifiers (URIs) to its individ-
ual fields. Figure A is an example RDF http://www.w3.org/2000/10/swap/pim/contact#Person
graph from the W3C RDF Primer
(www.w3.org/TR/rdf-primer), show- http://www.w3.org/1999/02/22-rdf-syntax-ns#type
ing a representation for a person
http://www.w3.org/People/EM/contact#me
named Eric Miller. As we create an
RDF graph of nodes and arcs, a URI
http://www.w3.org/2000/10/swap/pim/contact#fullName
reference used as a graph node iden-
tifies what the node represents; a URI
used as a predicate identifies a rela- Eric Miller
tionship between the things identi-
fied by the connected nodes. So, the http://www.w3.org/2000/10/swap/pim/contact#mailbox
RDF in Figure A represents
• individuals—such as Eric Miller, mailto:em@w3.org
identified by http://www.w3.org/
People/EM/contact#me; http://www.w3.org/2000/10/swap/pim/contact#personalTitle
• kinds of things—such as Person,
identified by http://www.w3.org/ Dr.
2000/10/swap/pim/contact#Person;
• properties of those things—such
as mailbox, identified by http:// Figure A. An RDF graph representing Eric Miller.
www.w3.org/2000/10/swap/pim/
contact#mailbox; and
• values of those properties—such as mailto:em@w3.org as
the value of the mailbox property (RDF also uses character
types such as integers and dates as property values).
RDF also provides an XML-based syntax called RDF/XML for
Eric Miller
recording and exchanging graphs. Figure B shows a small chunk
of RDF in RDF/XML corresponding to the graph in Figure A. Dr.
The Figure B rendering is actually quite clumsy syntactically,
and its lack of transparency and readability might have been a
factor inhibiting rapid adoption of RDF. However, there are
alternative forms that are easier to interpret; for example, see Figure B. A chunk of RDF in RDF/XML describing Eric Miller and
the N3 notation (www.w3.org/DesignIssues/Notation3.html). corresponding to the graph in Figure A.
phy (see http://marinemetadata.org/examples/ EU countries are developing similar pro- languages for sharing meaning. These lan-
mmihostedwork/ontologieswork). The need grams to implement the EU directive. guages provide a foundation for semantic
to understand systems across ranges of scale Despite these and other significant drivers interoperability.
and distribution is evident everywhere in sci- in defense, business, and commerce, it’s still In 1997, the W3C defined the first Resource
ence and presents a pressing requirement for apparent that the Semantic Web isn’t yet with Description Framework specification (see
data and information integration. us on any scale. the related sidebar). RDF provided a simple
Various e-government initiatives repre- So let’s review what progress we’ve made but powerful triple-based representation lan-
sent similar efforts. The United Kingdom and consider the various impediments to its guage for Universal Resource Identifiers
has developed an Integrated Public Sector global adoption. (URIs). It became a W3C recommendation
Vocabulary (www.esd.org.uk/standards/ by 1999—a crucial step in drawing attention
ipsv). The recently created UK Office of Progress to the specification and promoting its wide-
Public Sector Information (www.opsi.gov. Consistent with the need for a Web seman- spread deployment to enhance the Web’s
uk) is a response to an EU directive (2003/ tics, the user community, including standards functionality and interoperability.
98/EC, http://www.ec-gis.org/document. organizations like the Internet Engineering The original Web took hypertext and made
cfm?id=486&db=document). OPSI aims to Task Force and the World Wide Web Con- it work on a global scale; the vision for RDF
exploit the considerable amounts of gov- sortium (W3C), has directed major efforts was to provide a minimalist knowledge repre-
ernment data for citizens’ benefit. Several at specifying, developing, and deploying sentation for the Web.
MAY/JUNE 2006 www.computer.org/intelligent 97Universal Resource Identifiers stages for W3C recommendation status, is still needs tools and software development
URIs identify resources and so are cen- designed to fulfill this requirement. environments to support its production and
tral to the Semantic Web enterprise.3 Using application. These are starting to appear but
a global naming convention (however arbi- RDF translation as yet we have few means to routinely and
trary the syntax) provides the global net- Other significant progress includes GRDDL effortlessly generate Semantic Web annota-
work effects that drive the Web’s benefits. (Gleaning Resource Descriptions from tions using this or other languages at the point
URIs have global scope and are interpreted Dialects of Languages, www.w3.org/2004/ of content use or creation.
