CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 55
Martha M. Yee
Can Bibliographic Data be Put Directly
onto the Semantic Web?
This paper is a think piece about the possible future of bib-
liographic control; it provides a brief introduction to the
Semantic Web and defines related terms, and it discusses
granularity and structure issues and the lack of standards
for the efficient display and indexing of bibliographic
data. It is also a report on a work in progress—an experi-
ment in building a Resource Description Framework
(RDF) model of more FRBRized cataloging rules than
those about to be introduced to the library community
(Resource Description and Access) and in creating an
RDF data model for the rules. I am now in the process
of trying to model my cataloging rules in the form of an
RDF model, which can also be inspected at http://myee.
bol.ucla.edu/. In the process of doing this, I have discov-
ered a number of areas in which I am not sure that RDF
is sophisticated enough yet to deal with our data. This
article is an attempt to identify some of those areas and
explore whether or not the problems I have encountered
are soluble—in other words, whether or not our data
might be able to live on the Semantic Web. In this paper, I
am focusing on raising the questions about the suitability
of RDF to our data that have come up in the course of
my work.
T
his paper is a think piece about the possible future
of bibliographic control; as such, it raises more
complex questions than it answers. It is also a
report on a work in progress—an experiment in build-
ing a Resource Description Framework (RDF) model of
FRBRized descriptive and subject-cataloging rules. Here
my focus will be on the data model rather than on the
FRBRized cataloging rules for gathering data to put in
the model, although I hope to have more to say about the
latter in the future. The intent is not to present you with
conclusions but to present some questions about data
modeling that have arisen in the course of the experiment.
My premise is that decisions about the data model we
follow in the future should be made openly and as a com-
munity rather than in a small, closed group of insiders.
If we are to move toward the creation of metadata that is
more interoperable with metadata being created outside
our community, as is called for by many in our profes-
sion, we will need to address these complex questions as
a community following a period of deep thinking, clever
experimentation, and astute political strategizing.
n The vision
The Semantic Web is still a bewitching midsummer
night’s dream. It is the idea that we might be able to
replace the existing HTML–based Web consisting of
marked-up documents—or pages—with a new RDF–
based Web consisting of data encoded as classes, class
properties, and class relationships (semantic linkages),
allowing the Web to become a huge shared database.
Some call this Web 3.0, with hyperdata replacing hyper-
text. Embracing the Semantic Web might allow us to
do a better job of integrating our content and services
with the wider Internet, thereby satisfying the desire for
greater data interoperability that seems to be widespread
in our field. It also might free our data from the propri-
etary prisons in which it is currently held and allow
us to cooperate in developing open-source software to
index and display the data in much better ways than we
have managed to achieve so far in vendor-developed ILS
OPACs or in giant, bureaucratic bibliographic empires
such as OCLC WorldCat.
The Semantic Web also holds the promise of allow-
ing us to make our work more efficient. In this bewitch-
ing vision, we would share in the creation of Uniform
Resource Identifiers (URIs) for works, expressions, mani-
festations, persons, corporate bodies, places, subjects, and
so on. At the URI would be found all of the data about
that entity, including the preferred name and the vari-
ant names, but also including much more data about the
entity than we currently put into our work (name-title and
title), such as personal name, corporate name, geographic, and
subject authority records. If any of that data needed to be
changed, it would be changed only once, and the change
would be immediately accessible to all users, libraries,
and library staff by means of links down to local data such
as circulation, acquisitions, and binding data. Each work
would need to be described only once at one URI, each
expression would need to be described only once at one
URI, and so forth.
Very much up in the air is the question of what institu-
tional structures would support the sharing of the creation
of URIs for entities on the Semantic Web. For the data to
be reliable, we would need to have a way to ensure that
the system would be under the control of people who
had been educated about the value of clean and accurate
entity definition, the value of choosing “most commonly
known” preferred forms (for display in lists of mul-
tiple different entities), and the value of providing access
Martha M. Yee (myee@ucla.edu) is cataloging Supervisor at
the University of california, los Angeles Film and Television
Archive.
56 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
under all variant forms likely to be sought. At the same
time, we would need a mechanism to ensure that any
interested members of the public could contribute to the
effort of gathering variants or correcting entity definitions
when we have had inadequate information. For example,
it would be very valuable to have the input of a textual
or descriptive bibliographer applied to difficult questions
concerning particular editions, issues, and states of a sig-
nificant literary work. It would also be very valuable to be
able to solicit input from a subject expert in determining
the bounds of a concept entity (subject heading) or class
entity (classification).
n The experiment (my project)
To explore these bewitching ideas, I have been conduct-
ing an experiment. As part of my experiment, I designed
a set of cataloging rules that are more FRBRized than
is RDA in the sense that they more clearly differentiate
between data applying to expression and data apply-
ing to manifestation. Note that there is an underlying
assumption in both FRBR (which defines expression
quite differently from manifestation) and on my part,
namely that catalogers always know whether a given
piece of data applies at either the expression or the man-
ifestation level. That assumption is open to questioning
in the process of the experiment as well. My rules also
call for creating a more hierarchical and degressive
relationship between the FRBR entities work, expression,
manifestation, and item, such that data pertaining to the
work does not need to be repeated for every expres-
sion, data pertaining to the expression does not need
to be repeated for every manifestation, and so forth.
Degressive is an old term used by bibliographers for bib-
liographies that provide great detail about first editions
and less detail for editions after the first. I have adapted
this term to characterize my rules, according to which
the cataloger begins by describing the work; any details
that pertain to all expressions and manifestations of the
work are not repeated in the expression and manifesta-
tion descriptions. This paper would be entirely too long
if I spent any more time describing the rules I am devel-
oping, which can be inspected at http://myee.bol.ucla
.edu. Here, I would like to focus on the data-modeling
process and the questions about the suitability of RDF
and the Semantic Web for encoding our data. (By the
way, I don’t seriously expect anyone to adopt my rules!
