This thesis defines information publishing as a frequent knowledge work activity which includes gathering, filtering and disseminating information. Internet CAFE is a tool to support this activity. CAFE can be used to define an interest profile, retrieve relevant documents from the Internet, filter retrieved documents to select the most interesting ones, and publish a Web site pointing to those documents. The filtering component of CAFE uses a combination of traditional document similarity techniques, and a new evolutionary computing method--the parasitic genetic algorithm. Experimental results indicate that the performance of CAFE is sensitive to the domain of information being filtered. CAFE performs better with technical papers than with documents on topics of general interest. We identify a number of root causes for this sensitivity and propose the implementation of a user profile to assist in filtering out poor documents in general interest domains.