consistently across contexts. Associating a 01/rdxh/spec), which provides a means to
URI with a resource means that anyone can extract RDF from XML and XHTML docu- Rules and inference
link to it, refer to it, or retrieve a represen- ments using transformations expressed in But ontologies are only one part of the
tation of it. XSLT (Extensible Stylesheet Language) representation picture. Rules and inference
Given the Semantic Web’s aims, we want and associated with the original content. also need support. The OWL language itself
to reason about relationships. URIs provide This capability could potentially overcome is designed to support various types of infer-
the grounding for both our objects and rela- the RDF bootstrap problem by generating ence—typically, subsumption and classifica-
tions. They underpin the Semantic Web, sufficient RDF for serendipitous reuse to tion—and a range of automated reasoners are
allowing machines to process data directly. occur. The amount of XML and XHTML available (for example, see www.cs.man.ac.
In this way, the Semantic Web shifts the data on the Web, especially data generated uk/~sattler/reasoners.html). Because it’s diffi-
emphasis from documents to data. Much of from back-end databases, is considerable cult to specify a formalism that will capture
the motivation for the Semantic Web comes and offers good opportunities for RDF all the knowledge in a particular domain,
from the value locked in relational databases. conversion. there are other approaches to inference on
To release this value, database objects must the Web. Work has begun on the Rule Inter-
be exported to the Web as first-class objects change Format (www.w3.org/2005/rules), an
and therefore must be mapped into a system attempt to support and interoperate across a
of URIs. URIs have global scope. variety of rule-based formats. RIF will ad-
Languages have evolved to offer greater dress the plethora of rule-based formalisms:
opportunities for encoding meaning that can Associating a URI with a Horn-clause logics, higher-order logics, pro-
support information integration and interop- duction systems, and so on.
erability. RDF Schema became a recom- resource means that anyone Moreover, AI researchers have extended
mendation in February 2004. RDFS took the these various logics and modified them to
basic RDF specification and extended it to
support the expression of structured vocabu-
can link to it, refer to it, or capture causal, temporal, and probabilistic
knowledge. Causal logic, such as Glenn
laries. It has provided a minimal ontology Shafer proposed,5 developed out of action
representation language that the research
retrieve a representation of it. logics in AI, and it’s intended to capture an
community has adopted fairly widely. important aspect of commonsense under-
standing of mechanisms and physical sys-
Triple stores Web Ontology Language tems. Temporal logic formalizes the rules
As RDF and RDFS have gained ground, For those who required greater expres- for reasoning with propositions indexed to
the need for repositories that can store RDF sivity in their object and relation descrip- particular times; Zhisheng Huang and
content has grown. These so-called triple tions, the OWL (Web Ontology Language, Heiner Stuckenschmidt suggested tempo-
stores vary in their capabilities. Some focus www.w3.org/TR/2004/REC-owl-features- ral-logic approaches for ontology version
on providing a rich means to reason over 20040210) specification integrated several management.6 Probabilistic logics are cal-
the triples (for example, see http://jena. efforts. The W3C recommendation pre- culi that manipulate conjunctions of proba-
sourceforge.net), while others focus on sents three versions of OWL, depending on bilities of individual events or states. Perhaps
storing large quantities of data (see http:// the degree of expressive power required. the most well-known of these are Bayesian,
sourceforge.net/projects/threestore for an OWL’s core idea is to enable efficient rep- which you can use to derive probabilities for
example from the open source community resentation of ontologies that are also events according to prior theories about
and www.oracle.com/technology/tech/ amenable to decision procedures. It checks how probabilities are distributed. Bayesian
semantic_technologies/index.html for a an ontology to see whether it’s logically reasoning is commonplace in search engines.