They are radically different than the rules currently
being applied and would represent a revolution in cata-
loging practice that we may not be up to undertaking in
the current economic climate. Their value lies in their
thought-experiment aspect and their ability to clarify
what entities we can model and what entities we may
not be able to model.)
I am now in the process of trying to model
my cataloging rules in the form of an RDF model
(“RDF” as used in this paper should be considered
from now on to encompass RDF Schema [RDFS], Web
Ontology Language [OWL], and Simple Knowledge
Organization System [SKOS] unless otherwise stated);
this model can also be inspected at http://myee.bol
.ucla.edu. In the process of doing this, I have discovered
a number of areas in which I am not sure that RDF is yet
sophisticated enough to deal with our data. This article
is an attempt to outline some of those areas and explore
whether the problems I have encountered are soluble, in
other words, whether or not our data might be able to
live on the Semantic Web eventually. I have already heard
from RDF experts Bruce D’Arcus (Miami University) and
Rob Styles (developer of Talis, as Semantic Web technol-
ogy company), whom I cite later, but through this article
I hope to reach a larger community.
My research questions can be found later, but first
some definitions.
n Definition of terms
The Semantic Web is a way to represent knowledge;
it is a knowledge-representation language that provides
ways of expressing meaning that are amenable to com-
putation; it is also a means of constructing knowledge-
domain maps consisting of class and property axioms
with a formal semantics
RDF is a family of specifications for methods of
modeling information that underpins the Semantic Web
through a variety of syntax formats; an RDF metadata
model is based on making statements about resources in
the form of triples that consist of
1. the subject of the triple (e.g., “New York”);
2. the predicate of the triple that links the subject and
the object (e.g., “has the postal abbreviation”); and
3. the object of the triple (e.g., “NY”).
XML is commonly used to express RDF, but it is not
a necessity; it can also be expressed in Notation 3 or N3,
for example.1
RDFS is an extensible knowledge-representation lan-
guage that provides basic elements for the description of
ontologies, also known as RDF vocabularies. Using RDFS,
statements are made about resources in the form of
1. a class (or entity) as subject of the RDF triple (e.g.,
“New York”);
2. a relationship (or semantic linkage) as predicate of the
RDF triple that links the subject and the object (e.g.,
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 57
“has the postal abbreviation”); and
3. a property (or attribute) as object of the RDF triple
(e.g., “NY”).
OWL is a family of knowledge representation lan-
guages for authoring ontologies compatible with RDF.
SKOS is a family of formal languages built upon RDF
and designed for representation of thesauri, classification
schemes, taxonomies, or subject-heading systems.
n Research questions
Actually, the full-blown Semantic Web may not be exactly
what we need. Remember that the fundamental definition
of the Semantic Web is “a way to represent knowledge.”
The Semantic Web is a direct descendant of the attempt
to create artificial intelligence, that is, of the attempt to
encode enough knowledge of the real world to allow a
computer to reason about reality in a way indistinguish-
able from the way a human being reasons. One of the
research questions should probably be whether or not the
technology developed to support the Semantic Web can
be used to represent information rather than knowledge.
Fortunately, we do not need to represent all of human
knowledge—we simply need to describe and index
resources to facilitate their retrieval. We need to encode
facts about the resources and what the resources discuss
(what they are “about”), not facts about “reality.” Based
on our past experience, doing even this is not as simple
as people think it is. The question is whether we could do
what we need to do within the context of the Semantic
Web. Sometimes things that sound simple do not turn out
to be so simple in the doing.
My research questions are as follows:
1. Is it possible for catalogers to tell in all cases
whether a piece of data pertains to the FRBR
expression or the FRBR manifestation?
2. Is it possible to fit our data into RDF? Given that
RDF was designed to encode knowledge rather
than information, perhaps it is the wrong technol-
ogy to use for our purposes?
3. If it is possible to fit our data into RDF, is it possible
to use that data to design indexes and displays that
meet the objectives of the catalog (i.e., providing
an efficient instrument to allow a user to find a
particular work of which the author and title are
known, a particular expression of a work, all of
the works of an author, all of the works in a given
genre or form, or all of the works on a particular
subject)?
As stated previously, I am not yet ready to answer
these questions. I hope to find answers in the course of
developing the rules and the model. In this paper, I am
focusing on raising the questions about the suitability of
RDF to our data that have come up in the course of my
work.
n Other relevant projects
Other relevant projects include the following:
1. FRBR, Functional Requirements for Authority
Data (FRAD), Funtional Requirements for Subject
Authority Records (FRSAR), and FRBR-object-
oriented (FRBRoo). All are attempts to create con-
ceptual models of bibliographic entities using an
entity-relationship model that is very similar to the
class-property model used by RDF.2
2. Various initiatives at the Library of Congress (LC),
such as LC Subject Headings (LCSH) in SKOS,3
the LC Name Authority File in SKOS,4 the LCCN
Permalink project to create persistent URIs for
bibliographic records,5 and initiatives to provide
SKOS representations for vocabularies and data
elements used in MARC, PREMIS, and METS.
These all represent attempts to convert our exist-
ing bibliographic data into URIs that stand for the
bibliographic entities represented by bibliographic
records and authority records; the URIs would
then be available for experiments in putting our
data directly onto the Semantic Web.
