Component study of co-word analysis

Co-word analysis is a technique proposed to detect the intellectual structure of science by clustering words from scientific literature. It counts the co-occurrence of term pairs in texts, computes the association between terms using special indexes, and clusters terms into various graphical structures. Each graph (map) represents a theme in a research field and the graphs from several continuous periods can map the dynamics of the research field.; This research includes a series of systematic component comparison studies on the co-word analysis technique. On the one hand, a set of co-word analysis tests are conducted using two different strength indexes to compute the association between terms. The indexes being tested include: (i) the inclusion index (I); and, (ii) the equivalence index (E). On the other hand, a set of co-word analysis tests are carried out using different input term units. The input term sets being tested include: (i) terms from the subject heading field of articles (thesaurus terms); (ii) terms from article titles; All the tests are conducted on the same test collection drawn from the INSPEC database and use the same two-pass clustering procedure. Only one of the other components is changed in order to identify the effects of that component on the final results from the test. Outputs of all the tests are evaluated by human experts on some pre-selected criteria (correctness of linkage, specificity level, completeness, centrality, density). Statistical analyses are conducted to determine if the component being tested affects the results of co-word analysis significantly and how they affect the results.; The results of human evaluation and statistical analyses show that both the strength index component and input term component affect the output from co-word analysis in significant ways. The resulting maps from the inclusion index test are much more general than those from the equivalence index. The test with title terms gives the most specific maps while the test with title...