XTech 2007

I'll be giving a general talk about Annodex at XTech 2007 in Paris. The conference theme is "The Ubiquitous Web", and covers emerging standards for web technology.

My presentation slides and a short paper are available: Building video webs with Annodex.

Distributed video search

When you have a collection of data that is not distributed, that is all in one place, then it is fairly straightforward to query across the dataset. However when it is distributed you cannot possibly have an overview of the entire system. This is the problem of Web based information retrieval.

Multimedia resources compound the problem because they are continuous, and each resource may require special or custom processing in order to extract useful information. It is simply not possible to build a single query system that spans all types of video items; instead, you first need to build a layer which enables correlation of queries amongst different collections.

Annodex is such a layer.

Web Video

Web Video is not about the struggles of corporations to build the most popular store; it is not about streaming shows from the studios to the users in an overhyped mimicry of television.

Web Video is an opportunity to bring the social power and expressiveness of the Web to the medium of Video. By allowing links between freely published videos we build links between people and their ideas; by allowing anyone to host and display their own stories we open the power of visual expression to many thousands of communities and billions of individuals. Annodex is a technology that makes this happen.

FUCK TV!

Conclusion

We have the opportunity to bring about major social change by creating a new way for humans to create, find and use media content. The existing structures of media distribution are embodied in the way we divide those who produce content from those who consume it, and this division is embedded in the design of current hardware and software. We can create a better way.

Free Software and Free codecs provide the technology, and CreativeCommons provides the licensing. Gnome and GStreamer are a great platform for freely creating, editing, annotating, viewing and remixing content.

Annodex is a way to tie all this together so that anyone can publish media, such that it can be linked and indexed globally, and so that people can surf among millions of media sites seamlessly as browsing the web.

SydPUG meeting: change of location

Tomorrow evening's SydPUG meeting, has been moved to the Australian Computer Society on Castlereagh St, Sydney (between Market and Park Streets -- map).

I'll be talking about "Building Dynamic Video Webs", which involves using web CGI scripts and databases to recompose video, and other fun stuff that we've been hacking on at annodex.net.

So, it's an actual dry run. It's a fair guess we'll move to a nearby pub around 8pm :-)

IEEE Multimedia article is up!

Silvia and I wrote an article about video blogging for IEEE Multimedia magazine, which the editors awesomely titled " Video Blogging: Content to the Max". It's just been published in the April-June 2005 issue, and it's available online: [PDF]

MIT Media Lab

André and I did a few Annodex demos and chatted to people at MIT Media Lab today. Ted Selker naturally harshed us for not doing user studies before implementing the system infrastructure; and Henry Lieberman and some of his students quizzed us on CMML tracks and the scope of our metadata. Thanks to Barry Vercoe for arranging these meetings and our seminar.

Afterwards I couldn't help but record a cute squirrel outside the lab.

Boston Gnome Summit 2004

The first day of the Boston Gnome Summit was an excellently disorganized hackfest. Jon Trowbridge showed me how to write a metadata input filter for Beagle, which, to allow full-text search of Annodex media in the Gnome desktop, will involve a nearly trivial amount of C# to import the libannodex C API and pull metadata from disk files. As well as using these filters to index normal files, Beagle has drivers for indexing web browsing histories.