3. The DC-RDA Task Group project to put RDA data
elements into RDF.6 As noted previously and dis-
cussed further later, RDA is less FRBRized than my
cataloging rules, but otherwise this project is very
similar to mine.
4. Dublin Core’s (DC’s) work on an RDF schema.7 Dublin
Core is very focused on manifestation and does
not deal with expressions and works, so it is less
similar to my project than is the DC-RDA Task
Groups’s project (see further discussion later).
n Why my project?
One might legitimately ask why there is a need for a dif-
ferent model than the ones already provided by FRBR,
FRAD, FRSAR, FRBRoo, RDA, and DC. The FRBR and
RDA models are still tied to the model that is implicit
in our current bibliographic data in which expression
and manifestation are undifferentiated. This is because
publishers publish and libraries acquire and shelve mani-
festations. In our current bibliographic practice, a new
58 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
bibliographic record is made for either a new manifesta-
tion or a new expression. Thus, in effect, there is no way
for a computer to tell one from the other in our current
data. Despite the fact that FRBR has good definitions of
expression (change in content) and manifestation (mere
change in carrier), it perpetuates the existing implicit
model in its mapping of attributes to entities. For exam-
ple, FRBR maps the following to manifestation: edition
statements (“2nd rev. ed.”); statements of responsibility
that identify translators, editors, and illustrators; physi-
cal description statements that identify illustrated edi-
tions; and extent statements that differentiate expressions
(the 102-minute version vs. the 89-minute version); etc.
Thus the FRBR definition of expression recognizes that a
2nd revised edition is a new expression, but FRBR maps
the edition statement to manifestation. In my model, I
have tried to differentiate more cleanly data applying to
expressions from data applying to manifestations.8
FRBR and RDA tend to assume that our current bib-
liographic data elements map to one and only one group
1 entity or class. There are exceptions, such as title, which
FRBR and RDA define at work, expression, and manifes-
tation levels. However, there is a lack of recognition that,
to create an accurate model of the bibliographic universe,
more data elements need to be applied at the work and
expression level in addition to (or even instead of) the
manifestation level. In the appendix I have tried to con-
trast the FRBR, FRAD, and RDA models with mine. In my
model, many more data elements (properties and attri-
butes) are linked to the work and expression level. After
all, if the expression entity is defined as any change in
work content, the work entity needs to be associated with
all content elements that might change, such as the original
extent of the work, the original statement of responsibil-
ity, whether illustrations were originally present, whether
color was originally present in a visual work, whether
sound was originally present in an audiovisual work, the
original aspect ratio of a moving image work, and so on.
FRBR also tends to assume that our current data ele-
ments map to one and only one entity. In working on
my model, I have come to the conclusion that this is not
necessarily true. In some cases, a data element pertaining
to a manifestation also pertains to the expression and the
work. In other cases, the same data element is specific to
that manifestation, and, in other cases, the same data ele-
ment is specific to its expression. This is true of most of
the elements of the bibliographic description.
FRAD, in attempting to deal with the fact that our
current cataloging rules allow a single person to have
several bibliographic identities (or pseudonyms), treats
person, name, and controlled access point as three separate
entities or classes. I have tried to keep my model simpler
and more elegant by treating only person as an entity, with
preferred name and variant name as attributes or properties
of that entity.
FRBRoo is focused on the creation process for works,
with special attention to the creation of unique works
of art and other one-off items found in museums. Thus
FRBRoo tends to neglect the collocation of the various
expressions that develop in the history of a work that is
reproduced and published, such as translations, abridged
editions, editions with commentary, etc.
DC has concentrated exclusively on the description
of manifestations and has neglected expression and work
altogether.
One of the tenets of Semantic Web development is
that, once an entity is defined by a community, other
communities can reuse that entity without defining it
themselves. The very different definitions of the work
and expression entities in the different communities
described above raise some serious questions about the
viability of this tenet.
n Assumptions
It should be noted that this entire experiment is based on
two assumptions about the future of human intervention
for information organization. These two assumptions are
based on the even bigger assumption that, even though
the Internet seems to be an economy based on free intel-
lectual labor, and, even though human intervention for
information organization is expensive (and therefore at
more risk than ever), human intervention for information
organization is worth the expense.
n Assumption 1: What we need is not artificial intel-
ligence, but a better human–machine partnership
such that humans can do all of the intellectual
labor and machines can do all of the repetitive
clerical labor. Currently, catalogers spend too
much time on the latter because of the poor design
of current systems for inputting data. The univer-
sal employment provided by paying humans to
do the intellectual labor of building the Semantic
Web might be just the stimulus our economy
needs.
n Assumption 2: Those who need structured and
granular data—and the precise retrieval that results
from it—to carry out research and scholarship may
constitute an elite minority rather than most of the
people of the world (sadly), but that talented and
intelligent minority is an important one for the cul-
tural and technological advancement of humanity.
It is even possible that, if we did a better job of
providing access to such data, we might enable the
enlargement of that minority.
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 59
n Granularity and structure issues
As soon as one starts to create a data model, one encoun-
ters granularity or cataloger-data parsing issues. These
issues have actually been with us all along as we devel-
oped the data model implicit in AACR2R and MARC 21.
Those familiar with RDA, FRBR, and FRAD development
will recognize that much of that development is directed
at increasing structure and granularity in cataloger-
produced data to prepare for moving it onto the Semantic
Web. However, there are clear trade-offs in an increase
in structure and granularity. More structure and more
granularity make possible more powerful indexing and
more sophisticated display, but more structure and more
granularity are more complex and expensive to apply and
less likely to be implemented in a standard fashion across
all communities; that is, it is less likely that interoperable
data would be produced. Any switching or mapping
that was employed to create interoperable data would
produce the lowest common denominator (the simplest
and least granular data), and once rendered interoper-
able, it would not be possible for that data to swim back
upstream to regain its lost granularity. Data with less
structure and less granularity could be easier and cheaper
to apply and might have the potential to be adopted in a
more standard fashion across all communities, but that
data would limit the degree to which powerful indexing
and sophisticated display would be possible.
