The hard work for showing that a problem is NP-complete (=NP-easy and NP-hard) is pretty much always to show that the problem is NP-hard. That is what the paper by Demaine and Demaine show. Namely, that Eternity II is also NP-hard and so a polynomial-time solution to it (in the general case) gives a polynomial-time algorithm for every NP-easy problem.

In the other branch, often called higher complexity theory, we do pretty much same thing except we look at higher complexity classes, find relations between them, and find ways to deal with the hardness of the problems, e.g., with approximation. One topic in this field is that of exact algorithms, which derives upper and lower bounds on the computational resources of NP-hard problems resting on the assumption that there is no subexponential-time algorithm for 3SAT (called the exponential-time hypothesis). For Eternity II-type puzzles, it should be quite doable to derive such lower bounds from that hypothesis.

I think my point is – some classification systems are better than others.
Social tagging and recommender systems have their Achilles heel namely: how do we trust them?

My hunch is that even when e-books replace physical books in a library, it will still be worth while to get a human brain to verify whether a software-based keyword extractor / subject classification system did a reasonable job. I think I’m jaded about the endless unfulfilled promises that AI will replace humans (adequately).

The Dewey system is a workaround for our inability to index physical objects in rich ways. I can’t very well do full text search on physical books, can I?

In the digital age, we no longer need Dewey, except maybe at the public library… until we move to e-books… which should happen with a decade or so.

http://www.citeulike.org/groupforum/1606

Specifically, we need to have collected the metadata ourselves rather then manually entered or uploaded. Our article matching is derived from the URL so manually entered has not much to go on. (I’m over simplifying here.)

For the record, our recommender system isn’t one that Toine discussed in that poster.

Regards,
Eternity II Blogger.

This is very interesting. From a genre perspective, I can see the lines of “email” and “instant messaging” getting blurred here, which may or may not be a bad thing. It would be neat to see which register survives. You know, talkin liek im kool, eh? Or would I in fact, type like like a regular person when “emailing”?

I think it is really neat, but perhaps really distracting! I can just see myself not getting any work done because it would be so easy to join in wave after wave of (social) conversation. I do, however, see its great potential if you were to collaborate with coworkers on a project. Very cool.

On thing that made me especially curious was:
“Another great feature is the advanced spell checker that not only looks through a dictionary for spelling, but looks for context in your sentence. It’s crazy to think that this could work, but it does as they’re showing it off.”
Context! Incredible! How does this thing know semantics/pragmatics, given only text without any intonational markers?!

But — I remind myself that this will take a lot of time. I’m still waiting for a Google Chome port here in the Linux world…

The data I have is *very* sparse. On about 8M science articles, I’m only able (currently) to produce item-based recommendations (based on citations only) for about 1.8M of them. I’m now working on reducing the sparsity of my data….

There’s also going to be an OpenURL API, very much like ‘bX’ – submit an OpenURL to Synthese and you get back metadata about the recommended articles in XML. Should be done in September sometime.

Suppose that I say “No, I don’t want to cite paper W even though I plan to cite papers X, Y, Z”… then what does it say? Can you then focus on papers (if any) citing papers X, Y, Z but not paper W?

The fun thing is that these are small data sets (very sparse) so you can do the computations live with little effort.

Have you tried something like this: papers who have cited papers X, Y, Z have also cited paper W? I’d love to get an analysis of a paper I am about to submit to see whether I have omitted any reference… It would be cool to determine, mathematically, whether a set of reference is “complete” in some sense.

(There has been a lot of work done on mining frequent “minimal” item sets. I am guessing it could be applied to this problem.)

Minimally, just this feature: papers which have cited this paper have also cited this other paper… that’d be great.