So Google has just launched Google Scholar, which, according to this NY Times article...

...is a result of the company's collaboration with a number of scientific and academic publishers and is intended as a first stop for researchers looking for scholarly literature like peer-reviewed papers, books, abstracts and technical reports.

So, I thought: Cool.

I've been universally impressed with nearly everything Google does. So I took it for a test drive. And unfortunately, I'm pretty underwhelmed.

First of all, I consider myself an amateur AI researcher, and I read a fair amount of academic research. My primary tools have been CiteSeer and just good old-fashioned Google.

But Citeseer seems to have server problems fairly frequently, and doesn't seem to do a great job of indexing and referencing recent stuff (like from the past 2 or 3 years). So it'd be great to have an alternative. But at least in Beta, it doesn't look like Google Scholar is it...at least for the kind of stuff I look for and read.

Let's take a subject I'm working on right now.

I tried Google Scholar with "fingerprint classification". Out of the first ten hits I get 3 citations, 5 links to portals (that require membership or charge a fee to read the actual paper), and only 2 actual papers (neither of which are very prominent pieces of work). The next 10 hits are better, though. But it's a bit sad that the actual regular Google search for "fingerprint classification" actually yields better information, pointing, for example, to the work being done at MSU.

So I do another search, this time for "active vision".

This time 3 out of the first 10 are actual papers. The rest are portals or references. Not good.

I'd like to be able to sort hits by type (citation, portal, or actual document). And I'd like to be able to sort my hits by date...what researcher wouldn't?

Sorry, Google...in at least the Beta version, your newest tool, one that I'd love to use, just ain't that great.