Take the example of a personal name: Currently, we
demarcate surname from forename by putting the sur-
name first, followed by a comma and then the forename.
Even that amount of granularity can sometimes pose a
problem for a cataloger who does not necessarily know
which part of the name is surname and which part is
forename in a culture unfamiliar to the cataloger. In other
words, the more granularity you desire in your data,
the more often the people collecting the data are going
to encounter ambiguous situations. Another example:
Currently, we do not collect information about gender
self-identification; if we were to increase the granularity
of our data to gather that information, we would surely
encounter situations in which the cataloger would not
necessarily know if a given creator was self-defined as a
female or a male or of some other gender identity.
Presently, if we are adding a birth and death date,
whatever dates we use are all together in a $d subfield
without any separate coding to indicate which date is the
birth date and which is the death date (although an occa-
sional “b.” or “d.” will tell us this kind of information).
We could certainly provide more granularity for dates,
but that would make the MARC 21 format much more
complex and difficult to learn. People who dislike the
MARC 21 format already argue that it is too granular and
therefore requires too much of a learning curve before
people can use it. For example, Tennant claims that “there
are only two kinds of people who believe themselves
able to read a MARC record without referring to a stack
of manuals: a handful of our top catalogers and those on
serious drugs.”9 How much of the granularity already
in MARC 21 is used either in existing records or, even
if present, is used in indexing and display software?
Granularity costs money, and libraries and archives are
already starving for resources. Granularity can only be
provided by people, and people are expensive.
Granularity and structure also exist in tension with
each other. More granularity can lead to less structure (or
more complexity to retain structure along with granular-
ity). In the pursuit of more granularity of data than we
have now, RDA, attempting to support RDF–compliant
XML encoding, has been atomizing data to make it useful
to computers, but this will not necessarily make the data
more useful to humans. To be useful to humans, it must
be possible to group and arrange (sort) the data meaning-
fully, both for indexing and for display. The developers
of SKOS refer to the “vast amounts of unstructured (i.e.,
human readable) information in the web,”10 yet labeling
bits of data as to type and recording semantic relation-
ships in a machine-actionable way do not necessarily
provide the kind of structure necessary to make data
readable by humans and therefore useful to the people
the Web is ultimately supposed to serve. Consider the
case of music instrumentation. If you have a piece of
music for five guitars and one flute, and you simply
code number and instrumentation without any way to
link “five” with “guitars” and “one” with “flute,” you
will not be able to guarantee that a person looking for
music for five flutes and one guitar will not be given this
piece of music in their results (see figure 1).11 The more
granular the data, the less the cataloger can build order,
sequencing, and linking into the data; the coding must be
carefully designed to allow the desired order, sequenc-
ing, and linking for indexing and display to be possible,
which might call for even more complex coding. It would
be easy to lose information about order, sequencing, and
linking inadvertently.
Actually, there are several different meanings for the
term structure:
1. Structure is an object of a record (structure of docu-
ment?); for example, Elings and Waibel refer to
“data fields . . . also referred to as elements . . .
which are organized into a record by a data struc-
ture.”12
2. Structure is the communications layer, as opposed to
the display layer or content designation.13
3. Structure is the record, field, and subfield.
4. Structure is the linking of bits of data together in the
60 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
form of various types of relationships.
5. Structure is the display of data in a structured, ordered,
and sequenced manner to facilitate human understand-
ing.
6. Data structure is a way of storing data in a computer
so that it can be used efficiently (this is how computer
programmers use the term).
I hasten to add that I am definitely in favor of add-
ing more structure and granularity to our data when it
is necessary to carry out the fundamental objectives of
our profession and of our catalogs. I argued earlier that
FRBR and RDA are not granular enough when it comes
to the distinction between data elements that apply to
expression and those that apply to manifestation. If we
could just agree on how to differentiate data applying to
the manifestation from data applying to the expression
instead of our current practice of identifying works with
headings and lumping all manifestation and expression
data together, we could increase the level of service we
are able to provide to users a thousandfold. However, if
we are not going to commit to differentiating between
Figure 1b. example of encoding of musical instrumentation at the expression level based on the above model
5
guitars
1
flute
instrumentation of musical expression
original instrumentation of musical expression—number of a particular instrument
original instrumentation of musical expression—type of instrument
Figure 1a. extract from Yee rdF model that illustrates one technique for modeling musical instrumentation at the expression level (using a
blank node to group repeated number and instrument type)
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 61
expression and manifestation, it would be more intellec-
tually honest for FRBR and RDA to take the less granular
path of mapping all existing bibliographic data to mani-
festation and expression undifferentiated, that is, to use
our current data model unchanged and state this openly.
I am not in favor of adding granularity for granularity’s
sake or for the sake of vague conceptions of possible
future use. Granularity is expensive and should be used
only in support of clear and fundamental objectives.
n The goal: efficient displays and indexes
My main concern is that we model and then structure the
data in a way that allows us to build the complex displays
that are necessary to make catalogs appear simple to use.
I am aware that the current orthodoxy is that recording
data should be kept completely separate from indexing
and display (“the applications layer”). Because I have
spent my career in a field in which catalog records are
indexed and displayed badly by systems people who
don’t seem to understand the data contained in them, I
am a skeptic. It is definitely possible to model and struc-
ture data in such a way that desired displays and indexes
are impossible to construct. I have seen it happen!
