WordSieve is most appropriate for agents which monitor ongoing user document accesses, rather than for batch document indexing. Although it does not need to remember or refer to previously accessed documents to build a user access profile, it does need the documents to be presented in a sequence in which the user would actually access them. This is not a problem in a real-time agent environment, but would be in a large, unorganized corpus. Although TFIDF does not have this limitation, it cannot take advantage of the order of document use when that information is available.