A slight gripe about Google scholar

While the community gets hooked on Google Scholar (see articles in Nature) it is worth noting its flaws rather than becoming caught up in the general excitement. One of the hyped features of Google Scholar is that it returns the most cited papers first. However, the most cited papers are rarely what we want (highly cited papers are easily found through ISI's Web of Knowledge in any case); often, when we want to cast our net wider than the restricted journal set chosen by ISI, we are looking for good papers that have been missed out by the citation machine. It is the old complaint that 'popular' is not necessarily 'good', and this problem is magnified rather than mitigated by Google's search engine, which relentlessly seeks out the popular.

Even then, does it really find the most cited papers? As ever, I looked for myself, and found my most highly cited paper was actually the most difficult to find using keywords. In the end I had to resort to typing the exact title in (and yes, it was more highly cited according to Google Scholar than any of the other entries it had found). The problems were twofold: firstly, Google Scholar had indexed the same paper under several different entries; secondly, it had failed to pick up the freely available copy of the paper available through our publication server at UCL, eprints, and instead gone for the restricted access copy held by the publisher. Yes, these are presumably teething problems with what is still a Beta test, but it all sets a worrying trend for the future when academics will try to get their papers to the top of Google, and not strive to produce the best research they can.

More on Procedural Cities

Courtesy of the newly launched and very handy Google Scholar, a follow-up to last month's post on procedural urban modelling.

The system described in Parish and Müller's 2001 paper Procedural Modeling of Cities (pdf) is lovely, and generates some convincingly natural-looking urban forms:



In particular, the similarity between their Manhattan (above) and the real thing (below) is very impressive:



Greuter, Parker, Stewart and Leach build upon those ideas and provide a real-time system for urban environment generation in Real-time Procedural Generation of 'Pseudo Infinite' Cities (pdf), which also includes a neat way of generating interesting building models by modifying and extruding floor plans.



Bringing it back home to the Bartlett, Azari Mat Yasir (one of my colleagues on last year's MSc VE course) produced a VRML world called Hatch City as a part of his coursework. It requires the Cortona VRML plug-in from Parallel Graphics, but it's well worth a look. When the city develops far enough, little red cars appear which is cute touch.

Living Escher

On his Living Escher site, Marcello Bastéa-Forte has been experimenting with interactive animations of M.C. Escher drawings. The results so far are stunning.



Along similar lines, Escher For Real uses rapid prototyping to explore 3D manifestations of Escher's drawings.

Kinecity - World Trade Center Installation

Working as Kinecity, Marek Walczak and Jakub Segen have designed a podium light wall for 7 World Trade Center. The spatial studies by Martin Wattenburg are particularly interesting, it's a shame that only one of them is presented at full scale (seven stories of a building).



Elsewhere, Marek, Jakub and six others are taking part in FutureFace, a series of workshops on the future of the interface, held over Saturday brunch - what a pleasant idea!

Marek and Martin have also collaborated before on an increasingly diverse and intriguing selection of projects. Of their recent work, ThinkingTable and DialogTable are stand out pieces.