Tracking Latent Domain Knowledge - Mapping Scientific Frontiers: The Quest for Knowledge Visualization - page 228

Tabl e 7. 1 Seven discoveries of undiscovered public knowledge, all published in the biomedical

literature

Seven examples published
in the biomedical literature Year A Potential cause factors C Disease

1986 Swanson ( 1986a ) Fish oil Raynaud's syndrome

1988 Swanson ( 1988 ) Magnesium Migraine

1990 Swanson ( 1990 ) Somatomedin C Arginine

1994 Smalheiser and Swanson ( 1994 ) Magnesium deficiency Neurologic disease

1996 Smalheiser and Swanson ( 1996a ) Indomethacin Alzheimer's disease

1996 Smalheiser and Swanson ( 1996b ) Estrogen Alzheimer's disease

1998 Smalheiser and Swanson ( 1998 ) Calcium-independent

phospholipase A2

Schizophrenia

Swanson has been pursuing his paradigm since 1986 when he found two sizeable
biomedical literatures: one is on the circulatory effects of dietary fish oil and the
other is on the peripheral circulatory disorder, Raynaud's disease. Swanson noticed
that these two literatures were not bibliographically related: No one from one camp
cited works in the other (Swanson, 1986a , b ). On the other hand, he was pondering
the question that apparently no one had asked before: Is there a connection between
dietary fish oil and Raynaud's disease?
Prior to Swanson's research, no medical researcher had noticed this connection,
and the indexing of these two literatures was unlikely to facilitate the discovery of
any such connections. Swanson's approach can be represented in a generic form.
Given two premises that A causes B (A ! B) and that B causes C (B ! C), the
question to ask is whether A causes C (A ! C). If the answer is positive, the causal
relation has the transitive property. In the biological world, such transitive properties
may not always be there. Therefore scientists must explicitly establish such tran-
sitivity relationships. Swanson suggests once information scientists identify such
possibilities, they should recommend domain experts to validate (Swanson 2001 ).
Swanson's approach focuses on the discovery of such hypotheses from the vast
amount of implicit, or latent, connections. Swanson and Smalheiser ( 1997 )defined
the concept of non-interactive literatures. If two literatures have never been cited
together at a notable level, they are non-interactive - scientists have not considered
both literatures together. In the past 15 years, Swanson identified several missing
links of the same pattern, notably migraine and magnesium (Swanson 1988 ),
and arginine and somatomedin C (Swanson 1990 ). Since 1994, the collaboration
between neurologist Neil Smalheiser and Swanson led to a few more such cases
(Smalheiser and Swanson 1994 , 1996 ). Table 7.1 is a summary of various case
studies. They also made their software Arrowsmith available on the Internet
(Swanson 1999 ).
Swanson's approach relies on the identification of the two premises A ! Band
B ! C. In a large knowledge domain, it is crucial for analysts to have sufficient
domain knowledge. Otherwise, to find two such premises is like searching for nee-
dles in a haystack. Knowledge domain visualization (KDViz) can narrow down the
search space and increase the chance of finding a fruitful line of scientific inquiry.