Geencommentaar.nl

So I kinda stopped posting here a long time ago, but I am contributing to a Dutch group blog: geencommentaar, so everybody head over there!

Most Dangerous idea

Every year, The Edge asks a number of top scientists and other thinkers a single question. This year the question was: "WHAT IS YOUR DANGEROUS IDEA?"

Some very cool people contributed (Rodney Brooks, Daniel Dennett, Freeman Dyson, Richard Dawkins, V.S. Ramachandran, Robert Shapiro, to name a few). One of my favorites is the Daniel Dennetts most dangerous idea: There aren't enough minds to house the population explosion of memes. I love memes...

Sint Game

For our first anual Sinterklaas game is coming up and these are the rules:
Each one of us bought 2 or more pakjes with a combined value of 5 Euros. These pakjes are put on a big pile. Then everyone gathers around in a circle and on his or her turn throws two regular dice:

2 - Give one pakje away
3 - Put one pakje back on the pile
4 - Choose someone to take a pakje from the pile or from someone else
5 - Everyone gives their pakjes to their left neighbour
6 - Choose someone with a pakje to unwrap a pakje
7 - Nothing happens
8 - Take one pakje from the pile
9 - Unwrap one pakje of your own
10 - Swap a pakje with someone or the pile (if you have nothing, just take one)
11 - Take two pakjes from the pile or someone
12 - Take all pakjes from one person

Continue throwing the dice untill all pakjes have been distributed. After that, three more rounds are played. After that everyone keeps their pakjes and unwraps any unopened ones.

Blog du Olivier + Bullettime

A journalist friend of mine started a blog, so I thought I would help him skyrocket his google ranking by adding a link from my little blog. Here you go:

Olivier van Beemen's blog


Ok, while I'm typing away, I made a Matrix-esque 'bullet-time' video at a friends party. I guess it is pretty a in-crowd thing, but why hide my art from the world:

Biktie Bullettime (.wmv, not to big)

New Grouplog

Mattijs (the one from the tijsepijs log) came up with a great idea today. We should create a group blog for anyone at the HCS Lab.
Creating Blogs being so easy, we did just that and now we have the group blog: Human-Computer Blog (for current lack of a better name). In the following weeks we will have to see if this will be as big a succes as the Biktorrr blog :)

SSSW 2005

By the way, I attended the Semantic Web Summerschool 2005 in Cercedilla, Spain. IT was really cool and I met a lot of nice people working on similar topics. And it is really nice to hear the SW stuff from the 'Awesome Superstars of the Semantic Web' themselves.

Most of the students are now member of the sssw05 yahoo group, and a lot of people put their pictures online. I stole a couple of them and put them up on my own webspace.

One of the cool things about the Summerschool was that we were to do a lot of practical stuff. This was either in the form of the hands-on sessions or a mini-project. I was in a mini project with three other students (Rinke, Tom and Jan). Tom came up with a cool idea to envision a semantic framework for lonely hearts ads. This actually landed us the third prize! (An alarm clock)

Google API

Ok, so it has been over three months since my last post, but hey, who said I would post regularly?

I actually am making this new post for two reasons. First a Greek Ph.D. Student read my blog on Ontology Learning and sent me a link to his homepage. It seems he is working on this kind of stuff. That encouraged me to make a new blog post. The second reason was that I did some experiments with the Google API: content!

