Staff Papers

Picture this: a working combination of Z39.50, the Dublin Core and MARC21

A paper given by Debbie Campbell, Manager, Infrastructure Projects, National Library of Australia, at ALIA 2000 "Capitalising on knowledge the information profession in the 21st century" 24-26 October 2000 in Canberra



Abstract

PictureAustralia is an example of a service which capitalises on knowledge creation. The rich content attached to image collections which has been compiled over many decades is being re-purposed in a newer technical environment to increase public access to Australia's heritage in pictures.

Because of the breadth of possibilities provided by the technical environment which supports the World Wide Web, it has been necessary to invoke several standards to smooth out the anomalies when bringing together access to what were, until now, physically separate collections. The software behind PictureAustralia, Blue Angel Technologies' MetaStar Enterprise, invokes standards both old and new to the bibliographic world. A combination of MARC21, the Dublin Core and Z39.50 have been exercised for PictureAustralia, to support access to both text and images.

Keywords: MARC21, Z39.50, Dublin Core, resource discovery, PictureAustralia


Introduction

PictureAustralia is an example of a service which capitalises on knowledge [PICTUREAUSTRALIA]. The rich content attached to image collections which has been compiled over many decades is being re-used in a "new" technical environment to increase public awareness of Australia's heritage in pictures.

Because of the breadth of this relatively recent technical environment, the World Wide Web, it has been necessary to invoke several standards to smooth out the anomalies when facilitating joint access to what were, until now, physically separate collections.

The software behind PictureAustralia, Blue Angel Technologies' MetaStar Enterprise, also provides native support for standards which library services use to provide access to bibliographic content - MARC21, Dublin Core, and Z39.50 - but in PictureAustralia these standards have been used to support access to both text and images.

Some standards aren't discussed here, such as those required for the digitisation process, or the choice of image formats. (In the case of this service, it is the Joint Photographic Experts Group's standard JPEG). The software is able to manage variations in digitised image sizes and quality, but other environmental factors such as network bandwidth must be taken into account when providing an image service.


PictureAustralia's architecture

So how are Z39.50, Dublin Core, and MARC21 deployed ?

Many of the participants in Phase I of the PictureAustralia service, known as the ImageSearch Project, were libraries. Other State and public libraries are also joining the service. Those who already have large image collections, often made available online via a Web entry point separate from their public access catalogues, have rich descriptive content stored in MARC21 format or a variation. In many cases, these descriptive records are already repurposed between the OPACs and single visual format gateways.

However, PictureAustralia has also tested the mapping and conversion of the MARC21 records to the Dublin Core schema, in order to capture the record descriptions into a central repository.

The ImageSearch Project participants agreed to generate metadata according to the Dublin Core standard schema by mapping from their original metadata to provide descriptive content to PictureAustralia. It is, in fact, the only standard imposed on participation in the service. The use of the Dublin Core elements for mapping legacy data in a reasonably consistent way has been extended to the data of all participants, including the Australian War Memorial's military schema and the National Archives' government-sourced metadata.

Does this add an additional layer of complexity for searchers? When mapping from a granular schema such as MARC to a coarse schema, the information is not lost. There is a choice of not duplicating the rich schema in its entirety. As long the connection between the two is made explicit, via appropriate links, then the relationship should be clear even if some complexity remains. For example, one of the more detailed sets of metadata comes from the National Library of Australia:

Title Dan Kelly, Steve Hart, Ned Kelly, Joe Byrne, Kate Kelly.

Creator Marony, Patrick William.

Subject Kelly, Dan, 1861-1880 -- Portraits, etc.

Subject Hart, Steve, 1859-1880 -- Portraits, etc.

Subject Kelly, Ned, 1855-1880 -- Portraits.

Subject Byrne, Joe, 1856-1880 -- Portraits, etc.

Subject Kelly, Kate, 1862-ca. 1899 -- Portraits, etc.

Subject Bushrangers -- Victoria -- Portraits.

Subject Picture -- Portrait -- Oil.

Date [ca. 1894].

Date 1997/05/26

Type Image

Identifier 12819

Identifier 2263689

Identifier PIC R4407. Con 2121.

Identifier PCD96 3794-1024-1729 : IMG0019.PCD

Format 1 painting : oil on canvas ; 107 x 61.2 cm.

Coverage Australia -- Victoria

Fig: 1. Full Dublin Core metadata record

Image Record No.:12819

Creator:Marony, Patrick William.

Title:Dan Kelly, Steve Hart, Ned Kelly, Joe Byrne, Kate Kelly.

Publisher/Date: [ca. 1894].

Image Format:1 painting : oil on canvas ; 107 x 61.2 cm.

Coverage:Australia -- Victoria

Subject:

Kelly, Dan, 1861-1880 -- Portraits, etc.

Hart, Steve, 1859-1880 -- Portraits, etc.

