Reasoning about Form and Content for Multimedia IR∗

A multimedia retrieval model should involve a combination of concepts and techniques from 1) digital signal processing, which deals with the physical aspect (form) of documents and contributes algorithms for assessing the similarity between them, and 2) symbolic processing, which contributes conceptual models for representing the semantic aspect (content) of documents and algorithms for reasoning about it. Features pertaining to either aspect should be addressable from within the same expressions of a query language: however, this is not provided for in existing models. The works we have been doing in the last years can be seen as converging to this larger picture, of which we give a sketch here. The kernel of our model is a Description Logic (DL) that departs from standard DLs, such as Mirtl, in two important ways: