Alexander Johannesen wrote:
> On Wed, Mar 18, 2009 at 02:20, Weinheimer Jim <j.weinheimer_at_aur.edu>
> wrote:
> > It would be nice if it were that simple, but Google's algorithm (the
> entire strength of
> > Google) is based on trillions of links to all different sites (the page
> with most links
> > to it by the most linked = #1). There's nothing like that option in the
> library, and
> > even Google's algorithm isn't so hot in Google Books.
>
> Actually, this was true a few years ago. They've moved on, and other
> things are at play now. Besides, all it takes for this to work in
> libraries if links are (indeed still) the main stew booster, is for
> libraries to properly share their stuff! Not hard at all. C'mon, make
> it easier for Google to help you out.
Yes, other things are at play. For example, Google is a highly-secretive company and there have been some suspicions on tampering with results. One thing I heard on a great podcast recently with Siva Vadiyanthan, who is writing "The Googliization of Everything" http://gslis.simmons.edu/podcast/index.php?id=64, (I suggest it to everybody) is something that probably everyone has noticed. Have you noticed that rather suddenly when you do a search for something in Google, the Wikipedia page very, very often pops up to number 1? As he said, it could be that Wikipedia is becoming to be seen as "the" source by the world (which has implications in itself), or it is just as likely that Google may be tweaking the results for whatever reason.
We know they tweaked it for the major Google-bombing episodes, probably to save George Bush and the Republican party some embarrassment.
As I mentioned in a previous post, Google does change, but it is very difficult to conclude that the changes are improvements for us. Of course, they will definitely be improvements in Google's opinion, since that is the nature of a private corporation. But lacking any kinds of standards or yardsticks by which to measure Google search results, it just cannot be concluded that they are getting better. At least in Google Books, I can state that in the areas I happen to know quite a bit about in the bibliographical aspects, retrieval has gotten no better since it began, although there are more (wonderful!) things in there.
> > Google's ranking by "relevance" (a semi-propagandistic term
> since it means
> > something quite different from the normal sense of "relevance")
>
> No it doesn't; It means whatever it means in the context of where you
> are, just like in real-life. Within Google it is relevant to the words
> you typed in. Don't like the relevance? Switch your words, just like
> in real-life.
As I said: propaganda. When words cease to have meaning, or they are chosen to mean something quite different from the established idea, this is a basic point of advertising. So, in the U.S. when you go to a grocery store, what you may think that "no-salt" or "no-fat" means may be quite different from what it *really* means, that is, what the company and producers make.
I worked in grocery stores for many years, especially in produce, and I remember when it first struck me when I noticed all the tomatoes labeled "vine-ripened" were green! I discovered it wasn't a lie, it was simply a very different meaning of the words "vine-ripened" than I expected.
"Relevance" is an example of this sort of advertising propaganda. I'm sure the Google people thought long and hard over that term, but it is *absolutely not* "the relation of something to the matter at hand" (definition taken from "The Free Dictionary"). It is what is most linked to as determined primarily by bloggers, obsessive-compulsive internet freaks, and the "page-rank" experts hired by companies. We know Google plays with it to an extent. Who knows what governments are doing but I suspect they are working with it a lot.
Again, I am not criticizing Google but it is vitally important to understand the strengths and weaknesses of the tools that we use. Librarians have always used lots and lots of different tools for different purposes and have always taken note of the strengths and weaknesses of one index over another, or of one reference source over another. I guess this is why it comes so natural to us.
> > would need to be recreated in the catalog, but how? By items most checked
> > out (most popular?) By getting into publisher databases and trying to
> arrange
> > by printing statistics? Or by retail statistics and best-sellers?
>
> Ah, well *now* we're cookin'! :) I've got heaps of stuff about this,
> mostly prototypes and hacks before I quit the library world, things
> like "Heat Engine" which uses inverse cumulative histograms to track
> real popularity of books (without the dreaded short-term effects of
> 'peak', and deals with normative decline as opposed to pure
> statistics), or the "Memory Peak" (dealing with books borrow history,
> tracking subject headings over time and match it against keywords
> people search with in the OPAC), another system for mapping website
> searches against catalog searches and finding corrolations, or if you
> try http://ll01.nla.gov.au Kent Fitch (bless his heart!) played with
> the ABC news feed, pulling it down and try to find resources that
> somewhat matches the news items in question (right-hand side box).
> Funky and fun, and sometimes really helpful and relevant.
>
> In fact, library developers (and not just programmers) should be
> spending a lot of their time trying this stuff out and thinking about
> new ways to eal with what you've got, because, well, it's what you've
> got, and you won't get much else by the sound of it. :(
>
> > Or by "rate this book!" Let's say that Nietzsche's "Thus
> spoke Zarathustra"
> > got 200 votes while Kant's "Critique of the Pure
> Reason" only got 50. What
> > would somebody conclude?!
>
> Well, there's other and better ways. For example, make your OPACs and
> catalogs more in the vein of social websites, and introduce roles on
> it where librarians can overlay an expert layer over the data. By
> that, reference librarians can surf and search around, tagging their
> books, make lists of recomendations and so forth. Make your systems
> with more roles in them, the *same* system, and this will open up
> opportunities you just don't have right now.
>
> > While some of these tools are interesting, I'm not sure which ones really
> belong in a library....
>
> Again, a friendly reminder that your users are ... *everyone*. So yes,
> they probably belong in the library.
Yes, I agree they are worth a try. But in this age of diminishing resources and results-oriented management, it becomes extremely difficult to justify the time and expense.
Still, we need to try.
Jim Weinheimer
Received on Wed Mar 18 2009 - 04:46:39 EDT