The LC Working Group report states that “it will
be recognized that human users and their needs for
display and discovery do not represent the only use of
bibliographic metadata; instead, to an increasing degree,
machine applications are their primary users.”14 My
fear is that the underlying assumption here is that users
need to (and can) retrieve the single perfect record. This
will never be true for bibliographic metadata. Users
will always need to assemble all relevant records (of all
kinds) as precisely as possible and then browse through
them before making a decision about which resources
to obtain. This is as true in the Semantic Web—where
“records” can be conceived of as entity or class URIs—as
it is in the world of MARC–encoded metadata.
Some of the problems that have arisen in the past in
trying to index bibliographic metadata for humans are
connected to the fact that existing systems do not group
all of the data related to a particular entity effectively,
such that a user can use any variant name or any combi-
nation of variant names for an entity and do a successful
search. Currently, you can only look for a match among
two or more keywords within the bounds of a single
manifestation-based bibliographic record or within the
bounds of a single heading, minus any variant terms for
that entity. Thus, when you do a keyword search for two
keywords, for example, “clemens” and “adventures,”
you will retrieve only those manifestations of Mark
Twain’s Adventures of Tom Sawyer that have his real name
(Clemens) and the title word “Adventures” co-occurring
within the bounded space created by a single manifes-
tation-based bibliographic record. Instead, the preferred
forms and the variant forms for a given entity need to be
bounded for indexing such that the keywords the user
employs to search for that entity can be matched using
co-occurrence rules that look for matches within a single
bounded space representing the entity desired. We will
return to this problem in the discussion of issue 3 in the
later section “RDF Problems Encountered.”
The most complex indexing problem has always
proven to be the grouping or bounding of data related
to a work, since it requires pulling in all variants for the
creator(s) of that work as well. Otherwise, a user who
searches for a work using a variant of the author’s name
and a variant of the title will continue to fail (as they do in
all current OPACs), even when the desired work exists in
the catalog. If we could create a URI for the Adventures of
Tom Sawyer that included all variant names for the author
and all variant titles for the work (including the variant
title Tom Sawyer), the same keyword search described
above (“clemens” and “adventures”) could be made
to retrieve all manifestations and expressions of the
Adventures of Tom Sawyer, instead of the few isolated
manifestations that it would retrieve in current catalogs.
We need to make sure that we design and structure
the data such that the following displays are possible:
n Display all works by this author in alphabeti-
cal order by title with the sorting element (title)
appearing at the top of each work displayed.
n Display all works on this subject in alphabetical
order by principal author and title (with principal
author and title appearing at top of each work dis-
played), or title if there is no principal author (with
title appearing at top of each work displayed).
We must ensure that we design and structure the
data in such a way that our structure allows us to create
subgroups of related data, such as instrumentation for a
piece of music (consisting of a number associated with
each particular instrument), place and related publisher for
a certain span of dates on a serial title change record, and
the like.
n Which standards will carry out which functions?
Currently, we have a number of different standards to
carry out a number of different functions; we can speculate
about how those functions might be allocated in a new
Semantic Web–based dispensation, as shown in table 1.
In table 1, data structure is taken to mean what a
record represents or stands for; traditionally, a record
has represented an expression (in the days of hand-
62 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
press books) or a manifestation (ever since reproduction
mechanisms have become more sophisticated, allowing
an explosion of reproductions of the same content in
different formats and coming from different distribu-
tors). RDA is record-neutral; RDF would allow URIs to
be established for any and all of the FRBR levels; that is,
there would be a URI for a particular work, a URI for a
particular expression, a URI for a particular manifesta-
tion, and a URI for a particular item. Note that I am not
using data structure in the sense that a computer pro-
grammer does (as a way of storing data in a computer
so that it can be used efficiently).
Currently, the encoding of facts about entity relation-
ships (see table 1) is carried out by matching data-value
character strings (headings or linking fields using ISSNs
and the like) that are defined by the LC/NACO author-
ity file (following AACR2R rules), LCSH (following rules
in the Subject Cataloging Manual), etc. In the future, this
function might be carried out by using RDF to link the
URI for a resource to the URI for a data value.
Display rules (see table 1) are currently defined by
ISBD and AACR2R but widely ignored by systems, which
frequently truncate bibliographic records arbitrarily in
displays, supply labels, and the like; RDA abdicates
responsibility, pushing display out of the cataloging
rules. The general principle on the Web is to divorce data
from display and allow anyone to display the data any
way they want. Display is the heart of the objects (or
goals) of cataloging: The point is to display to the user the
works of an author, the editions of a work, or the works
on a subject. All of these goals only can be met if complex,
high-quality displays can be built from the data created
according to the data model.
Indexing rules (see table 1) were once under the control
of catalogers (in book and card catalogs) in that users had
to navigate through headings and cross-references to find
Table 1. Possible reallocation of current functions in a new Semantic Web–based dispensation
Function Current Future?
Data content, or content guidelines (rules
for providing data in a particular element)
Defined by AACR2R and MARC
21
Defined by RDA and RDF/RDFS/
OWL/SKOS
Data elements Defined by ISBD–based AACR2R
and MARC 21
Defined by RDA and RDF/RDFS/
OWL/SKOS
Data values Defined by LC/NACO authority
file, LCSH, MARC 21 coded data
values, etc.
Defined as ontologies using RDF/
RDFS/OWL/SKOS
Encoding or labeling of data elements
for machine manipulation; same as data
format?
