Think “Different”
by Karen Coyle
Keynote, Dublin Core, 2012 and Emtacl12
data from Amazon, but they also obtained library bib-
liographic data, from the Library of Congress and from
other libraries.
“Think Different” was an Apple company advertis- None of the Archive staff members had any prior
ing campaign, which may be familiar to you. It caused experience with library data and at first they thought:
a bit of a scandal in the U.S. because “think different” no problem. Then they started looking at the data and
is grammatically incorrect – it should be “think dif- decided: “ok, problem.” So I worked with them as a kind
ferently,” with differently being the proper adverbial of translator between their goals and the data in library
form. Around the time of the death of Steve Jobs, a man catalog records.
who really did think different, I read that 1) he knew
his grammar and 2) his intention was to use different There were some significant differences in the goals
as a noun. So the message was: what ever you would between the Open Library and most library catalogs, a
usually think, you would normally think, instead think big one being that the Open Library would be wiki-like
of something different – a different way of looking at and anyone, really anyone, could edit the data, just like
things, a different style, a different assumption about they can in Wikipedia. This meant that some of the par-
how things work. ticular details of library data had to be smoothed out so
that the input and editing could use simple “fill in the
This is essentially the same concept as the saying box” forms. Some aspects of library data just wouldn’t
“think outside the box,” except that latter has become an fit into this model.
overused cliché, and I must say that it is often not clear
to me where the box is that is being thought outside of.
I’m going to explain what I mean with some exam-
ples, because the concept is hard to define in words.
Case 1: The Open Library and
Alphabetical Order
The Internet Archive has a bibliographic catalog called
the Open Library. It is a large database with about 25
million bibliographic records, and it is also the access
to the over one million digitized books that the Archive
holds and provides access to. When they were begin-
ning to create the catalog they realized that they needed
some sources of bibliographic data. They took in some
1
This seemed to be working just fine, and then we ran When I told this story to a large auditorium of U.S. li-
into the “alphabetical order” problem. brarians, they howled with laughter. What do you mean
“Why do we need alphabetical order?” Of course you
The catcher in the rye need alphabetical order. Alphabetical order is the very
basis of library cataloging; it permeates our cataloging
Many titles have extra words at the beginning that rules, even the new rules that presumably are being de-
get in the way of an expected alphabetical order. Most signed, at least in part, for the Semantic Web.
users would look for this work under “C” not “T,” and
the number of entries under “The” in a large library Well, Google, the most popular place to search for
catalog would be enormous. So library data has various information on the entire planet, doesn’t present results
was to indicate what part of the title is to be ignored for in alphabetical order. Nor does Amazon. Even OCLC’s
the purposes of putting titles in alphabetical order. Worldcat, the world’s largest database of library biblio-
graphic data, does not present in alphabetical order as
There are some systems that use marks to indicate
the non-filing part:
/The /catcher in the rye
Others, like MARC, make the cataloger indicate the
number of bytes that must be ignored when sorting:
[4] The catcher in the rye
The rules for what entails an initial article for filing
purposes differs from language to language, so explain-
ing to users how to input this data was going to be very
complex, and in an open database with potentially any
internet user being an editor, this just was not going to
work.
So the Open Library folks did something absolutely WorldCat results for “Barack Obama”
different – they asked: why do we need alphabetical or-
der?
Google results for “public library” Amazon results for “Barack Obama”
2
its default. And even that absolutely classic alphabeti- It was the discovery mechanism because no other
cal list, the telephone book, is now accessed as a search, technology of discovery was available in the analog
and in some cases does not return its results in alpha- card catalog. For the last 50 years we have had database
betical order. management systems that can reach into any part of the
library data record and find any text string or combi-
nation of strings within the record. We no longer use,
and should of course no longer design our data for, the
linear retrieval of the past.
What is the nature of alphabetical order in terms of
knowledge organization? Here is a set of things. What is
the meaning of this set of objects?
A “white pages” search online
And yet, when librarians create bibliographic data
they are following rules that not only assume but actu-
ally dictate that the key decisions that are made about
the metadata must be done in support of alphabetical
order. They don’t have any obvious conceptual or semantic
relationship between them. They are, however, in alpha-
It’s true that alphabetical order was once the only dis-
betical order.
covery mechanism in the library catalog.
