Informetric theories and methods for exploring the Internet: an analytical survey of recent research literature

Library Trends, Wntr, 2002 by Judit Bar-Ilan, Bluma C. Peritz

Cui (1999) used citation analysis to rank health Web sites. Again, the hypertext links were viewed as citations. The study analyzed the links appearing on the homepages of the libraries of the top U.S. medical schools, as compiled and published by U.S. News and World Report.

Lawrence, Giles, & Bollacker (1999) took a completely different approach to citation analysis on the Web. Instead of studying hypertext links as analogues of citations in the academic world, they looked for citations in the classical sense, and their "Autonomous Citation Indexing" (ACI) system can automatically create a citation index from literature in electronic format. The rationale behind this project is that an increasing number of authors,journals, institutions, and archives make research articles available on the Web, mainly in PDF or Postscript formats. ACI is implemented for computer science literature at the "ResearchIndex" site (http://www.researchindex.com/). The system allows one to search articles and citations. When searching for citations, it provides citation context (in the citing article), citation statistics, and links to the citing articles. For full-text articles the system also displays the exact bibliographic reference, the list of citations and the list of references, similar documents (textual similarity), and related documents (based on cocitations). The user interface needs some improvement.

Garfield (1999) related to this project in an address delivered at a symposium in honor of Manfred Kochen: "... without aposteriori human intelligence, the Internet will remain at best a mixed blessing. Artificial intelligence will help but not suffice.... The Internet has made it practical for future citation index databases to generate annotated bibliographies and reviews containing contextual quotations based on autonomous citation indexing. To see how this works in the field of computer science just go to www.researchindex.com."

A recent paper (Goodrum, McCain, Lawrence, & Giles, 2001) compares the ISI SCISEARCH Citation Index to ACI in the area of computer science. A major difference between the two systems is that ACI indexes PDF and Postscript formatted publication on the Web, while SCISEARCH indexes only a selected list of journals in the area.

Cocitation and Coword Analysis

The objective of the study carried out by Larson (1996) was to explore the applicability of classical cocitations on the Web, when citations are substituted with hyperlinks. He carried out a cocitation analysis of a set of Earth Science related Web sites. The starting point were two authoritative sites on the topic. The list of pages pointing to these two sites was retrieved using Alta Vista, and the links appearing in the relevant pages were extracted. This set underwent a second round of relevance judgment by Larson, and a set of thirty-four "core" pages was created. Again, Alta Vista's link option was utilized to retrieve the number of URLs linking to each of the 544 cocitation pairs. The data were converted to a correlation matrix and multidimensional scaling (MDS) was used to create the cocitation map. Larson concluded, "the mappings ... seem to produce quite clear, reasonable and interpretable results."