CAMA 2007

June 25th, 2007 by Martin

I’m currently sitting in the lobby of the Bosman’s Hotel, trying not to fall asleep before my taxi arrives to take me to the airport…

The CAMA workshop yesterday was very interesting, there’s a lot going on in the field. For me, the main questions when trying to use CAM to provide some advanced services are still the following ones:

  • What is a good model for context? Do we maybe have to follow a multi-layer approach which can be used for different kinds of services? Defining a complex model would of course potentially allow for very advanced services, but how many objects will ever be annotated with such information? That’s why I’m a friend of a quite flat and simple model that eases a common understanding.
  • How can we motivate users to share CAM? This not only involves issues like good interfaces, simplicity and joy-of-use, but also privacy and security issues.

Concerning the second point, I expected to get some suggestions from the keynote of Seth Goldstein. AttentionTrust is nice to see, but in my opinion, the idea that users “own” their data and footprints they leave on the web is almost impossible to realize. If oyu have any idea of how this could be done, please leave it in the comments, because I want to get rich ;-)

20070623 vancouver 002

Later, Wayne Hodgins talked about mass personalization. I like his talks a lot (most of the stuff I already knew from his blog), still I disagree with some parts:

  • Mass personalization for me is just a second step in the development. First of all we should have mass enabling, we should allow users to autonomously choose the content they like, and the content should be accessible! Sometime in the future we’ll maybe have some real good recommendation services, but we should first give the control to the people, allow them to create mashups and to remix the content. Adaptability first, then adaptivity!
  • I still don’t get what constitutes a learning object. For me, any resource (this includes real objects that can be referenced via URI, e.g., a building) can be a learning object. That’s why I prefer the term learning resource. In addition to that, I think that the comparison between a learning object and objects in software engineering is not that helpful: It suggests that a learning resource has a well defined input- and ouput behaviour as some software part has - but reality is different in my opinion.
  • I also think that the “lego” metaphor doesn’t fit: Lego parts follow a quite strict definition (concerning length, width, the points of connection, etc), whereas some piece of content has potentially infinite many aspects that have an influence on whether it can be connected to another object or not. I think comparing a learning resource with a piece of lego is dangerous, because people might get the impression that they just can put arbitrary objects together in any prefered way. In reality, this doesn’t work at all.

It was a very inspiring event, and I hope we can continue to work together in this field!!!

Measuring the quality of conferences

June 22nd, 2007 by Martin

Today I had the pleasure to attend a very interesting talk given by Ergin Elmacioglu from the PIKE Research Group. He introduced the ConfRank system that judges the goodness of academic conferences by analyzing the programm commitee. Therefore, the social network of ACM authors is analyzed, and the results were really impressive. Maybe I’ll check our LOKMOL PC sometimes ;-) I can imagine that it would also be very useful to check scientific blogs to get some evidence about the quality of conferences. And what about a blacklist? I really avoid conferences when I see certain people in the PC!

He also provided some very interesting links which I thought might be quite useful for me - unfortunately, http://fakeconferences.org/ does no longer exist (they seem to offer a “positive” alternative, but I didn’t check this), and http://cs-conference-ranking.org/ is currently not available. Luckily, the Inauthentic Paper Detector still seems to work. Time to check - do you write like a robot or not? ;-)

JCDL 2007 in Vancouver

June 21st, 2007 by Martin

I’m currently in Vancouver attending the Joint Conference on Digital Libraries where I’ll give a talk in the CAMA Workshop on Saturday about “Sharing Contextualized Attention Metadata to Support Personalized Information Retrieval”. As there is no wireless at the conference (pretty strange for a conference in 2007), I can just blog from my hotel room (free WLAN there, hahaha).

The conference was opened today with a keynote by Google’s Daniel M. Russell about “What are they thinking? Searching for the mind of the searcher”. It was pretty interesting to see how Google copes with the different contexts, perspectives, and mental models of users. I guess the most difficult problem is related to what Wayne Hodgins is always refering to as “unlearning”: People don’t follow the optimal path, but the most familiar one. Tough work for Google ;-)

DSC01067

Another very inspiring talk was given by Rahul Nair from Yahoo! Research Berkeley lab (looking at the photo on his page almost prevented me from posting this, but it was really good anyway). He introduced the World Explorer - a great example of how we can make use of the mass of user-generated metadata: They use geotagged images from flickr to provide meaningful tags about any region in the world. A simple but very exciting approach - take a look!