Co-occurrence matrices and their applications in information science: Extending ACA to the Web environment

摘要: Co-occurrence matrices, such as cocitation, coword, and colink have been used widely in the information sciences. However, confusion controversy hindered proper statistical analysis of these data. The underlying problem, our opinion, involved understanding nature various types matrices. This article discusses difference between a symmetrical cocitation matrix an asymmetrical citation well appropriate techniques that can be applied to each respectively. Similarity measures (such Pearson correlation coefficient or cosine) should not but derive proximity matrix. argument is illustrated with examples. study then extends application co-occurrence matrices Web environment, which available data thus collection methods are different from those traditional databases Science Citation Index. A set collected Google Scholar search engine analyzed by using both multivariate new visualization software Pajek, based on social network graph theory. © 2006 Wiley Periodicals, Inc.