3
Alphabetical order is not about knowledge, it’s about
words. It’s not conceptual, it’s not semantic. It is an ac-
cident of language that this first object, an apple, was
given the word “apple” that precedes the word of the
second object, “book.” There is no meaning that would
place apples before books.
Alphabetical order is an accident of language, and
different languages have different accidents. This makes
alphabetical order a very poor discovery method in a
multilingual environment.
However, I can show you various groupings of things
that, even if you haven’t seen them before, probably
make sense to you.
This is not a quirk; cognitive scientists can explain,
at least to some extent, why our brains work well with
concepts. And yet, in spite of this evidence, and in spite
of what we know about “modern technology” (that is,
technology that has been around for 50 years) the most
recently developed library cataloging rules still have
within them in key areas the assumption of the pre-
dominance language terms in alphabetical order. This
is a failure to think clearly about a problem that needs
solving because you have a default solution that cannot
be questioned, a solution with a long history that has
been successful in the past. This is an inability to “think
different.”
4
It goes from being a mass-produced object with
all instances being the same to a display of continu-
Case 2: The Book ous, fluid text with no fixity of the meaning of “page.”
If the change the font, the amount of text that is display
on the virtual page changes. If your device is a differ-
The book is such an icon in our culture: it represents
ent size or shape from someone else’s, what is on your
learning, it represents libraries, it represents religious
page is different from what is on theirs. Page numbers
beliefs. We all know what the book is: it has pages, usu-
are meaningless in this environment. Some education-
ally paper with printing on it, it has a title page, chapters
al institutions did experiments with ebooks for their
with headings, a binding that holds it all together, and
classes, but they ended up rejecting them because it
page numbers.
was too hard, in the classroom, to get everyone on the
“same page.” Where once a professor could instruct
his class to “turn to page 87,” there is no equivalent in
the ebook. Using percentages doesn’t work because the
point measured as “17%” of a large book could cover a
great deal of text.
Unsolvable problem? Hardly. In fact, this problem
was solved before the advent of printing.
For over 500 years we’ve had this fairly standardized
object that we know so very well, and then, suddenly, it
changes: the e-book is invented.
When books were in manuscript form there was
no concept of page numbers because each copy was
unique. Page numbers came into use only with the mass
production of books through printing. As a book that
decidedly preceded printing, the Bible has developed a
highly effective way of numbering chapters and verses
such than anyone with a different copy can refer to ex-
actly the same portion of text.
5
To give a more modern example, Ludwig Wittgen- In a printed book these numbers have to be visible
stein, admittedly not your average author, wrote his and they are admittedly unattractive and distracting. In
philosophical works with numbers on each paragraph. ebooks these place markers can be hidden until needed,
This means that persons reading his work, even in dif- as appears to be the Kindle approach. The ebook soft-
ferent translations, can reference the same precise point ware could make it easy to cite individual points in the
in his books. book, and perhaps even ranges.
I’ll give one more example, which is a bit odd but
definitely a case of thinking different.
There is a group of people who have taken upon
themselves to translate some classical works that are in
the public domain to series of QR code barcodes. They
call their project “Books to Barcodes.” Each QR code
represents some amount of text, usually a few sentences
or maybe a short paragraph. This is obviously not an
ideal way to read a book like Pride and Prejudice, but
it does demonstrate that snippets of text can be turned
into a machine-readable format that can be decoded by
a variety of modern devices. This doesn’t in itself give
you location between texts on different devices or in
different formats, but this odd project may provide a
clue to how we can solve this problem in a cross-plat-
form way.
