Thursday, July 27, 2006

There is no shelf.

I just finished listening to the latest Talis Library 2.0 Gang podcast (on headphones though, so Tim wouldn't wince at the sound of his own voice), this week on tagging and folksonomies. I'm struck by one thing that Casey Bisson mentions. He relates a story of a patron asking if they can put all the books he's interested in, and just those, on one shelf--the answer of course, is to laugh, of course the library won't be rearranged just for that one person!

In a virtual library, however, you can pull all the books you're interested in, and just those books.

In a library, each book has a call number, which places it on the shelf somewhere, next to the other books that have similar "aboutness." I might start looking for At home in the studio: the professionalization of women artists in America in the history section, because I know the author is a historian--but I wouldn't find it there, because the Library of Congress Call Number is an N (Fine Arts).

The answer to the patron at Casey's library, is that it doesn't matter where the books are physically located in the library - as long as he can pull them together intellectually.

For the purposes of shelving, in a library, At home in the studio is "about" fine arts - it can only be on one shelf, after all.

If that book was in my LT catalog though (and it is), it can be "about" multiple things--there doesn't have to be one essential place that it lives. I could tag it (and I have) so I could find it when I was searching for my history books, and when I wanted my art books. It can be on both "shelves", on any shelf - because there is no shelf (oh, postmodernism)...

Ok. Enough rehashing. Go bend spoons with your mind.

8 Comments:

Blogger RJO said...

(Puts on contrarian hat. I know you know this stuff, but I think some of the computer people don't.)

Ah, but those who don't remember history are condemned to reinvent it and think they have something new.

What you're describing is ... a card catalog. More specifically, it's the radical innovation of old Charles Ammi Cutter, circa 1890: the dictionary catalogue. In Mr. Cutter's invention, when you look in the dictionary catalogue for "history of knitting" you find not only books with the title "History of knitting" but also (right next to them) works by the corporate author "The History of Knitting Society" (on a different floor, shelved with learned society publications) and even (right next to them) books titled "K1 P2" and "Textile Automation Methods" (shelved on a different floor again), all assuming the cataloger (read, "professional tagger") gave them subject added entries under "History of knitting".

For all the whiz-bang of LT (and it's very whiz-bang, of course), it is still chained to a one-book=one-record data model. Old Mr. Cutter broke that model over a century ago.

7/27/2006 8:26 PM  
Blogger Tim said...

I didn't enjoy this one so much--entirely for vanity reasons. I felt tongue-tied and out-of-step. It might have been the heat, but I credit it to the fact that I've spent so much time thinking about tags--my head is so crammed full of them--that I didn't know where to begin. This is why people like me give talks--to clarify and organize our thinking.

I think I came off a little defensive. Although tags-on-books aren't the only way that tagging and libraries connect, they are the most important way. And LibraryThing has orders more tags, and, I think, better tags than anyone else. (I also think this is likely to remain so for some time.) Working with them day-in and day-out has given me a lot of insight into what people are doing, and what it might be useful for. I have come to see that people tag books quite differently from how they tag, say photos in Flickr. And they tag them differently in Amazon and LibraryThing.

Meanwhile blog posts and papers come out about what Del.icio.us and Flickr's tagging means to library information. When LibraryThing *is* considered, analysis is very binary—"tags are great," or "tags are terrible." Hey people, they're both! Let's see how and where...

This is starting to change. I stood up and promoted LibraryThing data for academic discussions, and Paul Miller has blogged about it ( http://blogs.talis.com/panlibus/archives/2006/07/a_great_opportu.php ). We've had a few inquiries. Let's hope we get more.

7/27/2006 8:33 PM  
Blogger Tim said...

RJO. I appreciate contrarianness, but I'm confused about the point. Although LibraryThing does store it's data records book-by-book, it has an elaborate (and innovative) "works" system to bring it all together. That LibraryThing doesn't list a subject next to two titles next to an author next to a "see also," etc. as if they were all the same sort of thing is a strength, no?

Incidentally, someone should say that the "There is no shelf" motto is from Clay Shirky's "Ontology is Overrated." Although I think the brothers Wachowski get ultimate credit.

7/27/2006 8:38 PM  
Blogger RJO said...

I'm confused about the point. Although LibraryThing does store it's data records book-by-book, it has an elaborate (and innovative) "works" system to bring it all together. That LibraryThing doesn't list a subject next to two titles next to an author next to a "see also," etc. as if they were all the same sort of thing is a strength, no?

Well, I don't think it is necessarily, though it may be in certain instances. (I'm thinking this through by talking about it also, because it's so interesting.)

The simplest way to see the weakness of the one-book=one-record model is with multiple authors: if I have LT sort my catalog by author and browse down through "Wallace" I'll never see anything by "Darwin & Wallace" because there can only be one "copy" of each record in a given list. An old paper card catalog (dictionary catalog)) has many "copies" of each record: one for each author, one for the title, one for each of maybe ten subject headings, etc.

