Presenter:

-------Monica Berko---------

Information from the proposer:

Monica Berko,
Director, IT Applications
National Library of Australia

PictureAustralia

PictureAustralia is a web service based on a metadata index held at the National Library in Canberra which links to pictorial images held on the web sites of participating cultural agencies around Australia. The service has a "hybrid" architecture with a centralised search index and distributed images. The search index for the PictureAustralia service is based on metadata in Dublin Core format.

Users can search the central metadata index at the National Library for their subject of interest, view image thumbnails in their results, then go to a participating agency's web site to view a larger version of the image and order a high resolution copy if needed. Thematic trails have been produced with canned searches of the metadata, to highlight the images and true strengths of collections.

PictureAustralia consists of links to images of all forms of Australiana. Images may be of art, abstract art, fine art, portraits, photographs of people, places and events, posters and three-dimensional art.

Participating institutions are asked to create a web page for each of their images which includes in the HEAD tag an agreed set of Dublin Core Metatags describing the image. Every month the National Library gathers this metadata from each participating agency using harvesting software. The metadata is stored at the Library in an XML repository. An index is built using this metadata which PictureAustralia users' search when using the service. The thumbnail images in search results are retrieved from participants' web sites dynamically and are not duplicated in the repository. The software being used by the Library to gather the metadata and build the search index is MetaStar from Blue Angel Technologies, see: http://www.blueangeltech.com.

This architecture is simple and scalable and the technical investment needed to join the PictureAustralia service is low especially if image collections are already web enabled.

See http://www.pictureaustralia.org/

Panel Discussion Topics

Dublin Core - DC provides a "lingua franca" and common metadata schema that all types of cultural institutions, Libraries, Museums, Galeries, Archives etc can map their metadata to. But there are some issues with combining metadata from different organisations:

Subject thesauri - Picture Australia participants use different subject thesauri and mapping of terms to a common thesaurus has not been attempted. This limits the ablity to build browse interfaces that might use the subject thesaurus to navigate the PA information space.
Relevance ranking - We have found that this doesn't work very well when searching across creator, title, subject fields (the default search in PA) for images catalogued by different organisations with different practices.

Harvesting - Our full re-harvesting model won't scale. It would be easier to implement incremental harvesting if each PA participant used a database for their metadata records. However, most do not have the technical resources to do this and PA opted for simple web pages as the home for the metadata (either static or dynamic pages). As the harvesting component of the software product we use does not detect whether the data being indexed has changed or not, we have to do a full re-gather and re-index each time the metadata repository is to be refreshed. Of the current participants all but one are using static pages (although most are generating these from a database).
Architecture - A centralised search index with distributed images (which is analogous to Internet search engines) is a good simple architecture that works well. A more distributed architecture using Z39.50 would be possible, but is dependent on the larger volume participants having the technology in place to provide their metadata via a Z39.50 target. It is not needed for current volumes but would solve the issue of scalability.

Back to Culture Track Program