Currently, the authors investigate future document management services. One of these services is accessing dynamic document spaces, i.e. improving the access to document spaces which are frequently updated. This process is rather computational intensive. This paper describes the research conducted on software development for massively parallel processors. A prototype has been built which processes streams of information from specified newsgroups and transforms them into personal information maps. Although this technology does speed up the training part compared to a general purpose processor implementation its real benefits emerges with larger problem dimensions because of the scalable approach.