Yesterday I attended a research colloquium at university from an external speaker. Lynda Hardman of CWI talked about “Presenting Knowledge on the Semantic Web”.

She covered three interesting topics dealing with media and metadata which were quite interesting. We —the audience— didn’t make it easy for her, but I do feel that a lot of interaction took place and stuff was learned.
Vox Populi
The first project she showed us was a project called Vox Populi from Stefano Bocconi, which is an automated video composer which uses rhetorical metadata to automatically create documentaries supporting a point of view.
A possible schematic of reasoning:

In the example they interviewed people in New York after 9/11 (Interview with America) and asked them their opinions about the situation and about the looming war with Afghanistan. After cutting up the interviews and annotating them on semantic/rhetoric value they could automatically script edits either supporting war, opposing it, or providing a balanced view (full details in the paper).
This is a technically very interesting concept and the montage shown was quite convincing. Don’t take my word for it, you can try it out yourself.
Objections Interesting as it may be from an academic point of view, I have a lot of ethical concerns when faced with this project. If it sees deployment a tool such as this will only accelerate the current media culture of immediacy, imbalance and balkanization we are seeing right now.
Furthermore the opinions of the participants are grossly distorted. Their stories are cut to a mixable resolution discarding any overarching message and those cuts are used to support or detract from arbitrary points of view. I honestly don’t know who would agree with this treatment.
Vox Populi fits in well with the soundbyte-ization of culture while I think more insight may be gained from longer fragments, more balanced views and more thinking not less. In the process as it is shown the original video is reduced to a set of parameters to be consumed by a mathematical optimizer. That means most of the context, coherence, nuance and insight is lost.
What you then get out of the system is an aggregate of the stripped fragments which has a new set of values instilled upon it (like any documentary would) but which has only been directed by a set of input parameters working on the few dimensions of information distilled from the cuts.
Dystopia Talking with Oliver about this project we took it a bit further and discussed what would happen if the rhetorical annotation of the video could be automated and unleashed on the full dutch video archives.
Dutch Institute and Archive for Audiovisual Media:

Not only does Vox Populi pick content based on rhetorical value but it also tries to pick fragments which a specific viewer will feel affect and authority for. This means that if you have enough well-annotated content you could make a personalized messages for each viewer with content optimalized for their experience and persuasion.
The media landscape will be completely fragmented and people will be shown video for manipulative purposes by those in power —in power over the media.
P.S. Be sure to read that Wikipedia reference which says that the Vox Populi […] semper insaniae proxima sit (is always close to insanity).
E-culture
We got a quick demo of a collections browser for various collections of cultural objects. It displays metadata information about objects and relations between them and makes it possible to search for objects in the catalogue.
The comparison with Amazon was made which may be silly at first glance but solves how you can tackle a lot of the similar problems these collections are faced with succesfully. I am not convinced that each institution in the world has to come up with their own clever user experience silos.
Wouldn’t it be best for all involved if they would create nicely interlinked and Search Engine Optimized pages for each object in their collection and let the big search engines index that? Or am I being too naive?
NewsML
Lastly a standard for the exchange of newsworthy information: NewsML was treated very briefly.
I don’t know too much about this technology but the above description is awfully reminiscent of the Atom Syndication Format which is already widely understood and used to exchange news.
It still would be nice to have an exchange of ideas to foster goodwill and prevent stuff from getting duplicated. There might be advantages to using Atom/REST as the carrier medium of NewsML data. In any case the internet as a whole benefits strongly form more collaboration and less institutionalized strife.