Wittgenstein’s paragraph numbering
Books as QR codes
6
Thinking “Different” About Libraries
The classical library is a thing of great beauty. The wood- In fact, libraries are very “thing” oriented, and the
en shelves, leather-bound books and the cathedral-like library view of its collection is that of things that are
atmosphere of contemplation, the distant ceiling that owned and must be managed and controlled. Library
gives one a sense of reaching upward, transcending – cataloging is all about the physical format, and plac-
who of us who love books and learning wouldn’t want ing those things in a linear order. Retrieval is primarily
to study here? language-based, requiring that the user be able to name
This model is still at the heart of most modern li- what she is looking for.
braries. It may be the case that the library as conceived This concept of a controlled information world is out
even many thousands of years ago is so right that it is of date, as is the concept that information comes in a
still relevant. However, the technological changes over tangible form, pre-packaged. While print dominated
the last century, and in particular those that have led for a significant amount of time (from about 1450 to the
to a vastly larger variety of information media, must late 1800’s), our information has been in less tangible
certainly mean that the “books on a shelf ” model is no formats since the invention of the telegraph in 1844.
longer the most relevant one. Telephone, radio, television, and now the Internet have
Today’s media have an emphasis on interactivity that taken over the book’s previous monopoly on informa-
was inconceivable in the past. Books have been called a tion. In fact, something that we couldn’t even imagine
“slow conversation” between authors and readers, and just a few years ago is how the telephone has combined
some readers go on to extend that conversation in their with the Internet to become a multi-media information
own publications. In the traditional book world it is not and conversation machine.
easy to see how the books interact. There are some overt Yet in libraries we are still putting considerable ef-
connections, such as footnotes, but that is only part of fort into the linear arrangement of books on a shelf. It
the story. Technology today can help us see other con- is a myth that the library’s shelves are meaningful to the
nections, such as when documents share readers. user. We tell users that when they go to the shelf look-
We can also learn about interaction between docu- ing for one book they will find others on the same topic.
ments through their proximity in bibliographies, syl- That may happen sometimes, but the limitation of hav-
labi, and perhaps even the shelves they share. An aspect ing only one shelf location means that some books are
of the conversation is knowing what books an author or also kept far from each other. And what is the user to
reader had available in her time, which gives us a con- make of those numbers on the spines of the books? At
text for that person’s part of the conversation. And we no point does the library show users what those num-
know that books are active when they are read, and that bers mean. And they are meaningful – in fact they have
the reader is as important as the book itself. a whole knowledge structure in them, which is seen by
the one librarian who assigns the number, and no one
But libraries insist on treating books as objects; ob-
else. It’s a secret code. How can that help people find
jects to be organized, not as knowledge.
things?
7
We have to move libraries from organizing things However, the length of the page is not a good mea-
to knowledge discovery. This means that the library sure of how useful the catalog is. One problem that I see
can be only one part of the information picture since is that all of these still focus on a single item; they are
a great deal of information is outside of the library. It still very “thing” based. For Amazon that is a function
means emphasizing the relationships between things, of its purpose, which is to sell things. The library could
not their place in a linear order. It also means allowing take a different view. Some library catalogs are adding
interoperability and that means treating library users topic maps and various facets, but the focus is still on
as contributors to the knowledge universe, not just as the individual things, not relationships between them.
consumers. It means, yes, giving up control, and that is And in fact it is easier to organize things than it is to
going to be the hardest thing for librarians to do. manage a complex of relationships, but that’s not our
Libraries have been slow to augment their catalogs mission. In fact, I have a new mission statement for li-
with more information beyond the bare catalog record. braries:
WorldCat is now providing more links to resources
outside the catalog itself. Amazon still provides much The mission of the library
more information, and more interaction, than the li- is not to gather physical
brary catalog which still mostly does not let the user things into an inventory,
contribute at all.
but to organize human
knowledge that has been
very inconveniently pack-
aged.
To do this, we must un-package our data so that any
card
information can combine with any other information,
and let the user of the data determine what is the focus,
and what relationships are meaningful.
It should be possible to ask a question of the library
catalog, and to get an answer, not just a list of items.
“What were the most popular book sub-
jects from 1830-1840?”
(WorldCat answer, using kw:history because a
search on date alone isn’t allowed: 141,076)
Worldcat
Presenting results as a list of items no longer works
because library holdings have gone beyond human
scale. A search on the term “history” and limited to
books published from 1830-1840 retrieves over 140,000
records. Even with a few hundred it would be hard for
a human to look at these records and determine what
the most popular subjects were in that time period. We
need to apply computation to perform tasks that hu-
Amazon
mans cannot do on their own. That’s what computers
are good at.
8
An example of computation on library data is WorldCat identi-
ties, which shows timelines that give the user a quick snapshot of
the author.
Another example is on the subject pages in OpenLibrary, where
you can see that the term “human evolution” appears first in 1859,
with Darwin, but gains ground only after about 1960. The topic
“love” however has been written about at least since the beginning of
printing, and undoubtedly even before. This is useful information in
itself, and this is just using the library metadata.
Timeline for subject “human evolution”
Timeline for subject “love”
9
However, library metadata itself has some serious
problems as data, as exhibited by a record for “The ori-
gin of species” by Darwin, that has the publication date
2009. Nothing in the data tells us that the text is that of
1859, because the emphasis is on the package, not on
the content. And this means that the library user misses
some key context, which is how the slow conversation
of books has taken place. Darwin makes little sense if
you think his ideas are from 2009.
In fact, this view greatly interrupts the conversation
about science and evolution.
New Dimensions
We need to add some new dimensions to the library. The fifth dimension is people. What is in the library
The library today is 2-D, relying heavily on linear order was created by people, and people will use the services
-- linear order of shelves, linear order of headings in and materials of the library in unpredictable ways. Peo-
the catalog, and linear order of catalog records that are ple will understand and create new knowledge using the
retrieved with searches. library. That knowledge may be totally new to the world
The first dimension that we need to add is linking; or just new for that user. It will combine thoughts from
linking between items in the library based on any as- the person’s life and information previously encoun-
pect of the concepts they contain; linking from the li- tered.
brary to information outside of the library; and linking
from the main information resource in the users’ envi-
ronment, the Web, to the library. With links, the library
can become 3-D.
The fourth dimension is time. It needs to be possible
to follow thoughts and ideas in the library as the devel-
op over time. This includes knowing what works influ-
enced an author or inventor and what new discoveries
that person may have contributed to. Readers should
be able to reconstruct the context of information that
they encounter, so that they can better understand the
world that knowledge addressed.
10
Addendum: Is Linked Data the Answer?
I have beena proponent of exploration into the use of ticular one called “[Link]” to enhance the informa-
linked data for libraries for a number of years. Therefore tion that can be presented to users when they search the
you might expect that I would say: Yes, linked data is Web. Google refers to this as “rich snippets” and shows
the answer! examples that lead users directly to resources within or
Instead, I want to insert here a word of caution. No, related to a Web page:
I’m not going to declare that we should abandon ideas
of making use of linked data, but I do want to caution
against having “an answer” to the issues and problems
that we face. When you have “an answer” you tend to
stop looking for new ideas and new solutions. You also
tend to only consider problems that the “answer” can
address.
It is unreasonable, and even dangeous, to think that
any one technology will be the solution to every pro-
cess and service that we want to provide. So although
there is much to be gained by using linked data to cre- In a similar way, some searches might return a di-
ate a web that connects libraries to other information rect link to resources in the user’s library community:
resources, we have more to do than simply linking.
One strong assumption in the library field is that
what we have to contribute to the Web and the greater
information world is the contents of our bibliograph-
ic databases. Yet there would be little to be gained by
flooding the Web with hundreds of millions of records,
most of which already exist. In fact, the Web is awash in
bibliographic data, from booksellers like Amazon and
Barnes and Noble, to Google Books, which is a mix of
actual books and bibliographic data, to book fan sites Some of this can be facilitated with linked data, but
like LibraryThing and GoodReads. Although some of linked data in itself will not make it possible to effec-
the data in library catalogs may be rare or unique, most tively and efficiently provide these local holdings. To
of what we would contribute would be duplication of provide actual library services through general Web
data that already exists. software we will need to “think different” — even differ-
What libraries do have, however, that no one else ent from linked data.
does is that we know where the user can borrow or use
materials in her nearby community. It is library hold-
ings that is key to providing service and furthering
knowledge creation. It is also key to providing visibility
for libraries as users explore resources on the Web.
Connecting users to library resources from the Web
is an extention of what we already provide using the
OpenURL: offering the user access to copies of mate-
rials from within non-library contexts. There may be
more than one way to provide this service. The search
engines are exploring the use of microformats, in par-
11