Powerset October 30, 2007
Posted by Andre Vellino in Search, Semantics.2 comments
I’m looking forward to hearing from Powerset Labs so I can try their new “semantic” search engine from Xerox PARC. I registered, but there appears to be a backlog due to the buzz in the blogosphere. I find it interesting to see what kinds of positions they are advertising for: this the first time I’ve seen an post for a “relevance ranking engineer”.
2,3 Turing Machine October 25, 2007
Posted by Andre Vellino in Logic.1 comment so far
I wonder if the recent proof by Alex Smith that the smallest universal turing machine has 2 states and 3 colours is related to the fact that 3-SAT is NP-Complete. Propositional variables have 2 states (0,1) and 3-SAT is the boolean satisfiability problem for conjunctions of ternary disjunctions of propositional variables. This can’t be a coincidence!
RecSys2007 October 22, 2007
Posted by Andre Vellino in Collaborative filtering, Data Mining, Information retrieval, Recommender service.2 comments
I like small topic-specific conferences like RecSys2007! You get to know a few people, for one thing. But also, there isn’t anyone there, typically, who doesn’t have some very good reason for being there, so you’re almost guaranteed to bump into someone interesting.
I didn’t realize that RecSys2007 is, in fact, the first fully-fledged conference for recommender systems. This subset of the Information Retrieval community has been lingering in the world of workshops in larger conferences and occasional summer-schools, but with an attendance of ~120 participants with from dozens of countries, I’d say the field has come of age.
User Interfaces for a DL October 12, 2007
Posted by Andre Vellino in Collaborative filtering, Digital library, Enterprise Architecture, Visualization.add a comment
My colleague Glen Newton has developed some innovative search-and-browse interfaces for a digital library – they are part of his Ungava project. This web application has been deployed on our experimental platform CISTI Lab and you can try it out on a collection from the NRC Research Press.
Stephen Anthony has blogged at greater length about CISTI Lab and all the other stuff that’s on there, especially from the point of view of Enterprise Architecture.
One of the interfaces Glen is experimenting with is what he calls “drill clouds” and he has recently written about this on his blog. Although I’m not the world’s biggest fan of tag cloud-like things (I just don’t like text as a UI), I think this is a pretty cool experiment and I believe in having multiple interfaces (drill clouds, timelines, exhibit view) to achieve the same end.
I plan to use Ungava as a platform for generating user-data and article recommendations to users and the question arises of what interfaces will work best for displaying recommendation results. Sequential lists are understood by one and all, but they’re boring and not especially useful.
Glen recently pointed me to an interesting clustering UI for displaying recommendation results, but I don’t think I will try to emulate it. It looks nice, but I don’t find it informative.
I wish I could be more precise about how I think a recommendation result interface should look like. I’m drawn to something more like intention-driven search interfaces explored at Yahoo and also implemented at Microsoft Live Search Academic (Beta). These engines give you sliders and controllers that allow you to re-order search results according to relevance criteria. Something dynamic and interactive like that can be profitably done with recommendations as well.
Web Intelligence ‘07 October 11, 2007
Posted by Andre Vellino in Collaborative filtering, Recommender service.add a comment
The preliminary program for the International Conference on Web Intelligence and Intelligent Agents 2007 is finally out, barely 4 weeks before it begins. It must be a significant scheduling problem to organize such a conference, and I don’t envy the planners’ task: I counted no less than 220 people involved in its organization and / or peer-reviewing the papers!
I hope that our modest workshop paper will get some useful feedback. I think we’re on the right track, but there’s nothing like explaining your ideas to peers (such as, I hope, the GroupLens researchers who developed Techlens) to see where the gaps are.
