describing an original graph with a plurality of nodes and links, where said nodes represent information documents and said links represent pointers from one document to another;

forming a metagraph from said original graph by identifying a set of strongly connected components (SCC), replacing each SCC with a metanode, and retaining links between metanodes;

determining a gain for each SCC;

determining from said metagraph which SCCs are source SCCs;

increasing the gain of at least one source SCC by modifying an adjacency matrix representing link weights of the original graph, until the following three conditions are satisfied, thus creating a modified adjacency matrix:

(i) all source SCCs whose gain have been increased have a common gain;

(ii) the common gain is greater than that of any non-source SCC; and

(iii) said common gain is greater than a gain of any source SCC whose gain has not been so increased;

determining an importance measure for said plurality of nodes by performing a link analysis of said original graph using said modified adjacency matrix to compute link analysis node weights;

ranking two of the plurality of nodes with said importance measure; and

one of outputting or displaying a result relating to said importance measure,

wherein said step of determining the gain of each SCC comprises, for each SCC:

determining a second adjacency matrix representing intra-SCC link weights for a corresponding SCC; and

determining a dominant eigenvalue of said adjacency matrix; and

setting the gain based on said dominant eigenvalue.