Chad Hogg
Review of "Efficient Identification of Web Communities" by Flake, Lawrence, and Giles

	This paper contains a definition of a web community as a subset of the pages in the web such that pages in the community are linked more frequently to each other than to pages outside the community.  A method is explored for discovery of web communities based on a maximum-flow network analysis model.  Experimental results are provided for three focused crawls that use this method of discovery.
	There are several great points about this paper.  The addition of an appendix on the ISA algorithm and its time complexity was a nice touch.  The authors state clearly that they have not yet been able to analytically prove some of their results, but that they have shown them to be experimentally consistent.  This should not be remarkable, but many papers make claims without offering a note that they are as yet unproven.  The idea suggested here is entirely novel in the scope of web resource analysis, and has fascinating potential.
	The authors do not provide enough detail and references on the process of network flow analysis, which is the crux of their new technique.  In a field that may benefit from this point-of-view that is likely to be foreign to researchers, it would be helpful to fully explain the background.  It is not entirely clear how the theoretical framework the authors suggest is used to develop the experimental results discussed, nor why only pages within 2 link epochs of the seed set are considered.  Finally, the authors state that Kleinberg's HITS algorithm is "strongly related to ... methods used by the Google search engine."  Although Google has been known to use an algorithm that ranks pages based on social network analysis, this algorithm is actually quite different from that of HITS.
	This is a paper worthy of publication, but it would be helpful if the authors were able to solidify the connections between traditional web mining approaches and their network flow analysis model.