We are currently trying to determine exactly what criteria should be used to determine when two paragraphs show a "high similarity." These criteria will be based on example articles that have been linked by human judges. The next step will be to determine how articles can be related to one another through their lexical chains. This stage will most likely consist of a corpus study, considering related articles and their associated lexical chains. Articles that show a similar distribution of chains and chain words would be linked. It would seem that rules for linking articles in the same newspaper should be easily generalizable to linking articles from different newspapers. The process is still in its initial stages, but the results so far have been very promising.