Digital Media Libraries Program
The rapid uptake of the internet, digital television and multimedia
services is forcing organisations in the news, information and entertainment
industries to restructure their assets for interactive access for
in-house and external use.
The Digital Media Libraries (DML) Program was established to add
value to this restructuring. It investigated, developed and demonstrated
innovative technologies that are particularly relevant to digital
media libraries and on-line information services. To assist with
these tasks, the DML Program established the following projects:
These projects addressed key operating and strategic issues such
as data volume requirements, indexing and cataloguing facilities,
networking access to digital media collections, information retrieval
and delivery mechanisms, systems architecture, content models, and
tools for time-based media and semantic interpretation of the content.
Program Highlights
Developed P@NOPTIC, a set of web search and intranet management
tools providing high quality answers, high speed and high capacity.
P@NOPTIC uses the most effective scoring formula known (Okapi BM25),
and has been tested in TREC (the international Text Retrieval Competition).
Indexing is very fast, up to 13 gigabytes (2 million pages) per
hour on a single Pentium processor. It features a parallel net-friendly
spider combined with PADRE text retrieval software and a CGI web
search interface and document summariser. P@NOPTIC indexing and
query processing takes advantage of multiple networked machines
to improve speed and data handling capabilities.
Developed the Very Large Collection2 (VLC2) test collection, used
at TREC. It comprises 18.5 million web pages and a total of 100GB
of data. VLC2 has been distributed to more than 15 universities
and commercial groups worldwide.
Developed PADRE (Parallel Document Retrieval Engine) text retrieval
software. Initially developed for the Fujitsu AP1000, PADRE investigates
finding and ranking relevant documents within a large collection
of electronic text. It is an application of parallel and distributed
computing to information retrieval. PADRE was tested on VLC2, indexing
the data in 7.45 hours using only a single Pentium machine. Lanterna
Magica has included PADRE in a proposal for a nationwide search
and navigation system. PADRE has also been incorporated in P@NOPTIC.
Developed PASTIME (Parliamentary Sound Text and Image Environment)
software, which offers browsing tools suited to multimedia files
and the efficient management of large hyperbases. Tested at Parliament,
collaborating with the Department of the Parliamentary Reporting
Staff, it captured audio, imagery (still and motion) and text proceedings
in the House of Representatives. PASTIME is offered commercially
through Fujitsu.
Undertook a study into providing an assisted captioning system in
cooperation with the Australian Caption Centre (ACC). This led to
the development of the CACTUS demonstrator, which applies automatic
speech recognition techniques and audio-video processing in the
problem domain of automated captioning.
Experimental running of the Film Researchers Archival Navigation
Kit (FRANK) on Telstra’s broadband network, demonstrating its
media-asset management capabilities for audio-visual material. FRANK
was also tested on the ABC’s television archives.
Key participant in the TransACT Trial, providing on-line access
to high quality digital video content in cooperation with ScreenSound
Australia.
Conducted a high profile consultancy for SGI on the impact of multimedia
with the Australian Parliament.
Presented a study on the impact of broadband networks to the ACT
Government Department of Urban Services.
|