Kelly, Ned, 1855-1880 -- Portraits.

Byrne, Joe, 1856-1880 -- Portraits, etc.

Kelly, Kate, 1862-ca. 1899 -- Portraits, etc.

Bushrangers -- Victoria -- Portraits.

Picture -- Portrait -- Oil.

Immutable:2263689

Identifier:abn83061538

Last Update:1997/05/26

Call No.:PIC R4407. Con 2121.

Master:PCD96 3794-1024-1729 : IMG0019.PCD

Fig. 2 Original National Library of Australia MARC-based record.

PictureAustralia is an example of a cross-domain information retrieval application, as recently expressed by William Moen in his discussion of resource discovery using Z39.50 [MOEN]. The Dublin Core Metadata elements are represented as Use attributes in Z39.50 for searching. The heterogeneity of application of the elements by the participating agencies has some derived 'semantic equivalence' through the Dublin Core mapping. Contrary to many Z93.50 implementations, all of the metadata is indexed equally, and made accessible in both composite indexes and on an individual element basis. The indexes themselves and their relationships as expressed in the boolean operators - and, or, not - are driven by Z39.50 syntax. Phrase, keyword, and truncation capability are also dictated as a Z39.50 implementation in MetaStar Enterprise.

Other Z39.50 functions, such as sort and merge, are under development. However, as Z39.50 is supported by MetaStar Enterprise in native mode, it is possible to search distributed Web collections without harvesting the metadata into the central repository. It would require each participant in the service to support Z39.50 on their own servers, and this is where the plan for a fully distributed model comes unstuck. PictureAustralia is a collaborative and cross-sectoral service. It is essentially only bibliographic systems which support Z39.50. Systems for museums and archives rarely do. However, a distributed Z39.50 capability may lead to other problems, such as an inability in some systems to integrate the results from different locations. For now, PictureAustralia will continue to harvest the Dublin Core metadata on a routine basis. The experience of using a centralised Z39.50 metadata index in this service has shown that it is comparatively easier to achieve, and more scalable than distributed Z39.50 indexes.


Search functionality

In summary, the descriptive information of the combined metadata collection is able to be searched by using a single interface, because it has been harvested and copied into a central database at the National Library. The images themselves remain on the Web servers of each participating agency, to be managed by each institution. The access point from the database to the pictorial image resident on each institution's server is the Universal Resource Locator (URL), one of the few stable protocols in the Web environment. The URL is stored in the identifier element of the Dublin Core metadata schema. Some of the linkages may vary slightly, depending on environmental factors at an individual site, but the aim is to deploy a strategy of technical convenience.

In PictureAustralia, keyword searching is available in a simple mode, where the creator-title-subject elements are combined. In this mode, 'and' is the default for a phrase. For example, entering the keywords wild flowers will find images where their metadata contains both the words wild and flowers. Entering the phrase Arthur Boyd will find works painted by Arthur Boyd as well as works containing Arthur Boyd as the object of attention. This search functionality is, in essence, a form of fuzzy matching.

In advanced search mode, each data element may be mined separately or in combination through the use of the boolean operators. Right-hand truncation is also supported. In the original ImageSearch prototype, the default search was cast across all of the metadata. Important details such as the names of individuals in photographs are stored in the date and description elements, but because the latter can hold a lot of disparate information, there are too many hits in the results set when searching half a million metadata records. The advanced search functionality overcomes this situation.

While all of the metadata is therefore available for easy searching, PictureAustralia has received some requests to provide more interpreted content.


Interpretive context

Each of the image providers to the PictureAustralia service has made available a Web entry point to their online pictorial collections. However, not all of them currently provide a database-style search service. There appears to be a trade-off in the creation of contextual information, where information presented statically is able to contain more detail, and metadata created to support keyword searching which is easily expanded as new images and their metadata records are created. PictureAustralia provides for both.

Trails, featuring highlights and sub-collections, will continue to be developed over time, and may provide a form of contextual information. For example, they may be sourced from virtual exhibitions of particular artists, or from co-located images of many different institutions in a related theme. A lot of interest in this level of functionality has been shown by the higher education sector, and PictureAustralia will be exploring this enhancement in the near future.

Searching habits are being recorded in PictureAustralia. This is considered essential to improve the search functionality of the service, and also allows it to be compared to the access functionality for visual materials in Online Public Access Catalogues (OPACs). An interesting statistic recently provided by the State Library of Victoria [SLV] showed that searching habits for visual materials are quite different from those deployed for other sorts of materials. In the State Library of Victoria's MultiMedia Catalogue, which hosts pictorial materials, the topic keyword is used far more extensively than any other element available to the searcher. The standard OPAC shows a more distributed search pattern across title, author and subject. The title of an artwork may not closely reflect its 'aboutness' and metadata creation may need to reflect this.