commercial example). Some operate as consistent or to determine whether a partic- In domains where reasoning under uncer-
plug-ins to current Web browsers (http:// ular concept falls within the ontology. tainty is essential, such as bioinformatics,
simile.mit.edu/piggy-bank) and others as sys- OWL uses the linking provided by RDF Kenneth Baclawski and Tianhua Niu have
tems that can operate with a range of existing to allow ontologies to be distributed across suggested using Bayesian ontologies to
third-party databases (www.openrdf.org). systems. Ontologies can become distrib- extend the Web to include such reasoning.7
As the stores themselves have evolved, uted, as OWL allows ontologies to refer to
the need has arisen for reliable and stan- terms in other ontologies. In this way OWL Data exposure and
dardized data access into the RDF they is specifically engineered for the Web and viral uptake
hold. The SPARQL language (www.w3.org/ Semantic Web.4 So far, we’ve focused on languages, for-
TR/rdf-sparql-query), now in its final review OWL is seeing increased adoption but malisms, standards, and semantics. For this
98 www.computer.org/intelligent IEEE INTELLIGENT SYSTEMSwe make no apologies. The Semantic Web bases on proteins, but we can’t provide the The issue for a Semantic Web built from
can’t exist without carefully developed and URI for a Uniprot protein and then simply these conventions is to know when parts
agreed standards, just as the existing Web read off or determine its properties. Rather, need revision.
couldn’t have existed without HTTP, HTML, the server passes us a zipped bundle of data This brings us to an often quoted concern
and XML. But languages and standards are downloaded as a blob. Moreover, the Life about the Semantic Web—the cost of ontol-
of no consequence without uptake, and Science Identifier (http://lsid.sourceforge. ogy development and maintenance. In some
uptake requires increasing the amount of net) naming-scheme standards that life areas, the costs—no matter how large—will
data exposed in RDF. (We identify RDF scientists use aren’t HTTP compatible. A be easy to recoup. For example, an ontology
because of the often-encountered principle process is needed that routinely gives URIs will be a powerful and essential tool in well-
of least power—the less expressive the to such objects and entrusts their manage- structured areas such as scientific applica-
language, the more reusable the data.) ment to individuals and communities who tions. In certain commercial applications,
Uptake is about reaching the point where care about consistent and explicit reference the potential profit and productivity gain
serendipitous reuse of data, your own and methods. from using well-structured and coordinated
others’, becomes possible. We’ve mentioned vocabulary specifications will outweigh the
the development in the life sciences. Expe- Ontology development and sunk costs of developing an ontology and
rience suggests that an incubator commu- management the marginal costs of maintenance.
nity with a pressing technology need is an The challenges here are real. The ontolo- In fact, given the Web’s fractal nature,
essential prerequisite for success. In the gies that will furnish the semantics for the those costs might decrease as an ontology’s
original Web, this community was high- Semantic Web must be developed, man- user base increases. If we assume that ontol-
energy physicists who needed to share large aged, and endorsed by committed practice ogy building costs are spread across user
document sets. It’s easier to mobilize 10 communities, the number of ontology engi-
percent of a small but focused community neers required increases as the log of the user
than 10 percent of the general populace— community’s size. The amount of building
these early adopters are critical. The ontologies that will furnish time increases as the square of the number of
It’s also instructive to consider typical engineers. These are naïve but reasonable
Semantic Web projects of the past five years. the semantics for the Semantic assumptions for a basic model. The conse-
They demonstrate a distinctive set of charac- quence is that the effort involved per user in
teristics. Typically, they generate new ontolo- Web must be developed, building ontologies for large communities
gies for the application domain—whether it’s gets very small very quickly.10
information management in breast diseases8
or computer science research.9 They either
managed, and endorsed by Not all ontologies have the same charac-
teristics and, in general, we can distinguish
import legacy data or else harvest and rede- deep from shallow ontologies. Deep ontolo-
posit it into a single, large repository. Then
practice communities. gies are often those encountered in science
they carry out inference on the RDF graphs and engineering, where considerable efforts
held within the repositories and represent go into building and developing the concep-
the information using a custom-developed communities. Whether the subject is mete- tualization. For domains such as proteomics
interface. orology or bank transactions, proteins or and medicine, the ontology is in a very real
These projects have been important prov- engine parts, we need concept definitions sense the data of interest. This becomes
ing grounds for a number of techniques and that we can use. apparent when we use an ontology to clas-
methods. They show how to facilitate har- Although some denotations are more sify complex sets of properties as constitut-
vesting and semantic integration by using persistent than others, we must recognize ing certain sorts of object.
