Bascule, Phil, and Yourdad won! Longshot guess proved right.

[Martin]

This?

http://www.slac.stanford.edu/spires/

 

that's right! spires is a good database. I don't think their "topcites" lists are quite as extensive as they used to be. they changed format a year or so ago.

but on the whole it's very good.

 

they have librarians in Germany that go through the papers and TAG them with various keywords like "quantum gravity". the librarians' definitions and criteria may sometimes not be perfectly useful----they may call a lot of stuff "quantum gravity" and tag it, stuff that I don't want to know about. they also might make human error.

 

Have a keyword list you want me to check?

 

Not at the moment, thank you. I'm interested in quantum cosmology but I don't think the librarians have "quantum cosmology" in their list of keywords.

The list of keywords used by the DESY librarians (the ones responsible) is available in the spires "Help" section if you are curious.

 

But you have enough curiosity and imagination to think of something to search for.

 

You can boolean----so you can search by author AND by topic AND by date AND by 100+ citations etc. etc.

 

So... think this could be automated?

 

No, I don't. The catch is the human element. When I do a keyword quantum gravity search I always have to go thru the list manually to pick out the stuff that is LQG-related or at least non-string QG.

 

But you could automate ALMOST all of it. The important thing is to remember to pick the option where you get the hits sorted by approximate citation count. You have to find the thingee where they suggest you select NO SORT (fastest) and select from that menu. Get it ordered by citation count and then the highly cited papers will come first---and they will tend to be LQG-related and have a larger fraction of interesting ones.

 

Just for fun, try this

FIND K QUANTUM GRAVITY AND DATE = 2006

with "sort by approximate citation count" as the ordering option

and the default "standard" format

You will see that #1 and #2 are loop, and then #3 and #4 are string and #5 is God-knows-what about a quintessence particle called a "quintom"

But then #6 and #8 are loop-related----#8 is Martin Reuter who is featured at Loop conferences although his approach is nominally something else.

So the citation sort does a lot of the filtering!

We get #1, 2, 6, 8, 12,....

 

It's EASY but I cant think of any way to automate it. You need someone who can recognize that #3,4,5,7,...don't belong.

[bascule]

It's EASY but I cant think of any way to automate it. You need someone who can recognize that #3,4,5,7,...don't belong.

 

A Bayes classifier comes to mind... if you had a large enough set of papers flagged "loop" vs "string" vs "neither" to train it with...