Defined by ISO 2709–based
MARC 21
Defined by RDF/RDFS/XML
Data structure (i.e., what a record
stands for)
Defined by AACR2R and MARC
21; also FRBR?
Defined by RDF/RDFS/OWL/
SKOS
Schematization (constraint on structure
and content)
MARC 21, MODS, DCMI abstract
model
Defined by RDF/RDFS/OWL/
SKOS
Encoding of facts about entity
relationships
Carried out by matching data
value strings (headings found
in LC/NACO authority file and
LCSH, ISSN’s, and the like)
Carried out by RDF/RDFS/OWL/
SKOS in the form of URI links
Display rules ILS software, formerly ISBD–
based AACR2R
(“Application layer”) or Yee rules
Indexing rules ILS software SPARQL, “application layer,” or
Yee rules
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 63
what they wanted; currently indexing is in the hands of
system designers who prefer to provide keyword index-
ing of bibliographic (i.e., manifestation-based) records
rather than provide users with access to the entities they
are really interested in (works, authors and subjects),
all represented currently by authority records for head-
ings and cross-references. RDA abdicates responsibility,
pushing indexing concerns completely out of the catalog-
ing rules. The general principle on the Web is to allow
resources to be indexed by any Web search engines that
wish to index them. Current Web data is not structured at
all for either indexing or display.
I would argue that our interest in the Semantic Web
should be focused on whether or not it will support
more data structure—as well as more logic in that data
structure—to support better indexes and better displays
than we have now in manifestation-based ILS OPACs.
Crucial to better indexing than we have ever had before
are the co-occurrence rules for keyword indexing, that
is, the rules for when a co-occurrence of two or more
keywords should produce a match. We need to be able
to do a keyword search across all possible variant names
for the entity of interest, and the entity of interest for
the average catalog user is much more likely to be a
particular work than to be a particular manifestation.
Unfortunately, catalog-use studies only have studied
so-called known-item searches without investigating
whether a known-item searcher was looking for a par-
ticular edition or manifestation of a work or was simply
looking for a particular work in order to make a choice
as to edition or manifestation once the work was found.
However, common sense tells us that it is a rare user
who approaches the catalog with prior knowledge about
all published editions of a given work. The more com-
mon situation is surely one in which a user desires to
read a particular Shakespeare play or view a particular
David Lean film and discovers that the desired work
exists in more than one expression or manifestation
only after searching the catalog. We need to have the
keyword(s) in our search for a particular work co-occur
within a bounded space that encompasses all possible
keywords that might refer to that particular work entity,
including both creator and title keywords.
Notice in table 1 the unifying effect that RDF could
potentially have; it could free us from the use of multiple
standards that can easily contradict each other, or at least
not live peacefully together. Examples are not hard to
find in the current environment. One that has cropped
up in the course of RDA development concerns family
names. Presently the rules for naming families are dif-
ferent depending on whether the family is the subject of
a work (and established according to LCSH) or whether
the family is responsible for a collection of papers (and
established according to RDA).
n Types of data
RDA has blurred the distinctions among certain types
of data, apparently because there is a perception that
on the Semantic Web the same piece of data needs to be
coded only once, and all indexing and display needs can
be supported from that one piece of data. I question that
assumption on the basis of my experience with biblio-
graphic cataloging. All of the following ways of encod-
ing the same piece of data can still have value in certain
circumstances:
n Transcribed; in RDF terms, a literal (i.e., any data
that is not a URI, a constant value). Transcribed
data is data copied from an item being cataloged.
It is valuable for providing access to the form of the
name used on a title page and is particularly useful
for people who use pseudonyms, corporate bodies
that change name, and so on. Transcribed data is an
important part of the historical record and not just
for off-line materials; it can be a historical record of
changing data on notoriously fluid webpages.
n Composed; in RDF terms, also a literal. Composed
data is information composed by a cataloger on
the basis of observation of the item in hand; it can
be valuable for historical purposes to know which
data was composed.
n Supplied; in RDF terms, also a literal. Supplied data
is information supplied by a cataloger from outside
sources; it can be valuable for historical purposes
to know which data was supplied and from which
outside sources it came.
n Coded; in RDF, represented by a URI. Coded data
would likely transform on the Semantic Web into
links to ontologies that could provide normalized,
human-readable identification strings on demand,
thus causing coded and normalized data to merge
into one type of data. Is it not possible, though,
that the coded form of normalized data might
continue to provide for more efficient searching
for computers as opposed to humans? Coded data
also has great cross-cultural value, since it is not
as language-dependent as literals or normalized
headings.
n Normalized Headings (controlled headings); in RDF,
represented by a URI. Normalized or controlled
headings are still necessary to provide users with
coherent, ordered displays of thousands of entities
that all match the user’s search for a particular
entity (work, author, subject, etc.). The reason Google
displays are so hideous is that, so far, the data
searched lacks any normalized display data. If
variant language forms of the name for an entity
64 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
are linked to an entity URI, it should be possible to
supply headings in the language and script desired
by a particular user.
n The RDF model
Those who have become familiar with FRBR over the
years will probably not find it too difficult to transition
from the FRBR conceptual model to the RDF model. What
FRBR calls an “entity,” RDF calls a “subject” and RDFS
calls a “class.” What FRBR calls an “attribute,” RDF calls
an “object” and RDFS calls a “property.” What FRBR calls
a “relationship,” RDF calls a “predicate” and RDFS calls a
“relationship” or a “semantic linkage” (see table 2).
