Copy the page URI to the clipboard
Dietze, Stefan; Benn, Neil; Domingue, John; Conconi, Alex and Cattaneo, Fabio
(2009).
DOI: https://doi.org/10.1007/978-3-642-10543-2_9
Abstract
The increasing availability of multimedia (MM) resources, Web services as well as content, on the Web raises the need to automatically discover and process resources out of distributed repositories. However, the heterogeneity of applied metadata schemas and vocabularies – ranging from XML-based schemas such as MPEG-7 to formal knowledge representation approaches – raises interoperability problems. To enable MM metadata interoperability by means of automated similarity-computation, we propose a hybrid representation approach which combines symbolic MM metadata representations with a grounding in so-called Conceptual Spaces (CS). In that, we enable automatic computation of similarities across distinct metadata vocabularies and schemas in terms of spatial distances in shared CS. Moreover, such a vector-based approach is particularly well suited to represent MM metadata, given that a majority of MM parameters is provided in terms of quantified metrics. To prove the feasibility of our approach, we provide a prototypical implementation facilitating similarity-based discovery of publicly available MM services, aiming at federated MM content retrieval out of heterogeneous repositories.