Multidisciplinary research thrusts from co-word analysis

Summary form only given. It is noted that the main documented criticism of co-word analysis has been that co-occurrences were evaluated on index, or key, words of published articles rather than directly on the text, thereby generating sources of bias and error. This has limited the credibility and scope of textual material available to co-word analysis. The author presents a technique that obviates the need for indexed words, uses all words which appear in any type of text, and weights their importance by frequency of occurrence. All single, double, and triple words in a body of text are extracted and analyzed. For each high-frequency content word or word combination W_i, the frequency of co-occurrence of the other high-frequency content words in a limited region around W_i is computed. The co-occurrence frequency is normalized, and clusters of highly co-occurring words are obtained by a mapping procedure. The body of text used described research opportunities in many fields; the high-frequency terms were the pervasive research thrust areas; and the clusters represented pervasive multidisciplinary research thrust areas. For selected pervasive thrust areas, related mid-frequency words (supporting research areas) which had high normalized co-occurrences were obtained