The difficulty in any data-modeling exercise lies in
deciding what to treat as an entity or class and what to
treat as an attribute or property. The authors of FRBR
decided to create a class called expression to deal with any
change in the content of a work. When FRBR is applied
to serials, which change content with every issue, the
model does not work well. In my model, I found it useful
to create a new entity at the manifestation level, the serial
title, to deal with the type of change that is more relevant
to serials, the change in title. I also created another new
entity at the manifestation level, title-manifestation, to deal
with a change of title in a nonserial work that is not asso-
ciated with a change in content. One hundred years ago,
this entity would have been called title-edition. I am also
in the process of developing an entity at the expression
level—surrogate—to deal with reproductions of original
artworks that need to inherit the qualities of the original
artwork they reproduce without being treated as an edi-
tion of that original artwork, which ipso facto is unique.
These are just examples of cases in which it is not that
easy to decide on the classes or entities that are necessary
to accurately model bibliographic information. See the
appendix for a complete comparison of the classes and
entities defined in four different models: FRBR, FRAD,
RDA, and the Yee Cataloging Rules (YCR). The appendix
also shows variation among these models concerning
whether a given data element is treated as a class/entity
or as an attribute/property. The most notable examples
are name and preferred access point, which are treated as
classes/entities in FRAD, as attributes in FRBR and YCR,
and as both in RDA.
n RDF problems encountered
My goal for this paper is to institute discussion with
data modelers about which problems I observed are
insoluble and which are soluble:
1. Is there an assumption on the part of Semantic Web
developers that a given data element, such as a publisher name,
should be expressed as either a literal or using a URI (i.e., con-
trolled), but never both? Cataloging is rooted in humanistic
practices that require careful recording of evidence. There
will always be value in distinguishing and labeling the
following types of data:
n Copied as is from an artifact (transcribed)
n Supplied by a cataloger
n Categorized by a cataloger (controlled)
Tim Berners-Lee (the father of the Internet and the
Semantic Web) emphasizes the importance of record-
ing not just data but also its provenance for the sake of
authenticity.15 For many data elements, therefore, it will
be important to be able to record both a literal (tran-
scribed or composed form or both) and a URI (controlled
form). Is this a problem in RDF? As a corollary, if any
data that can be given a URI cannot also be represented
by a literal (transcribed and composed data, or one or the
other), it may not be possible to design coherent, readable
displays of the data describing a particular entity. Among
other things, cataloging is a discursive writing skill. Does
RDF require that all data be represented only once, either
by a literal or by a URI? Or is it perhaps possible that data
that has a URI could also have a transcribed or composed
form as a property? Perhaps it will even be possible to
store multiple snapshots of online works that change
over time to document variant forms of a name for works,
persons, and so on.
2. Will the Internet ever be fast enough to assemble the
equivalent of our current records from a collection of hundreds
or even thousands of URIs? In RDF, links are one-to-one
rather than one-to-many. This leads to a great prolifera-
tion of reciprocal links. The more granularity there is in
the data, the more linking is necessary to ensure that
atomized data elements are linked together. Potentially,
every piece of data describing a particular entity could be
represented by a URI leading out to a SKOS list of data
values. The number of links necessary to pull together
Table 2. The FRBR conceptual model translated into RDF
and RDFS
FRBR RDF RDFS
Entity Subject Class
Attribute Object Property
Relationship Predicate Relationship/
semantic linkage
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 65
all of the data just to describe one manifestation could
become astronomical, as could the number of one-to-one
links necessary to create the appearance of a one-to-many
link, such as the link between an author and all the works
of an author. Is the Internet really fast enough to assemble
a record from hundreds of URIs in a reasonable amount
of time? Given the often slow network throughput typical
of many of our current Internet connections, is it really
practical to expect all of these pieces to be pulled together
efficiently to create a single display for a single user? We
yet may feel nostalgia for the single manifestation-based
record that already has all of the relevant data in it (no
assembly required).
Bruce D’Arcus points out, however, that
I think if you’re dealing with RDF, you wouldn’t neces-
sarily be gathering these data in real-time. The URIs
that are the targets for those links are really just global
identifiers. How you get the triples is a separate matter.
So, for example, in my own personal case, I’m going
to put together an RDF store that is populated with
data from a variety of sources, but that data popula-
tion will happen by script, and I’ll still be querying a
single endpoint, where the RDF is stored in a relational
database.16
In other words, D’Arcus essentially will put them
all in one place, or in one database that “looks” from a
URI perspective to be “one place” where they’re already
gathered.
3. Is RDF capable of dealing with works that are identified
using their creators? We need to treat author as both an
entity in its own right and as a property of a work, and
in many cases the latter is the more important function
for user service. Lexical labels, or human-readable identi-
fiers for works that are identified using both the principal
author and the title, are particularly problematic in RDF
given that the principal author is an entity in its own
right. Is RDF capable of supporting the indexing neces-
sary to allow a user to search using any variant of the
author’s name and any variant of the title of a work in
combination and still retrieve all expressions and mani-
festations of that work, given that author will have a URI
of its own, linked by means of a relationship link to the
work URI? Is RDF capable of supporting the display of a
list of one thousand works, each identified by principal
author, in order first by principal author, then by title, then
by publication date, given that the preferred heading for
each principal author would have to be assembled from
the URI for that principal author and the preferred title for
each work would have to be assembled from the URI
for that work? For fear that this will not, in fact, be pos-
sible, I have put a human-readable work-identifier data
element into my model that consists of principal author
and title when appropriate, even though that means the
preferred name of the principal author may not be able to
be controlled by the entity record for the principal author.
Any guidance from experienced data modelers in this
regard would be appreciated.
According to Bruce D’Arcus, this is purely an inter-
face or application question that does not require a solu-
tion at the data layer.17 Since we have never had interfaces
or applications that would do this correctly, even though
the data is readily available in authority records, I am
skeptical about this answer!