Issues arising in the searching of heterogenous metadata

Since the quantity of metadata, almost half a million records, has been provided to PictureAustralia, issues of heterogeneity of metadata have become more apparent. For example, relevance ranking has no relevance - it's not appropriate given the wide mix of semantic usage of the metadata elements. Therefore, other strategies need to be deployed to ensure that each collection of metadata is equally accessible.

Highlighting particular metadata practices may also be an issue, especially where the features of each institution's metadata are different. For this reason, PictureAustralia strongly recommends providing a few of the elements which are optional in the Dublin Core. We are also creating a metadata registry as part of the PictureAustralia Web site, which allows metadata creators to see the diversity of practices at a glance. This will take us beyond a recent approach that we saw described in a French educational metadata schema where metadata is created 'in a very free and intuitive way' [EUN].

The quantity of metadata applied to each image may also result in a qualitative issue, in the sense that applying broader concepts to defined scenarios may result in false matches when the user needs a facilitated match. The issue has best been described by Lloyd Sokvitne in his recent paper to the Communication Research Forum [SLT]. The paper discusses the fact that incomplete, incorrect, or invisible metadata may override the value of qualitative metadata. PictureAustralia needs to do more research in this area.

A related issue arises from the visual nature of this and many other Web services. The easy accessibility of images discourages people from reading instructions or advice, and they certainly don’t want to click to a help or an advanced search screen which may tell them about the metadata quirks. Users are just interested in typing and clicking, sometimes to test out the service. PictureAustralia plans to address this issue in part by providing online training in the form of a visitors' centre concept.

A concern in the future is likely to be the identification of the best use of thesauri in ways to benefit both the metadata creator and the user. At the moment, subject keywords from many different thesauri, including Library of Congress Subject Headings, are accepted into the subject element. However, for those agencies not yet using a thesaurus, PictureAustralia recommends the adoption of the newly released Australian Pictorial Thesaurus [APT].


Summary

Through the implementation of the PictureAustralia service, we have found that the Dublin Core has been a great equaliser. It has been possible to take the rich descriptive information created within specific communities such as archives, museums, galleries, historical societies and libraries and bring it together in a single meaningful structure which enables our target sector (in this case, the average Australian, sometimes known as the ‘taxpayer’) to see the value and beauty of the collections they have funded.

The Dublin Core schema has enabled the specificity and specialist nature of an archivist’s understanding or an art historian’s knowledge of our cultural heritage to be conveyed to a broader audience. The audience doesn’t have specialist knowledge and doesn’t need it to discover what’s inside our ‘memory institutions’. In addition, it has also revealed the breadth of our collections to people overseas who want to learn more about Australian culture and history. Perhaps this was what was envisioned by the Dublin Core community all along. It has certainly been envisioned by the applicants of Z39.50, and possibly even exponents of MARC21. PictureAustralia has shown it can become a reality.


Acknowledgements

Tony Boston,
Director, Digital Services Project, tboston@nla.gov.au

Danielle Freeman,
PictureAustralia Service Manager, dfreeman@nla.gov.au

Simon Jacob,
Senior Programmer, sjacob@nla.gov.au

Cliff Law,
IT Division, claw@nla.gov.au


Bibliographical References

[APT] The Australian Pictorial Thesaurus

Available: http://www.picturethesaurus.gov.au 12 October 2000.

[EUN] Metadata styles

Available: http://www.en.eun.org/etb/metatypes.htm 12 October 2000.

[MOEN] Resource Discovery Using Z39.50: Promise and Reality

Available: http://lcweb.loc.gov/catdir/bibcontrol/moen_paper.htm.

12 October 2000.

[PICTUREAUSTRALIA] Available: http://www.pictureaustralia.org 12 Oct 2000

[SLT] Policy and design issues affecting the development of an Information Architecture for a Government Resource Discovery Service, Communications Research Forum 2000. Available: http://www.service.tas.gov.au/papers/comresearchpaper.asp [Accessed 24 October 2002]

Lloyd Sokvitne, State Library of Tasmania, October 2000.

[SLV] Anne Beaumont, Technology Services Department, State Library of

Victoria, September 2000.


Biography

With an Information Technology background, Debbie Campbell provided research services to the National Library's PANDORA project and the MetaWeb project. In 1998, she became the NLA's Metadata Coordinator with a particular responsibility for assisting national academic and government gateway projects to achieve the vision originally intended for the adoption of standard metadata schemas. Her most recently implemented project is the National Library's new PictureAustralia service.

Ms. Debbie Campbell
Manager, Infrastructure Projects
Coordination Support Branch
National Library of Australia
Canberra, ACT, 2601
Australia
ph: +61 2 6262 1673
fax: +61 2 6273 2545
email:dcampbel@nla.gov.au

Top

About This Site | Copyright | Privacy | Accessibility | Site Map | Site Search | Content A-Z | Contact Us