Friday, September 21, 2007
Another Player in Semantic Search
A startup called Powerset debuted a new system of semantic natural-language Web searching on September 17. It's currently in a closed alpha, so I haven't been able to poke at it, but it sounds both ambitious and wonderful. Plus, it's build by PARC—the same folks who invented computer mice and GUIs. Here is the page for PARC's natural language processing research project, which is the technology used by Powerset.
Saturday, September 15, 2007
One more reason to be skeptical about information published in journals
I really need to be preparing for my job interview on Monday (everyone wish me good luck and pray that Northwest Airlines manages not to make a hash of my flights!), but this story is so fascinating that I had to blog it right away.
Dr. John Ioannidis, an epidemiologist, has compiled evidence indicating that the results of the majority of published studies in the sciences are incorrect. He has also analyzed some of the reasons for this, stemming from the incentive structures of publishing and academia. Ioannidis doesn't seem to use the term confirmation bias—one of myhobbyhorses areas of research interest—but I think he'd agree that confirmation bias probably plays a pretty big role here too.
Dr. John Ioannidis, an epidemiologist, has compiled evidence indicating that the results of the majority of published studies in the sciences are incorrect. He has also analyzed some of the reasons for this, stemming from the incentive structures of publishing and academia. Ioannidis doesn't seem to use the term confirmation bias—one of my
Tuesday, September 4, 2007
More about e-book sales
Looks like I may have spoken too soon about ebooks not catching on in the U.S. According to the International Digital Publishing Forum, U.S. trade ebook sales broke $8 million in the second quarter of 2007, up from under $2 million as recently as the fourth quarter of 2002. And that "does not include library, educational or professional electronic sales." It's still peanuts compared to print publishing revenues, but it's a sharper upward trend than I thought.
This is why one should always check the numbers and not just believe whatever the conventional wisdom is on a subject.
This is why one should always check the numbers and not just believe whatever the conventional wisdom is on a subject.
What was that about e-books being doomed?
In Japan, e-books designed for cellphones outsold print books in the first six months of 2007. (Hat tip: LISNews.)
I wish this sort of thing would catch on in the U.S. In general I prefer reading on a screen to reading on paper—I appreciate having the ability to search the full text for words and to control the font, the type size, etc. And, since it's not particularly uncommon for me to spend 12 hours a day at the computer, I've invested in a nice setup—a big high-quality LCD screen, ergonomic keyboard and trackball, and a good desk chair—so I'm actually more comfortable sitting at the computer than I am sitting on the sofa or in bed or all of those other places that people say they prefer to read. But I very rarely read true e-books (of the type carried by NetLibrary or ebrary) because the interfaces on them are so awful. Actually, for the longest time I couldn't use ebrary books even if I wanted to, because their proprietary reader didn't work on Linux and I wasn't about to boot into the other side of my dual-boot setup to access their books. (I'm a messy-desktop person—I usually have several dozen Firefox tabs, a dozen or so Thunderbird windows, and half a dozen word processing documents open at once. Closing them all, booting up into Windows, and then going back to Linux and trying to remember what I had open and why and re-opening them all is a pain that I'm only willing to go through for a very small number of things.) I've tried NetLibrary, but I got frustrated at my session timing out and losing my place. I was trying to use a NetLibrary book to write a paper, so I wanted to be able to refer to the book, refer to other stuff, write for awhile, pace around for awhile, and then go back and refer to the book again. No dice—every time I went back I had timed out and I had to start from the beginning to find the book and my page again. Also, at this point the majority of the time when I'm reading a book I'm doing so with an eye towards using an excerpt from the book in one or another anthology that I'm editing for Greenhaven, which means I need to be able to print a copy of the chapter I want to use. Except (at least the last time I tried this) NetLibrary is understandably not so keen about people printing out entire chapters because of the potential for copyright infringement.
But my point is, these aren't problems with e-books per se; they're problems with the current e-book interfaces. Unfortunately, I don't know what it's going to take to convince the e-book vendors to either improve their proprietary readers or to serve books in plain old let-me-do-what-I-want-with-it HTML. *sigh* Maybe I should just move to Japan.
