PhD thesis,
A comparison of self-organizing maps and pathfinder networks for the mapping of co-cited authors
Drexel University, Ph.D., (2003)
Abstract
Author co-citation analysis (ACA) is a methodology for determining the relationships between author pairs as specified by patterns of repeated co-citation. These patterns are visualized through computerized mapping techniques that now include self-organized maps or SOMs (also known as Kohonen feature maps) and Pathfinder networks (PFNETs). In this research, ACA maps of authors in the humanities were produced as both SOMs and PFNETs using an experimental Web-based system called AuthorLink based on author co-citation data from the Arts & Humanities Citation Index for the years 1988-1997. The intent was to test how well each map type corresponded to a set of mental maps of experts, the "gold standard," for the same authors. The mental maps were elicited from 20 experts in the humanities with card sorts. The first research question was: How well do SOMs and PFNETs based on one famous author--here, Plato--correspond to the experts' The second research question was: How well do SOMs and PFNETs based on personally chosen authors correspond to the experts' The third research question was: Is one type of map, SOM or PFNET, preferred by experts? The 20 experts were statistically equal, 11 for the PFNET to 9 for the SOM. However, during interviews with the experts it was suggested that the two types are complementary and that both are needed to do a thorough exploration.
