• LibraryThing
  • Book discussions
  • Your LibraryThing
  • Join to start using.

Findability vs. Accuracy

Build the Open Shelves Classification

Join LibraryThing to post.

This topic is currently marked as "dormant"—the last message is more than 90 days old. You can revive it by posting a reply.

Mar 9, 2009, 7:26pm Top

This was inspired by some discussions in other threads, but I hope the examples I've chosen aren't real. My intent isn't to hold anyone who has a particular misconception up for ridicule but to start a discussion about the general issue of what to do when findability and accuracy may come into conflict.

There are a number of cases where some fraction of the population won't look in the "right" place for a book. Someone might look under "Insects" for a book about spiders, or under "Planets" (rather than something like "Other Solar System objects") for a book about Pluto. These aren't the places where someone familiar with the subject matter would look for those books, of course, and in an objective sense they're the "wrong" place -- but if that's where people will look, is that where they should go?

I cynically think that people who are informed enough to know the "right" place to look for books in cases like this are also more likely to be able to use the library's computers to find where the books actually ended up being filed, while someone who's operating under the wrong assumptions might be less likely to do so. Is this at all a reasonable thing to think?

Mar 9, 2009, 7:55pm Top

This message has been deleted by its author.

Mar 9, 2009, 8:17pm Top

2> But there are solutions in place already, which we shouldn't forget, but make sure we design OSC to make proper use of them. One is, after the groups are determined, how do we sequence the groups on the shelf?

Sure, and that's by far the most natural approach -- unless the top-level groupings are themselves designed such that the "wrong" place would be very far away. Probably the ideal solution is to put everything in the "right" place, and make sure that there aren't these huge gaps.

Part of what prompted this post was the realization that many people define "animal" in odd ways, which made me rethink my opposition to "Pets" as a top-level (preferring to see it replaced by "Animals", as it is in many bookstores). If someone doesn't actually think of a snake or a tarantula or a fish as an animal, and the category on caring for these critters is under "Animals" rather than "Pets", they might be totally lost, while although "Science" might not be the most immediately obvious place for a book about feeding birds it's at least the most obvious of the remaining alternatives.

Mar 9, 2009, 8:39pm Top

This message has been deleted by its author.

Mar 9, 2009, 8:53pm Top

This is where there seems to be a disconnect with the goals of OSC, as proposed by Tim. This isn't supposed to be some grand scheme to organize all knowledge; it's supposed to be base on how people actually catagorize things. And since I don't think that many people tag their books with Living Things, I don't think it will fly as a category.

Mar 9, 2009, 9:32pm Top

This message has been deleted by its author.

Mar 9, 2009, 10:08pm Top

>5 jjwilson61:, Isn't it just unrealistic to design a classification system that's based on the way that things are tagged? At the end of the day, the system generated by OSC will be effectively one-dimensional; that is, there will be a single correct shelf order, just like in Dewey or Library of Congress. As a result, each book can go under only one top-level category, and all subcategories have to be fully nested within their top-level categories. Even if OSC builds in options for libraries to select among, each library individually will have to come up with the single shelf order that best suits its needs.

Tags are much more flexible than that. Just as an example, I have the Bhagavad-Gita tagged as both "Religion" and "Poetry," both of which would be helpful to me in searching my library, but neither category could be said to include the other. I think that tags are just too flexible to be used as a basis for something as one-dimensional as shelf-order.

>1 lorax:-4, You guys have described a good test for any cataloging system: can the average patron, whose understanding of both the subject matter and the cataloging system is imperfect, make his way to a shelf close enough to what he's looking for, such that he'll be able to find it with a minimum of trouble? In designing the system, as you suggest, we can take advantage of predictable misunderstandings, so books on arachnids should be shelved beside books on insects, etc.

But what I'm now wondering is how bad a job do Dewey and LC really do on this front. I mean, do we really need to reinvent the wheel here? OSC is an interesting project, but if the problem can be fixed with tweaks, then why not just tweak? Libraries have found it efficacious to pull genre fiction out of the Dewey 800s. Why not do the same for books on pets, if the library determines it has a need? Pull all the books on pets into a separate collection, on a separate set of shelves, with a big sign above them that says "PETS," and shelve the books by Dewey call number on those shelves. Same for CAREER, and FAMILY & RELATIONSHIPS, and any other grouping that a particular library finds its patrons are always asking for, but have a hard time finding.

I'm beginning to think that the ambition of this project needlessly exceeds the difficulty of the problems that it's supposed to solve.

Edited: Mar 10, 2009, 3:49pm Top

Tagging is a very 21st century, post-modern concept that works in the non-linear world of data. It really has changed the way think about organizing things. (Or perhaps, the concept of tagging was invented because we had changed the way we thought about organizing things.) Classification is a very 19th century concept that suggests every single thing has a single correct place. That doesn't quite work these days.
However, physical objects can only be placed in one place, so we do need to find if not the correct places then the best places for them. Basing a classification system on a tagging system may be doomed to conceptual failure. But I do think it is worth a shot.

Maybe an entirely faceted system is where we're headed, as a lot of the discussions--comics, biographies, science v. pets.--have ended up suggesting an optional facet as a solution.

Mar 11, 2009, 10:14pm Top

This message has been deleted by its author.