ontologies as mediators. They have served that they aren’t fixed over all time. Even Shallow ontologies comprise relatively
as a development context for RDF stores terms used to classify medical diseases few unchanging terms that organize very
and a whole range of important Semantic change as new procedures and understand- large amounts of data—for example, terms
Web middleware. In general, however, they ing emerges. We need to regard such ontolo- such as customer, account number, and
lack real viral uptake. Moreover, in most gies as living structures. Some might endure overdraft used in banking and financial
cases, we aren’t able to look up a URI and over long periods—for example, terms contexts or the basic relations that define
have the data returned. The data exposure describing the elements of the periodic geospatial information. Some might argue that
revolution has not yet happened. table. Others are much more volatile: the we’ve spent rather too much time extolling
URIs provide our symbol grounding in 18th-century concept of phlogiston doesn’t the virtues of deep ontologies at the expense
the Web. An RDF triple as a triple of URIs have a place in a modern ontology of chem- of the shallow ones that deliver very large
should dereference to terms whose mean- istry, but it was once thought to be essential amounts of reusable data. Shallow ontolo-
ings are defined in ontologies. Often, how- to explaining combustion and other chemi- gies require effort but over much simpler
ever, the URIs refer to objects that aren’t so cal reactions. Communities and practice sets of terms and relations.
defined. will change norms, conceptualizations, and
Consider a life science example: Uniprot terminologies in complex and sociologi- Folksonomies: Web-scale tagging
(www.ebi.uniprot.org/index.shtml) is the cally subtle ways. We shouldn’t be surprised The complexity of deep ontologies has
world’s most comprehensive set of data- or attempt to resist these reformulations. led some to eschew ontologies altogether in
MAY/JUNE 2006 www.computer.org/intelligent 99favor of a different approach. Folksonomies potential tasks in a domain, or to the opera- This next wave of data ubiquity will pre-
are a development generating considerable tion of context.14 This perception might be sent us with substantial research challenges.
interest at the moment. They represent a related to the idea of developing a single How do we effectively query huge numbers
structure that emerges organically when consistent Ontology of Everything—like of decentralized information repositories of
individuals manage their own information Cyc,15 for example. Such a wide-ranging varying scales? How do we align and map
requirements. Folksonomies arise when a and all-encompassing ontology might well between ontologies? How do we construct
large number of people are interested in have interesting applications, but it clearly a Semantic Web browser that effectively
particular information and are encouraged won’t scale and its use can’t be enforced. visualizes and navigates the huge connected
to describe it—or tag it (they may tag self- If the Semantic Web is seen as requiring RDF graph? How do we establish trust and
ishly to organize their own content retrieval widespread buy-in to a particular point of provenance of the content?
or altruistically to help others). Rather than view, then it’s understandable that emer- Provenance—that is, the when, where, and
a centralized form of classification, users gent structures like folksonomies begin to conditions under which data originated—has
can assign keywords to documents or other seem more attractive.16 But this isn’t a become a key requirement in a range of appli-
information sources. Semantic Web requirement. Ontologies cations. We might well need the help of
Well-known examples of applications are a rationalization of actual data-sharing researchers in areas as diverse as social
that harness and exploit tagging are Flickr practice. We can and do interact, and we do network analysis17,18 and epidemiology to
(www.flickr.com, a photography publica- it without achieving or attempting to achieve understand how information and concepts
tion and sharing site) and del.icio.us (http:// global consistency and coverage. Ontologies spread on the Web and how to establish
del.icio.us, a site for sharing bookmarks). are a means to make an explicit commitment their provenance and trustworthiness.
These applications, driven by decentralized to shared meaning among an interested com- We must not lose sight of the fact that
communities from the bottom up, are some- the Web, and indeed many of our most
times called Web 2.0 or social software. important digital environments, depends
Tagging on a Web scale is certainly an fundamentally on certain general assump-
interesting development. It provides a poten- Ontologies are attempts to more tions about social behavior. The Web relies
tial source of metadata. The folksonomies on people serving useful content; it relies
that emerge are a variant on keyword searches. carefully define parts of the on content generally being on the end of
They’re an interesting emergent attempt at links. We also require that people observe
information retrieval. But folksonomies data world and to allow copyright rules. Creative Commons (www.
serve very different purposes from ontolo- creativecommons.org) is an RDF-based
gies. Ontologies are attempts to more care-
fully define parts of the data world and to
interactions between data representation of copyright policy to facili-
tate and maximize appropriate reuse.
allow mappings and interactions between Policy-aware research takes this further,
data held in different formats. Ontologies
held in different formats. attempting to express the civic rules of
refer by virtue of URIs; tags use words. behavior expected in a Semantic Web
Ontologies are defined through a careful, environment.
