We have identified the equation R = c ( A + E * 1) R as an effective approach for defining the significance (PageRank) of Web pages.

Page et al [2] make several additional observations:

• This version of PageRank is not very susceptible to manipulation by manipulation to owners of Web pages. [p. 12]

• In computing eigenvectors (PageRank), the iterative algorithm can be made quite efficient by an appropriate initial guess. 1998 for 75 million URLs, [2] indicates that one iteration can be done "in about 6 minutes on a typical workstation", and the how process for computing PageRank (including handling dangling links) takes about 6 hours. [p. 7]

• If a surfer choses any Web page at random for a new start, then each entry of the starting vector E could be set the same constant. However, this gives relatively high ranking to commonly-linked pages, such as "copyright notices, disclaimers, and highly interlinked mailing list archives". [p. 11]

• If a surfer chose a specific starting point (Page et all experimented with the Netscape home page and the home page of John McCarthy — a famous computer scientist), then these pages get the highest PageRank. This could allow a personalized PageRank, with priorities given by each user. This approach did not seem to give excessive rankings to copyright notices and the like. [pp. 11-12]

created: 10 February 2007 last revised: 16 February 2007

previous  next