TI A Multimedia Information System with Automatic Content
Retrieval 

AU Michael Hugh O'Docherty 

RT MSc thesis 

LT UMCS-93-2-2 

OR UMCS 

AV ftp ftp.cs.man.ac.uk:pubTRUMCS-93-2-2.ps.Z 

AV http
http:www.cs.man.ac.ukcsonlycstechrepAbstractsUMCS-93-2-2.html 

AV email techreports@cs.man.ac.uk 

MN February 

YR 1993 

AB Traditional databases use simple data types - chiefly numbers
and strings - to represent information such as payroll records or
scientific data. Queries are commonly expressed in relational
calculus or algebra and the results require little more than a
textual terminal for presentation.  The natural successor to the
traditional database is the Multimedia Information System which
also uses richer data types - images,text, sound and so on - to
describe some application domain.  This calls for more advanced
technologies such as graphical workstations for data creation and
presentation, high-bandwidth networks for fast access to
distributed data and large repositories for the storage of
objects.  These technologies are now widely available.  Ideally,
queries put to a Multimedia Information System should refer to
the content of stored objects - content retrieval - and results
should contain complexas well as simple data types.  Manual entry
of content descriptions is time-consuming, error-prone and
subjective; better approach is to draw on the emerging fields of
image interpretation, text interpretation, speech interpretation
and knowledge representation to provide automatic content
retrieval.  This thesis serves as an introduction to Multimedia
Information Systems and related fields and describes in detail a
pilot system with automatic content retrieval built by the
Multimedia Group at Manchester University; the author has been
instrumental in the design and implementation of this system,
especially those components that cater for the gereration and
querying of content descriptions.