explicit process that attempts to remove munity, but anyone can use these ontologies The critical factors that led to the Web’s
ambiguity. The definition of a tag is a loose to describe their own data. Similarly, anyone success will also be important to the suc-
and implicit process where ambiguity might can extend or reuse elements of an ontology cess of our Semantic Web enterprise. As
well remain. The inferential process applied if they so wish. we’ve seen, some of these factors are social;
to ontologies is logic based and uses opera- others have their origin in elementary and
tions such as join. The inferential process The next wave fundamental design decisions about the
used on tags is statistical in nature and The Semantic Web we aspire to makes Web’s architectural principles. For exam-
employs techniques such as clustering. substantial reuse of existing ontologies and ple, the URL concept embodied the princi-
This doesn’t mean that tags will always data. It’s a linked information space in which ple that every Web address is equal and all
replace shallow ontologies. Where a per- data is being enriched and added. It lets content one jump away. Other critical fea-
ceived need for ontologies exists, light- users engage in the sort of serendipitous tures included the ability to let links fail (the
weight but powerful ones do emerge and reuse and discovery of related information 404 error).
are widely used. Two examples are Friend- that’s been a hallmark of viral Web uptake. A great deal of the success relates to
of-a-Friend11 and associated applications We already see an increasing need and a what we might call the ladder of authority.
such as Flink.12 This fits in general with rising obligation for people and organiza- This is the sequence of specifications (URI,
calls for the dual and complementary tions to make their data available. This is HTTP, RDF, ontology, and so on) and reg-
development of Semantic Web technologies driven by the imperatives of collaborative isters (URI scheme, MIME Internet content
and technologies that exploit the Web’s self- science, by commercial incentives such as type, and so on), which provide a means for
organization.13 making product details available, and by a construct such as an ontology to derive
Some people perceive ontologies as top- regulatory requirements. We believe this meaning from a URI. Another example is
down, somewhat authoritarian constructs— could bring about a revolution in how, for the construction of a standards body that’s
unrelated, or only tenuously related, to example, scientific content is managed been able to promote, develop, and deploy
people’s actual practice, to the variety of throughout its life cycle. open standards.
100 www.computer.org/intelligent IEEE INTELLIGENT SYSTEMST hese reflections lead us to ask how
we understand the present Web and what
data mining tools, approaches to inference,
ontological engineering and knowledge rep-
resentation. All of these are fundamental to
Nigel Shadbolt is a
professor of artificial
intelligence in the
developments we anticipate. This is a deep pursuing a Web Science agenda and realizing School of Electronics
and Computer Sci-
question, and we believe the history of sci- the Semantic Web. ence at Southampton
ence has something to teach us here. There University. Contact
was a time when our understanding of the him at nrs@ecs.
world was either a purely philosophical, soton.ac.uk.
reflective exercise or else craft-based and
References Tim Berners-Lee is
rooted in hard-won experience. Empirical the director of the
methods eventually gave rise to the branches 1. V. Bush, “As We May Think,” Atlantic World Wide Web
of natural philosophy that became physics, Monthly, vol. 176, no. 1, 1945, pp. 101–108. Consortium, a senior
chemistry, and biology. Traces of this legacy researcher at the
2. T. Berners-Lee, J. Hendler, and O. Lassila, Massachusetts Insti-
can still be found: until quite recently the tute of Technology’s
study of physics at Oxford was termed “The Semantic Web,” Scientific Am., May
2001, pp. 34–43. Computer Science
experimental philosophy. More recently, and Artificial Intelli-
areas that were once considered amenable 3. T. Berners-Lee, R.T. Fielding, and L. Masin- gence Laboratory, and a professor of com-
ter, “Uniform Resource Identifier (URI): puter science in the Department of Electron-
only to analytic thought—areas such as
Generic Syntax,” IETF RFP 3986 (standards ics and Computer Science at Southampton
epistemology and logic—are to some University. Contact him at timbl@w3.org.
track), Internet Eng. Task Force, Jan. 2005;
extent operationalized in computers and www.ietf.org/rfc/rfc3986.txt.
computer infrastructures. Knowledge rep- Wendy Hall is a
resentation and ontology engineering are 4. J.A. Hendler, “Frequently Asked Questions professor of com-
on W3C’s Web Ontology Language (OWL),” puter science in the
about trying to capture aspects of shared School of Electronics
conceptualizations. W3C, 2004; www.w3.org/2003/08/owlfaq.