If my LT catalog has 2000 books in it, I can never generate a display with more than 2000 rows. But a card catalog of the same collection might have 10,000 "rows" (cards).

The same would be true for any attribute of a record that can occur in multiple instances per record, such as tags. What LT can beautifully do is subset a collection and sort those subsets, but it can't "superset" a collection (multipply access points within a single view), as a paper card catalog does.

Another straightforward example would be: show me my collection, sorted by LC subject headings. This would necessitate multiplying the records in the display so there would be one instantiation for each heading. Or if you prefer tags: show me my collection sorted by tags, in such a way that each book will appear on the screen as many times as it has tags.

(This may well be exactly what you will produce in LT when you get multiple authors in place, for all I know.)

7/27/2006 9:06 PM  
Blogger Tim said...

>[A] straightforward example would be: show me my collection,
>sorted by LC subject headings.

This, I think would be useful. Or, at least a list of the subjects IN your collection would be useful. In general, the LT solution would be to provide a list of subjects (or tags, or authors, etc.) and then allow drill-down. Providing a long list you have to scan through is what you do when you don't have the option to search.

>If my LT catalog has 2000 books in it, I can never generate
>a display with more than 2000 rows. But a card catalog of
>the same collection might have 10,000 "rows" (cards).

This is what I mean. While a card catalog has uses, it's use is to allow searching on data in a non-digital environment. LibraryThign COULD provide a card-catalog-like display, taking 200 web pages of unique books and turning them into 1,000 or 2,000 web pages of books-subjects-authors, etc. But paging through to page 760 or jumping there ("whoops, too far / whoops, to near") seems like insanity to me when you could just use a search box, or click on an instance of the term.

Now, there's a question of whether LibraryThing should be able to output card-catalog like output FOR THE PHYSICAL WORLD...

7/27/2006 11:35 PM  
Blogger RJO said...

(This is all exceedingly interesting. And where, if not here? We have broad enough experience to be able to take on the issues, and aren't so professionally narrow that our thinking is boxed in. We know just enough to be dangerous.)

>>[Sorting your catalog by LCSH] I think would be useful. Or, at least a list of the subjects IN your collection would be useful.<<

A display that showed all the subject headings, one per line, with an indented short-title line below each heading that listed the books under that heading might be nice. But see below for an even better solution.

>>In general, the LT solution would be to provide a list of subjects (or tags, or authors, etc.) and then allow drill-down. Providing a long list you have to scan through is what you do when you don't have the option to search. [Emphasis added]<<

I disagree with the last point. I could humorously invert it and say: search is what you do when you already know the answer to your question; scanning is what you do when you're trying to formulate a question to ask. The first is mechanical; the second is creative. (As a noted philosopher of my acquaintance once said, "Dictionaries don't help you do philosophy"; that is, you can't find the meaning of life by looking up "life" in the dictionary.)

>>LibraryThing COULD provide a card-catalog-like display, taking 200 web pages of unique books and turning them into 1,000 or 2,000 web pages of books-subjects-authors, etc. But paging through to page 760 or jumping there ("whoops, too far / whoops, to near") seems like insanity to me when you could just use a search box, or click on an instance of the term.<<

I think you missed the point here, or more likely I didn't explain it clearly. It has nothing to do with one-per-page display, it has to do with more-than-one-display per book. Here's an anomaly that may illustrate the point:

By historical accident (I suspect), in LT you cannot sort on the Subject Headings column, but you *can* sort by the tags column. Given the way LT works currently, the behavior of the Subject Headings column is really the right one, since all that sorting the tags column will do is sort by whichever tag happens to be first. If I tag one book "dogs, cats" and another "cats, dogs" then each book will appear only once, at widely separated locations. There is no way to browse tags. ("But there is!" you cry: "just go to the tags page and drill down!" But drill-down is rarely my favored mode of browsing because it's always back and forth, back and forth; it imposes much greater cognitive load than scanning down a single sequence; but hang on....)

Here's what I'm talking about: when I click on the "sort by tags" column, what LT *should* do is automatically explode my catalog into as many records as there are tags. As I scroll down through the C's I should find "The Big Book of Cats and Dogs" because it's tagged "cats"; and then as I get down to the D's the same book should appear again because it's also tagged "dogs". That's the behavior of a "dictionary catalog" (a term of art), which is what most physical library catalogs have been since the late 1800s.

The same behavior should occur when I sort on the LCSH column: LT should automatically explode the catalog into as many rows as there are subject headings. (It's also how to handle multiple authors.) That would be majorly cool.

7/28/2006 9:26 PM  
Blogger RJO said...

Subject line I should have used for this thread: "Loaves, fishes, and exploding catalogs"

7/28/2006 9:53 PM  
Anonymous kbronte said...

Please I need help. I can't understand how to download book covers. I tried to download one from Amazon for the group I started but,alas, after a frustraing hour I gave up! I even tried to capture one from my catalog to put on the group page but that doesn't work either. Is there a place on the site where members can see a help menu??PLEASE HELP.kbronte

7/29/2006 9:41 PM  

Post a Comment

<< Home