Buckingham Shum, S. (2004) Contentious, Dynamic, Information-Sparse…

Buckingham Shum, S. (2004) Contentious, Dynamic, Information-Sparse Domains...and
Ontologies? IEEE Intelligent Systems, Jan/Feb 2004 (Trends & Controversies), 80-81


  Contentious, Dynamic, Information-Sparse Domains...and
                        Ontologies?

                Simon Buckingham Shum, The Open University
  As journalists and politicians know too well, sometimes simply asking a
particular question is enough to make a point. Swinging the spotlight onto issues
that people rarely discuss is a good mental hygiene exercise for the Semantic
Web at this point in its young life.
  It's about time to consider a 6-month health-check for this increasingly active
toddler, where we verify more rigorously than its doting parents can manage that
the child is indeed seeing and hearing clearly.


                                 The challenge

    In the spirit of this collection of essays, I adopt a contentious stance to make
my point. I focus on arguably the most controversial application proposed for the
Semantic Web--namely, knowledge management in organizations. (The fact that
it's not always seen as controversial demonstrates that all is not well). Now that
KM has begun to mature, those who didn't see through the technocentric hype in
the early days are rapidly realizing what others sought to emphasize above the
roar of computing vendors and AI researchers revving their engines: the
dominant metaphor in much real-world knowledge work is not the abstracted,
indexed, textual knowledge object, but rather the situated, embodied
sensemaking process.
    Organizations need to make sense of rapidly changing environments where the
questions (never mind the answers) might not be clear, where organizational and
other politics make certain ideas untenable, where incomplete knowledge and
backgrounds make understanding perspectives central, where expertise must be
combined and reconfigured in the light of discussion, and where information must
be interpreted in a timely manner. Contentious, dynamic domains, requiring good
enough, timely, collective sensemaking on incomplete, multimedia information--
quite a sobering reality check for any prospective knowledge infrastructure to
confront. However, such considerations are well documented by leading thinkers
such as Karl Weick1 and John Seely Brown.2


                               Enter ontologies

  What do ontologies require to operate?

* Consensus: an ontology is an agreed conceptualization of how the world is.
* Hand-crafting: we can't automatically construct nontrivial ontologies at
  present.
* Maintenance: our world view changes, and so must our ontologies, or we're
  modeling a fiction.
* Textual expression: "if you can't type it, it doesn't mean anything" is not a
  promising precondition for a world where meaning is clothed in multiple
  modalities.

  On the face of it, ontologies don't shape up as promising contenders for the
knowledge infrastructure backbone.
                   May many Semantic Webs bloom

   Clearly, ontology-based knowledge representation is for stable, well-
understood problems with well-known problem-solving methods. Organizations
have a huge requirement for database integration and machine-machine
interoperability. In such domains, we can even trust ontology-based agents to
negotiate autonomously within the well-defined boundaries. The clear implication
is that if deployed for KM, we're talking about innumerable Semantic Webs--
islands of coherence whose members subscribe to that world view enough to
publish and consume services with a degree of trust.
   But away from these quiet backwaters, on the wilder rapids of organizational
sensemaking, the brittle ontological canoe might snap. So, many conclude that
this vessel is simply not the one to ride when shooting these rapids.


                     Making ontologies less brittle

   Semantic Web adherents, however, keep the faith and are demonstrating how
we can make the canoe more flexible. To adopt a less confrontational stance, I'm
more than happy to recognize that this is where we find some of the most
interesting work at present. The emphasis by people such as Jim Hendler on
"scruffy" reasoning is absolutely right, as exemplified in the Advanced Knowledge
Technologies project (www.aktors.org), some of whose work I'll turn to next.
   Simple reasoning over multiple databases might prove an interesting strategy.
The ontologies assist with the data capture, data integration, and the reasoning
service definitions, resulting in added value through productive combinations of
previously disparate data sources (see
www.aktors.org/technologies/csaktivespace). Language technologies hold at least
one key for addressing the capture bottleneck that plagues any formal
representation. Agents can harvest text from the Web or other live sources and
interpret it ontologically to keep ontologies populated with up-to-date instances.3
This doesn't, however, revise the ontological structure itself, only instances.
Services that we can rapidly define, publish, and configure within a corporate
Semantic Web intranet can exploit distributed expertise.4 In principle, service
providers with completely different ontologies could still interoperate, but we
know how hard making this work really is, and in business practice, the case is
not yet proven.


   Mixing formality and informality to support collaborative
                       knowledge work

   My own work deals with collaborative analysis and sensemaking. A large
"collaboration technologies graveyard" exists of over-engineered systems that
didn't recognize the target end-user community's work practices, and so were
dumped.
   My strategy combines semiformal and formal semantics with the informality
inherent in collaborative work. For example, the e-Science CoAKTinG project
(www.aktors.org/coakting) combines free-text instant messaging, visual online
presence cues, and "dialogue mapping" using a semiformal concept mapping tool
with audio and video records of virtual meetings. There's also a project ontology
of the people, events, technologies, and organizations to provide integration with
other ontology-based resources and services.5 How people want to communicate
leads the requirements that the tools deliver. If the tools don't help the work you
have to do, then you just don't use them.
    Supporting conflicting interpretations and perspectives

   A second example is the Scholarly Ontologies project. What can the Semantic
Web offer in domains where there is little consensus, no master view, and
conflicting perspectives? In the Scholarly Ontologies project
(www.kmi.open.ac.uk/projects/scholonto), we're developing a semantic digital
library server that provides services for researchers whose business is, of course,
constructing and debating world views. Our tools provide a discourse ontology for
making, extending, and challenging "claims." Although we still want to deliver
useful knowledge services, we must relax many of our normal knowledge
engineering assumptions for nonengineers who want to construct distributed,
collaborative knowledge bases.6 Although currently being applied to research
literatures, the underlying approach applies to any domain where it's as
important to capture principled disagreement as it is to capture consensus.


                                    Diagnosis

   So, health-check over; is the infant okay, doctor? It's probably too early to tell.
Some early heart murmurs might cause concern, but they could pass with time.
The key is not to smother the child in cotton wool. The recommended regime is
lots of exercise out in the dirt with other children to make sure that the child is
properly socialized and develops the right immunities in the rough and tumble
competitive world. If in the end, no one will play with him, you'll only have
yourself to blame, Mrs. Ann O'Tate. Next, please.

Acknowledgements
  The views in this piece are my own and are not necessarily shared by my
colleagues in the cited projects and publications. I gratefully acknowledge the
support of the UK EPSRC, who fund the ScholOnto, AKT, and CoAKTinG projects.

References
 1. K.E. Weick, Sensemaking in Organizations, Sage Publications, 1995.
 2. J.S. Brown and P. Duguid, The Social Life of Information, Harvard Business
    School Press, 2000.
 3. F. Ciravegna, "Challenges in Information Extraction from Text for Knowledge
    Management," IEEE Intelligent Systems, vol. 16, no. 6, Nov./Dec. 2000, pp.
    84­86.
 4. E. Motta et al., "IRS-II: A Framework and Infrastructure for Semantic Web
    Services," Proc. 2nd Int'l Semantic Web Conf. (ISWC 2003), Lecture Notes in
    Computer Science vol. 2870/2003, Springer-Verlag, pp. 306­318.
 5. S. Buckingham Shum et al., "CoAKTinG: Collaborative Advanced Knowledge
    Technologies in the Grid," 2nd Workshop on Advanced Collaborative
    Environments, Held in conjunction with 11th IEEE Int'l Symp. High-
    Performance Distributed Computing, 2002. Available as Advanced Knowledge
    Technologies Project ePrint 156: http://eprints.aktors.org/archive/00000156
 6. V. Uren et al., "Interfaces for Capturing Interpretations of Research
    Literature," Workshop on Distributed and Collaborative Knowledge Capture
    (DC-KCAP), Oct. 26, 2003, Held in conjunction with Int'l Conf. Knowledge
    Capture (K-CAP 03), 2003. Available as Technical Report 130, Knowledge
    Media Institute, Open University, UK:
    http://kmi.open.ac.uk/publications/papers/kmi-tr-130.pdf