So I read this paper by Rudy Cilibrasi and Paul Vitanyi on the Normalized Google Distance (NGD), which deals with using Google to extract meaning from the web by exploiting the redundancy of knowledge. It is a really nice paper, very intuitive and firmly grounded in complexity theory as well. In short: the NGD between two concepts is determined by taking the number of hits each concept (term) has on a Google query (NrHits1 and NrHits2) and comparing this to the number of hits the Google query composed of the boolean combination of the two concepts (NrHit1+2). This is normalized by the total number of Google-indexed pages (M). The complete formula is:

(MAX(LOG(NrHits1), LOG(NrHits2)) - LOG(NrHits1+2))/(LOG(M) - MIN(LOG(NrHits1),LOG(NrHits2)))

Anyway, I decided to play around a little bit with this Google distance in my chosen domain: Artists and Art styles. I ran a couple of test. One consisted of calculating the NGD between an Art Style concept ('Impressionism') and Artist Names ('Vincent van Gogh','Manet',...). I found these results (Table shows NGD and name of Artist):

0.171111166168	Monet, Claude
0.41242976086	Hassam, Childe
0.42292288857	Frieseke, Frederick Carl
0.425858890051	Gogh, Vincent van
0.438108128715	Pissarro, Camille
0.456466856759	Morisot, Berthe
0.479547692307	Caravaggio, Michelangelo Merisi da
0.488549764746	Nolde, Emil
0.488922434217	Manet, Edouard
0.496153921128	Rembrandt Harmensz. van Rijn
0.497940393553	Degas, Edgar
0.505376793035	Warhol, Andy
0.519882855212	Goya y Lucientes, Francisco Jose de
0.525532648974	Picasso, Pablo
0.55359075495	Munch, Edvard
0.565804822367	Dali, Salvador

Especially the big difference between number 1 and 2 is pretty weird. Anyway I decided to check on the actual number of hits the google API returned to me, this is what I found (Table shows Name of Artist; Number of hits accoriding to Google API; The number of hits according to the manual search on the Google web page and how these two are related):

	API	Google Web Page	API/Web Page
Monet, Claude	67200	368000	5.476190476
Gogh, Vincent van	7520	40200	5.345744681
Manet, Edouard	34500	184000	5.333333333
Warhol, Andy	44500	237000	5.325842697
Goya y Lucientes, Francisco Jose de	318	681	2.141509434
Degas, Edgar	44400	237000	5.337837838

As you can see, the Google API is normally off by a factor 5.3, but sometimes (in Goya's case) by a completely different factor. Either the Google API is wrong or the normal web based search estimate of the number of hits is off.

The google API news group actually reported the same problem with the number of hits Google API returns. Apparently Google is aware of this problem but doesn't really fix it. And since automatic invocation of Google web pages through http modules is not in compliance with the Googles Terms of Use I wonder if there is a legal way to obtain good values to use in calculating NGD values. (I hope Cilibrasi and Vitanyi didn't use the values the API gave them in their experiments)

Films I have seen and you should too

(A highly subjective, yet alphabetical, list)

2001, A space oddyssey Aguirre, der Zorn Gottes Akira Alien Aliens Apocalypse Now Barbarella Ben-Hur Blade Runner Blow Out Blow Up Blue Velvet Brazil Breaking the Waves Casablanca Casablanca Citizen Kane Clerks Close Encounters of the Third Kind Das Boot Delicatessen Dogville Donnie Darko Dr. Strangelove Epidemic E.T. Eternal Sunshine of the Spotless mind Europa ExistenZ Fargo Faster Pussycat Kill Kill Festen Fitzcarraldo Full Metal Jacket Ghost in the Shell 1 + 2 Himmel Uber Berlin Iron Monkey Jaws Jules et Jim Lost in Translation M Magnolia Manhattan Metropolis Mony Python and the Holy Grail Mulholland Drive Nosferatu On the Waterfront One Flew Over the Cuckoos Nest Paris, Texas Pi Raging Bull Raiders of the Lost Ark Ran Rashomon Rear Window Riget I+II Salo: 120 days of Sodom Seven Samurai Seventh Seal Spirited Away Star Wars Stranger than paradise Taxi Driver The Big Lebowski The Day the Earth stood still The Deer Hunter The Godfather 1 + 2 the Good, the Bad and the ugly The Graduate The Shining The Shining The Silence of the Lambs The Wizard of Oz The Wizard of Oz Tonari No Totoro Twelve Monkeys