Submission
Government Access Architecture Workshop
23-24 September 1996, National Library of Australia, Canberra
A group of invited 'experts' from Commonwealth and State bodies plus representatives from research and private bodies were asked to discuss the concepts in the background paper that had been distributed to the group, and to make recommendations on future actions. The background paper and attendee list are at Attachment A
Workshop Objectives
The workshop agenda was to seek a collective strategy towards a technical architecture that would allow discovery and retrieval of Australian Government information from multiple disparate sources in a coherent and focussed manner. The agenda was to include both the browse paradigm and the search paradigm.
The Access Architecture in this instance was defined as comprising of indexing and searching as two related but separate activities. The key objectives of the workshop were:
- to find mechanisms that would allow the development of user interfaces which provide unified and coherent views of govt information; and
- to design an architecture that would scale with the increasing volume and complexity of Government information, and would allow interoperability among different agencies' systems without compromising the autonomy of agency-level information retrieval mechanisms.
The position paper and the detailed agenda are at: http://www.adfa.oz.au/DOD/imsc/imsctg/workshop/bckgrnd.htm.
Workshop Processes
Presentations were given outlining work done to date by the IMSC-TG (Information Management Steering Committee - Technical Group) as presented in the IMSC-TG report - http://www.adfa.oz.au/DOD/imsc/imsctg/imsctg1.rtf).
Invited speakers gave outlines of access architectures and search capacities employed by Educational Network of Australia (EdNA) and Australian Legal Information Institute (AustLII). A representative from Digital gave an overview of the Alta Vista product. Distributed Systems Technology Centre (DSTC) gave a presentation on their Open Information Locator (OIL) prototypes and on an information focussing prototype HyperIndex Browser (HIB), and the National Library of Australia (NLA) demonstrated their Metasearcher pilot over a number of distributed and internal sources using public domain as well as legacy search engines.
These presentations set the scene for workshop breakout sessions at which participants addressed a number of issues arising out of the background paper.
The specific questions addressed by the workgroups were:
Question 1:
Metasearcher
Is it a valid option as a whole-of-Government
information access mechanism, either in the short term and/or in the longer
term ?
- If so, how can we improve the currently available Metasearchers
?
- Is there a need for a 'lead-agency' mechanism to take things further in
this area ?
Question 2: Metadata
Is
there a willingness among the group to adopt and implement the metadata models
proposed by recent initiatives such as the IMSC-TG report, July 1996 ?
If so, which 'views' or categories of metadata are regarded of highest priority ?
Question 3: Storage of Metadata
Where should metadata reside ?
- embedded within data ?
- outside of data, but within the
agency, in a separate database ?
- in a separate database, but outside the
agency ?
Question 4: Indexing Mechanism
- who creates the primary indices? the owning agency or a central
agency?
- where do the primary indices reside ?
Question 5: Is full text indexing necessary?
Question 6: Persistent Locators of resources
Is such a mechanism necessary, and if so how to design and manage it ?
What are the benefits, if any, of improved access to Goverment information ?
Workshop Outcomes - A Summary
The longer term recommendations of all three workgroups converged very closely, whereas there was more divergence on immediate actions. The majority emphasis was skewed towards a longer term architectural approach and a minimisation of short term solutions.
The principal features of the consensus can be summarised as follows:
- that the user layer and the data provider layer be loosely coupled, with a middle layer mitigating the disparities of data sources and user requirements;
- that individual agencies be encouraged to create their own resource descriptions (metadata) and indices, and to store these locally;
- that the above agency-level metadata form a Government-wide repository of metadata, and that this repository be utilised in guiding and directing search requests, by the middle layer mentioned in 1 above;
- that centralised or Government-wide indexing be minimised, to focussed second-level indices, as opposed to detailed, full-text indices;
- regarding management and resourcing of the above, a lead agency model was favoured, with small-group pilots involving appropriate agencies, but the details of these mechanisms were left for discussion in other appropriate forums, such as IMSC or equivalent..
The principal divergence of views was over the issue of deploying metasearcher(s) as a short term measure, pending the establishment of a more structured middle-layer based design. While it was acknowledged by the Workshop that a metasearcher approach offers the advantage of a relatively quick solution and allows a clear separation between the query interface layer and the rest of the search mechanism, thus affording maximal autonomy to the individual information providers in their search engine strategy, one of its main problems is lack of scalability as the number of target databases increased.
Also, while this (metasearcher) approach demands less of a coordinated search infrastructure, it does call for agreed standards on description and interchange of technical metadata about the participating search engines, so that the query interface can synthesise a common interpretation of the total retrieval. Such technical metadata would also need to include collection descriptions so that the query interface can choose the most appropriate data sources for a given query.
Some of these technical metadata issues are currently being addressed by the STARTS (Stanford Protocol for Internet Retrieval and Search) initiative through the formulation of a lightweight Internet information retrieval protocol. The issues addressed include query syntax, results syntax, and other aspects of distributed information retrieval such as ranking algorithms used by individual search engines, with an aim to facilitating a coherent and integrated retrieval set through interchange of such technical metadata among participating data sites and search engines. Further discussions on these issues may be found at: http://www-db.stanford.edu/~gravano/starts.html.
A more detailed account of the Workshop's deliberations follows.
