Visual Analytics - Mapping Scientific Frontiers: The Quest for Knowledge Visualization - page 330

Challenge 8 Automatic labeling. The ability to generate informative and accurate
labels boils down to classification and categorization skills. Is it possible to pass on
such skills to algorithms?
The challenge is to choose labels that will make the most sense to the intended
audience. Studies have found that human beings tend to choose relatively broader
label terms than algorithms that are configured to differentiate groups of co-cited
references. A promising strategy is to make use of domain-specific knowledge and
adapt to the knowledge level of the audience.
Challenge 9 Individual differences. One user's daydream could be the other's
nightmare. The same visual-spatial configuration may send different messages to
different individuals. Personalization is a related issue.
The situation is similar to that of Challenge 8. The challenge boils down to how to
maintain an effective communication between technology and human information
and reasoning needs. Effectively incorporating and accessing background knowl-
edge is a long-term challenge that has been addressed by artificial intelligence.
Challenge 10 Ethical constraints. Moving from information-oriented search tools
to knowledge-oriented ones shifts the focus from documents to scientists and scien-
tific networks. The knowledge of invisible colleges has been privileged. Sometimes
this is the knowledge that distinguishes an expert from a new comer. Re-think
the famous quotation from Francis Bacon ( 1561-1626 ), “Knowledge is power.”
What are the ethical issues we need to take into account?
Much of the competitive edge results from the asymmetric procession of
knowledge. The techniques of making atomic bombs and cloning human beings are
just few examples of what decisions the society as a whole has to make to ensure
the humanity has a healthy future.
In terms of the technical capabilities, I forecasted the following developments in
2002. Which ones have been achieved and which ones are still not reachable?
For the next 3 - 5 years , between 2002 and 2005 , several routes of research and
development are likely to emerge or establish . In particular , the need for expanding
the current coverage of citation databases to include conference proceedings and
patent databases can trigger further research in automatic citation indexing and
large - scale text analysis . The need for timelier disciplinary snapshot should also
drive much research onto this route . Automatic extraction of citation context will
become increasingly popular . Software agents will begin to emerge for summarizing
multiple citation contexts - an important step in resolving the bottleneck of stream-
lining quantitative and qualitative approaches to science mapping . The recent surge
of interest in small - world networks is likely to continue . One can expect to see more
specific studies of scientific networks as small - world networks , including Web - based
resources analysis , traditional citation databases , and patent databases . Research
in small - world networks is likely to draw much of attention to network analysis tools
that can handle large - scale scientific networks . Cross - section comparisons should
increase .