David Gabriel , Universidade do Porto, Portugal
Cristina Ribeiro , Universidade do Porto, Portugal
Session: Design & Evaluation: Representation
Largely due to the facilities provided by the technology, there is currently a large effort put into the organization of large electronic repositories of documents, covering both present-day materials and those of historic interest. It is also technologically feasible to gather heterogeneous documents, and the same collection may include text, images, video and audio items, either independently or as multimedia objects. Professional and lay users have great expectations on the possibilities of searching such materials, partly due to the success of search engines on the Web. Content-based retrieval on multimedia objects is a hard task, however, and results similar to those obtained with text retrieval are not foreseeable in the near future for images or video. Retrieval of multimedia objects must therefore rely on the wealth of associated information, available as metadata of diverse nature.
The paper describes a metadata model where contextual information, in an archival perspective, is combined with the description of contents, as required for content-based search. Contextual information can be valuable for search, and is important when communicating search results, where issues such as date, access rights or age rating may condition the usage of retrieved documents. The metadata model adopts a flexible multi-level approach to both context and content description, to allow for different granularities in the description of materials. It may be impracticable to provide descriptions at the individual item level in the digitization of an historic photo archive. Instead, a top-down approach to the description of collections can be a good compromise. For the video materials of a broadcast station, on the other hand, there is automatically produced information for each recorded shot that can be also accommodated in the model.
The model has been tested in a prototype database system where specialized interfaces are offered for cataloguing and for search. The database contains data from historic, photo and video archives. The prototype also provides tools for filtering information into standard formats for interchange. The relations of the proposed model with existing and emerging metadata standards are presented. An application based on the prototype will be used for testing user behavior with data from an historic archive.