Tracking Latent Domain Knowledge - Mapping Scientific Frontiers: The Quest for Knowledge Visualization - page 225

interleaved phrases of normal science and scientific revolutions. A period of normal
science is typically marked by the dominance of an established framework. The
foundations of such frameworks largely remain unchallenged until new discoveries
begin to cast doubts over fundamental issues - science falls into a period of crises.
To resolve such crises, radically new theories are introduced. New theories replace
with greater explanatory power the ones in trouble in a revolutionary manner.
Science regains another period of normal science.
Kuhn suggested that a paradigm shift in science should lead to a corresponding
change of citation patterns in scientific literature; therefore the study of such patterns
may provide indicators of the development of a scientific paradigm. Indeed, a
number of researchers pursued this line of research since 1970s. For example,
Henry Small studied the movement of highly cited publications on the topic of
collagen as a means of tracking major paradigm shifts in this particular field
(Small 1977 ). White and McCain used INDSCAL to depict changes in author
co-citation maps over consecutive periods (White and McCain 1998 a). We have
started to investigate how information visualization can help us characterize the
dynamics of scientific paradigms (Chen et al. 2001 , 2002 ). In particular, our focus
is on contemporary puzzle-solving topics in science and medicine: What caused
dinosaurs' mass extinction? Are Bovine Spongiform Encephalopathy (BSE) and
the new variant Creutzfeldt-Jakob Disease (vCJD) connected? What powers active
galactic centers, super-massive black holes, or something else?
In this chapter, we introduce an approach to visualizing latent domain knowledge.
We demonstrate how one can accommodate latent domain knowledge and the main-
stream domain knowledge within the same visualization framework. We include two
case studies: Pathfinder network applications and theories of Bovine Spongiform
Encephalopathy (BSE), commonly known as mad cow disease. The rest of the
chapter is organized as follows. First, we outline existing work, including citation
analysis, knowledge discovery, and examples. We then extend our domain visual-
ization approach to visualize latent domain knowledge. We apply this approach to
two cases in which visualizing latent domain knowledge is involved: (1) tracing
applications of Pathfinder networks and (2) connecting a controversial theory of
BSE, mad cow disease, to the mainstream intellectual structure of research in BSE.

7.2 Knowledge Discovery

The advances of information visualization have revived the interest in a number of
challenging issues concerning knowledge tracking. Here we contrast two strands of
research the citation-based paradigm of knowledge discovery and the undiscovered
public knowledge approach. The key prerequisite for the citation-based paradigm is
a target scientific literature that is rich in citations, whereas the undiscovered public
knowledge deals with exactly the opposite situation when citation links are missing
or are considerably rare. A synergy of the two would lead to a more powerful tool
to facilitate knowledge discovery and knowledge management in general.