Week 6: Annotating multimedia

We read: Annotation of Multimedia Using OntoMedia, by K. Faith Lawrence et al.

To show dedication and commitment, we didn't cancel the meeting because Michael is out of the country.  Oh no!  Instead, he G+ Hangout-ed* in from Romania.

(Views of Bucharest)

The paper was written in 2006, and contains a few outdated concepts, and things that aren't around any more, but all in all the ontology it presents seems solid.  I (Amy) like the idea of annotating the contents of media, on top of just the meta-data-y stuff.  OntoMedia might find a use in my work.  Particularly for improving search options - if you can vaguely remember a video or comic based on something that happened, having linked data describing characters and events can be really valuable.

Main problem is it's all annotated by hand, and it'd be quite a job persuading people to break down and fill in details about every incident in a narrative.  Their Mediate software is a good solution for people who want to annotate but don't want to hand-craft RDF, but even with a nice form interface, those people are in a tiny minority.  It would be easy enough to automate annotation of scripts for the most part, but videos and comics would be a different story.. Paolo suggests facial recognition of actors, plus knowledge of the characters they play, but we all struggled to think of an automated solution for (examples in the paper) Event LOSE GUN or character state Transition ALIVE to DEAD.

I'd be interested to see an example of OntoMedia used to annotate comics (even short webcomics).  The paper mentions that it can be used for all kinds of multimedia, including images, but only elaborates on one scene of Total Recall, and associated screenplay and novel excerpts.


Years of complicated research into facial recognition and tracking, and this is what it yields?!

* Not quite the same ring to it as 'Skyped', I'll admit.

Read Amy's more detailed but less coherant notes from reading this paper.

No comments

Post a Comment

© Ontologies reading group for first year PhDs in CISA
Design:Maira Gall.