I wish this sort of thing would catch on in the U.S. In general I prefer reading on a screen to reading on paper—I appreciate having the ability to search the full text for words and to control the font, the type size, etc. And, since it's not particularly uncommon for me to spend 12 hours a day at the computer, I've invested in a nice setup—a big high-quality LCD screen, ergonomic keyboard and trackball, and a good desk chair—so I'm actually more comfortable sitting at the computer than I am sitting on the sofa or in bed or all of those other places that people say they prefer to read. But I very rarely read true e-books (of the type carried by NetLibrary or ebrary) because the interfaces on them are so awful. Actually, for the longest time I couldn't use ebrary books even if I wanted to, because their proprietary reader didn't work on Linux and I wasn't about to boot into the other side of my dual-boot setup to access their books. (I'm a messy-desktop person—I usually have several dozen Firefox tabs, a dozen or so Thunderbird windows, and half a dozen word processing documents open at once. Closing them all, booting up into Windows, and then going back to Linux and trying to remember what I had open and why and re-opening them all is a pain that I'm only willing to go through for a very small number of things.) I've tried NetLibrary, but I got frustrated at my session timing out and losing my place. I was trying to use a NetLibrary book to write a paper, so I wanted to be able to refer to the book, refer to other stuff, write for awhile, pace around for awhile, and then go back and refer to the book again. No dice—every time I went back I had timed out and I had to start from the beginning to find the book and my page again. Also, at this point the majority of the time when I'm reading a book I'm doing so with an eye towards using an excerpt from the book in one or another anthology that I'm editing for Greenhaven, which means I need to be able to print a copy of the chapter I want to use. Except (at least the last time I tried this) NetLibrary is understandably not so keen about people printing out entire chapters because of the potential for copyright infringement.
But my point is, these aren't problems with e-books per se; they're problems with the current e-book interfaces. Unfortunately, I don't know what it's going to take to convince the e-book vendors to either improve their proprietary readers or to serve books in plain old let-me-do-what-I-want-with-it HTML. *sigh* Maybe I should just move to Japan.
Sunday, September 2, 2007
More Holiday Weekend Reading
I've had the June 2007 issue of Webology open in a Firefox tab since, oh, June or so, because it had a couple of articles on folksonomies and ontologies that I intended to read and blog about. Today I finally got around to reading them . . . and I have nothing to say about them except, read them. There's nothing too surprising in either of them, but they're still worth a look.
Saturday, September 1, 2007
Speaking of Economists and Libraries....
This article on how patients would get more accurate diagnoses if doctors used computer algorithms rather than their own “clinical judgment” has been getting some attention on econ blogs.
What does this have to do with libraries? Read the debate that's been raging on the NCG4LIB listserv about the potential that a well-coded algorithm using Bayesian inference could possibly do authority control work as well as if not better than human catalogers, and ponder how much the catalogers who decry this possibility might have in common with the doctors discussed in the article.
What does this have to do with libraries? Read the debate that's been raging on the NCG4LIB listserv about the potential that a well-coded algorithm using Bayesian inference could possibly do authority control work as well as if not better than human catalogers, and ponder how much the catalogers who decry this possibility might have in common with the doctors discussed in the article.
Link Roundup
An interview with the man behind Google Scholar. (Hat tip: LISNews.org.)
How many books should you start?, by Tyler Cowen, who is on the economics faculty at George Mason University. More LIS folks should read econ blogs like Marginal Revolution. One of the big movements in economics right now is behavioral economics—drawing on both psychology and economics to understand why people do what they do with the resources they control, including their time. I think that if more LIS folks understood the incentives that people respond to, they'd have an easier time designing services that get used and selling their services to their funding authorities.
How many books should you start?, by Tyler Cowen, who is on the economics faculty at George Mason University. More LIS folks should read econ blogs like Marginal Revolution. One of the big movements in economics right now is behavioral economics—drawing on both psychology and economics to understand why people do what they do with the resources they control, including their time. I think that if more LIS folks understood the incentives that people respond to, they'd have an easier time designing services that get used and selling their services to their funding authorities.
Subscribe to:
Posts (Atom)
