A Multiple Decomposed Approach for Relevant Functions in Information Retrieval

A model on indexing for information retrieval is presented in this paper. As a way to reduce the limitations over relevant functions, we propose an architecture which incorporates multilevel functional decomposition to ascribe the documents based on users' concept spaces. The higher level can obtain taxonomies for documents based on cata-loguers' Boolean indexing opinions. The lower level will provide the degrees for these taxonomies based on cataloguers' authorities. Based on diierent demands from users, two kinds of aggregating relevance approaches are possible in this model, which can avoid the disadvantages of using Dempster's rule in the Dempster-Shafer indexing models.