MIND: An architecture for multimedia information retrieval in federated digital libraries
Henrik Nottelmann, Norbert Fuhr · 2001
Introduction Today, people have routine access to a huge number of heterogeneous and distributed digital libraries. To satisfy an information need, relevant libraries have to be selected, the information need has to be reformulated for every library w. r. t. its schema and query syntax, and the results have to be fused. This is an ineffective manual task for which accurate tools are desirable. MIND (which we are currently developing in an EU project) is an end-to-end solution for federated digital libraries which covers all these issues. We start from information retrieval approaches which focus on retrieval quality, but mostly only consider monomedial and homogeneous sources. We will extend these approaches for dealing with different kinds of media (text, facts, images and transcripts of speech recognition) as well as handling heterogeneous libraries (e.g., with different schemas). Another innovation is that MIND also considers non-co-operating libraries which only provide the