Mar 11, 2009, 11:10pm Top

Just a few comments:

A patron who is looking for something quite specific usually asks for help or goes to the catalog. However, a lot of people do what might be called a browse/look, where they want they want to see what's around in a general area. In public libraries, this is very common for new books.

My own local pubic library has three locations for each non-fiction call number: regular shelf, reference shelf, new books shelf. For fiction, there are the regular 4 week shelves, the regular two week shelves, the rental two-week shelf, the rental one-week shelf. To add to that, some adult fiction is filed under Young Adult, and occasionally some fiction is on both the YA and the regular shelf. Some literary fiction is filed in the 800's. Then there are foreign language shelves, usually for fiction. So, there is a a lot of malleability, and things are not quickly findable even if you know the call number or authors last name.

I like the idea of paying some attention to sequencing. In DDC, it is not surprising that religion (200's) and philosophy (100's) are not distantly located with respect to numbers, but science (500's) is. I'm not sure how much attention has been paid to what kind of letter or number systems has been proposed for OSC.

With regard to changing call number systems, it is the academic libraries that have done the most change in the last 50 years. A large percent of university libraries were using DDC in 1950, but nearly all are now in LOC, if only for books acquired in the last 35 years. The only Dewey change that I remember from the time I was a kid was the elimination of 656 for railroads (as a business - leaving 385 and 625).

But I do remember when I was using the Boston Public Library in the 1960's that there must have been at least five different systems in use, depending quite a bit on when the book was purchased. But most of their books were in closed stacks, including those in off-site storage. The problem here for the patron was trying to figure out how many subject headings might have been used for the same subject. In some sense, how the very largest libraries catalog is not necessarily useful for OSC, which presumes very open access to all the books.

Mar 12, 2009, 5:20am Top

People don't tag books with "Living Things" at least in part because that is implicit knowledge not explicit knowledge. If you tag the book "cats" you know that it is about living things. In the same way I don't tag my SF or horror or detective books "fiction".

Also using tags to determine the boundaries is not totally cut and dried.

Animals (as a tag) is used over 86,500 times. Pets only 10,380. Obviously both will also contain fiction.
Tag mashing shows "non-fiction, pets" at 250 books, and "animals, non-fiction" at 248 books with a reasonable overlap.

Mar 12, 2009, 9:05am Top

10: The only Dewey change that I remember from the time I was a kid was the elimination of 656 for railroads (as a business - leaving 385 and 625).

Dewey is revised continuously. The last big change was about 2 years ago when religion and the early 300s got a very thorough overhaul. The vast majority of libraries don't reclass. Small public libraries assume that old books will age out of the system and large libraries can't afford it. It not only requires cataloging staff to put the new numbers on, but the books have to be relabeled, not to mention they're not on the shelf while you do the work. (I also suspect that very few libraries of any size would move to OSC for these reasons as well.)

However, a lot of people do what might be called a browse/look, where they want they want to see what's around in a general area.

One of those 19th century ideas that Dewey was based on is that browsing could lead to discovery. A patron browses Insects. They don't find spiders. They go back to the beginning of insects where the general books on insects are. They open and book and find out that insects only have six legs. Hmm, where are spiders? They backup some more (in Dewey they remove the last number from the general Insect book and move "up" the class). They find a more general book on Invertebrates (making this up, don't have the schedule in front of me). They open it. Ahh, spiders are Invertebrates called Arachnids. They now browse along the shelf (In Dewey they can just look at all the books that have additions to the Invertebrate number) until they find either books with spider or books with arachnids in the title.

I can see now that all my work here is pretty useless, since I've been working on models like the one above, more trying to fix Dewey than get with the program. BISAC just bunches books together that might be related somehow, but doesn't say much about how (which now explains why I spend so much more time browsing in bookstores than I do in libraries, where I can generally get books in 15 minutes or so, even on topics I don't know well.)

In parting, I'll say that I think that there are two solutions to the Insect/Spider problem in the BISAC model. One is that the person doesn't find any books on spiders and goes and asks the clerk. The second is that the bookstore, aware that spiders and insects are confused, makes a more general category Bugs.

Mar 12, 2009, 3:26pm Top

One of the limitations of Library of Congress is that their systems (classification and subject headings) are based on their own collection. If it is not in their collection, they are not interested. My institution (not my library) has published a couple of books about gift cards, quantity thereof, how they are used, etc. There is no appropriate SH for gift cards in LCSH although they are pretty big in business right now (particularly at Christmas). Among other things, they represent a lot of money essentially given to businesses because the gift cards get lost or expire or whatever before they are used. I contacted LofC to see if they would establish a SH for the gift cards (given that there was literary mandate to do so) and I was informed that this would not happen until the Library had collected works on the topic -- regardless of its relevance. Rather arrogant and not terriblu useful.

Mar 12, 2009, 3:33pm Top

This message has been deleted by its author.

Mar 20, 2009, 10:44am Top

Please check out this thread (http://www.librarything.com/topic/60594) for a link to the new OSC blog and a call for specific volunteer involvement. Thanks!


This topic is not marked as primarily about any work, author or other topic.


No touchstones

You are using the new servers! | About | Privacy/Terms | Help/FAQs | Blog | Store | APIs | TinyCat | Legacy Libraries | Early Reviewers | Common Knowledge | 114,440,969 books! | Top bar: Always visible