Somewhat related: a PageRank-inspired approach to web page similarity:

http://www-cs-students.stanford.edu/~glenj/simrank.pdf