[f2f participants] Usecase

Hope you enjoyed the lunch (I also had a dinner).

 

So, I roughly describe now my usecase: automatic audiovisual content
annotation.

 

Several organisations like digital libraries, museums, companies, archives
but also personal users have digital audiovisual content that need to be
annotated in order to be easily accessed. This is, in general, a
time-consuming and costly process, so automatic (or semi-automatic) systems
for this aim are in the main focus of the research in the area of image,
video and audio processing and analysis. Lately, the proposed techniques
focus in knowledge-based analysis and the need for rules is obvious in
several steps like object and event recognition, alignment of different
multimedia ontologies (for example two different domain ontologies that have
some same concepts) etc. Of course in the above cases, we need both a rule
language (there are many) and a rule interchange format between different
systems (if I understand correctly the terminology).

 

The most important thing is that we need efficient tools, since the process
of recognition is really complicated and the data sources contain really
huge data. Thus, it is necessary to start from a quite restrictive core rule
language and then extend it.

 

On the other hand, it is also important that in order for the automatic
analysis tools to be realistic there is need for more sophisticated
characteristics like uncertainty handling, context-based reasoning, temporal
reasoning, etc. Thus, we need to keep this language ready to be extended for
covering these issues.

 

Regards,

Giorgos