The relevance density method in information retrieval

The authors propose a new information retrieval method, the relevance density method (RDM) for selecting relevant documents. The method can be used whenever the documents and the terms are represented by vectors in a multi-dimensional document-term space, such that the vectors corresponding to documents and terms dealing with closely related topics are close to each other. They model relevance as a continuous quantity whose distribution over the document-term space is a probability density. The Bayes rule is used to incorporate evidence about the user´s interests obtained at different stages of retrieval into the density. RDM addresses a long standing problem of responding to users whose information needs are best answered by two or more distinct sets of documents. In addition, RDM can incorporate detailed user models