and Computer Sci-
As we build ever more complex compu- 5. G. Shafer, “Causal Logic,” Proc. 13th Euro- ence at Southampton
tational artifacts and information infrastruc- pean Conf. Artificial Intelligence (ECAI 98), University. Contact
John Wiley & Sons, 1998, pp. 711–720; her at wh@ecs.soton.
tures, we observe that large-scale behavior
www.glennshafer.com/assets/downloads/ ac.uk.
emanates from small-scale and local regu-
article62.pdf.
larity. We need engineering methods to
ensure that our structures conform to reli- 6. Z. Huang and H. Stuckenschmidt, “Reason-
able and repeatable design requirements. ing with Multi-Version Ontologies: A Tem-
We need scientific analysis to understand poral Logic Approach,” The Semantic Web— 13. G.W. Flake, D.M. Pennock, and D.C. Fain,
ISWC 2005: 4th Int’l Semantic Web Conf., “The Self-Organized Web: The Yin to the
and predict the behaviors that result. When Semantic Web’s Yang,” IEEE Intelligent Sys-
LNCS 3729, Springer, 2005; www.cs.vu.
we build new opportunities for interaction, nl/~heiner/public/ISWC05a.pdf. tems, July/Aug. 2003, pp. 72–86.
we’re engaged simultaneously in a syn-
thetic and an analytic project. New rules of 7. K. Baclawski and T. Niu, Ontologies for 14. K. Sparck-Jones, “What’s New about the
interaction such as peer-to-peer protocols Bioinformatics, MIT Press, 2005. Semantic Web? Some Questions,” SIGIR
Forum, vol. 38, no. 2, 2004; www.acm.org/
result in new macro behaviors—behaviors 8. S. Dasmahapatra et al., “Facilitating Multi- sigir/forum/2004D/sparck_jones_sigirforum_
we can exploit and also analyze. These Disciplinary Knowledge-Based Support for 2004d.pdf.
micro rules can occur at different levels of Breast Cancer Screening,” Int’l J. Healthcare
abstraction—the rules of Wikipedia are Technology and Management, vol. 7, no. 5, 15. D.B. Lenat, “Cyc: A Large-Scale Investment
2006, pp. 403–420; http://eprints.ecs.soton. in Knowledge Infrastructure,” Comm. ACM,
beguilingly simple but lead to overall vol. 38, no. 11, 1995, pp. 32–38.
ac.uk/8955/01/ijhtm03-miakt.pdf.
coherence. Local-scale changes in Web
architectures and resources can lead to 9. N. Shadbolt et al., “CS AKTive Space, or 16. C. Shirky, “Ontology Is Overrated: Cate-
large-scale societal and technical effects. How We Learned to Stop Worrying and Love gories, Links and Tags,” 2005; www.shirky.
How so? the Semantic Web,” IEEE Intelligent Systems, com/writings/ontology_overrated.html.
vol. 19, no. 3, 2004, pp. 41–47.
We expect the developments, method- 17. D.J. Watts, P.S. Dodds, and M.E.J. Newman,
ologies, challenges, and techniques we’ve 10. T. Berners-Lee, “The Fractal Nature of the “Identity and Search in Social Networks,” Sci-
discussed here to not only give rise to a Web,” working draft, 1998–2005; www.w3. ence, vol. 296, 2002, pp. 1302–1305.
Semantic Web but also contribute to a new org/DesignIssues/Fractal.html.
18. G. Kossinets and D.J. Watts, “Empirical
Web Science—a science that seeks to de- Analysis of an Evolving Social Network,”
11. D. Brickley and L. Miller, “FOAF Vocabulary
velop, deploy, and understand distributed Specification,” working draft, 2005, http:// Science, vol. 311, 2006, pp. 88–90.
information systems, systems of humans xmlns.com/foaf/0.1.
and machines, operating on a global scale.
AI will be one of the contributing disciplines. 12. P. Mika, “Flink: Semantic Web Technology
for the Extraction and Analysis of Social Net-
AI has already given us functional and logic works,” J. Web Semantics, vol. 3, no. 2, 2005, For more information on this or any other com-
programming methods, ways to understand pp. 211–223; www.websemanticsjournal.org/ puting topic, please visit our Digital Library at
distributed systems, pattern detection and ps/pub/2005-20. www.computer.org/publications/dlib.
MAY/JUNE 2006 www.computer.org/intelligent 101You can also read