Perhaps Bruce’s suggestion under item 9 of designat-
ing a sortName property for each entity is the solution
here as well. My human-readable work identifier con-
sisting of the name of the principal creator and uniform
title of work could be designated the sortName poperty
for the work. It would have to be changed whenever
the preferred form of the name for the principal creator
changed, however.
4. Do all possible inverse relationships need to be expressed
explicitly, or can they be inferred? My model is already quite
large, and I have not yet defined the inverse of every
property as I really should to have a correct RDF model.
In other words, for every property there needs to be an
inverse property; for example, the property isCreatorOf
needs to have the inverse property isCreatedBy; thus
“Twain” has the property isCreatorOf, while “Adventures
of Tom Sawyer” has the property isCreatedBy. Perhaps
users and inputters will not actually have to see the huge,
complex RDF data model that would result from creating
all the inverse relationships, but those who maintain the
model will have to deal with a great deal of complexity.
However, since I’m not a programmer, I don’t know how
the complexity of RDF compares to the complexity of
existing ILS software.
5. Can RDF solve the problems we are having now because
of the lack of transitivity or inheritance in the data models that
underlie current ILSes, or will RDF merely perpetuate these
problems? We have problems now with the data models
that underlie our current ILSes because of the inability of
these models to deal with hierarchical inheritance, such
that whatever is true of an entity in the hierarchy is also
true of every entity below that entity in the hierarchy. One
example is that of cross-references to a parent corporate
body that should be held to apply to all subdivisions of
that corporate body but never are in existing ILS systems.
There is a cross-reference from “FBI” to “United States.
Federal Bureau of Investigation,” but not from “FBI
Counterterrorism Division” to “United States. Federal
Bureau of Investigation. Counterterrorism Division.” For
that reason, a search in any OPAC name index for “FBI
Counterterrorism Division” will fail. We need systems
that recognize that data about a parent corporate body
is relevant to all subdivisions of that parent body. We
need systems that recognize that data about a work is
relevant to all expressions and manifestations of that
work. RDF allows you to link a work to an expression
66 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
and an expression to a manifestation, but I don’t believe
it allows you to encode the information that everything
that is true of the work is true of all of its expressions and
manifestations. Rob Styles seems to confirm this: “RDF
doesn’t have hierarchy. In computer science terms, it’s a
graph, not a tree, which means you can connect anything
to anything else in any direction.”18
Of course, not all links should be this kind of tran-
sitive or inheritance link. One expression of work A is
linked to another expression of work A by links to work
A, but whatever is true of one of those expressions is not
necessarily true of the other; one may be illustrated, for
example, while the other is not. Whatever is true of one
work is not necessarily true of another work related to it
by related work link.
It should be recognized that bibliographic data is rife
with hierarchy. It is one of our major tools for expressing
meaning to our users. Corporate bodies have corporate
subdivisions, and many things that are true for the par-
ent body also are true for its subdivisions. Subjects are
expressed using main headings and subject subdivisions,
and many things that are true for the main heading (such
as variant names) also are true for the heading combined
with one of its subdivisions. Geographic areas are con-
tained within larger geographic areas, and many things
that are true of the larger geographic area also are true
for smaller regions, counties, cities, etc., contained within
that larger geographic area. For all these reasons, I believe
that, to do effective displays and indexes for our biblio-
graphic data, it is critical that we be able to distinguish
between a hierarchical relationship and a nonhierarchical
relationship.
6. To recognize the fact that the subject of a book or a film
could be a work, a person, a concept, an object, an event, or a
place (all classes in the model), is there any reason we cannot
define subject itself as a property (a relationship) rather than a
class in its own right? In my model, all subject properties
are defined as having a domain of resource, meaning
there is no constraint as to the class to which these subject
properties apply. I’m not sure if there will be any fall-out
from that modeling decision.
7. How do we distinguish between the corporate behavior
of a jurisdiction and the subject behavior of a geographical loca-
tion? Sometimes a place is a jurisdiction and behaves like
a corporate body (e.g., United States is the name of the
government of the United States). Sometimes place is a
physical location in which something is located (e.g., the
birds discussed in a book about the birds of the United
States). To distinguish between the corporate behavior of
a jurisdiction and the subject behavior of a geographical
location, I have defined two different classes for place:
Place as Jurisdictional Corporate Body and Place as Geographic
Area. Will this cause problems in the model? Will there be
times when it prevents us from making elegant general-
izations in the model about place per se? There is a similar
problem with events. Some events are corporate bodies
(e.g., conferences that publish papers) and some are a
kind of subject (e.g., an earthquake). I have defined two
different classes for event: Conference or Other Event as
Corporate Body Creator and Event as Subject.
8. What is the best way to model a bound-with or an issued-
with relationship, or a part–whole relationship in which the
whole must be located to obtain the part? The bound-with
relationship is actually between two items containing
two different works, while the issued-with relationship
is between two manifestations containing two different
works (see figure 2). Is this a work-to-work relation-
ship? Will designating it a work-to-work relationship
cause problems for indicating which specific items or
manifestation-items of each work are physically located
in the same place? This question may also apply to those
part–whole relationships in which the part is physically
contained within the whole and both are located in the
same place (sometimes known as analytics). One thing to
bear in mind is that in all of these cases the relationship
between two works does not hold between all instances
of each work; it only holds for those particular instances
that are contained in the particular manifestation or item
that is bound with, issued with, or part of the whole.
However, if the relationship is modeled as a work-1-
manifestation to work-2-manifestation relationship, or a
work-1-item to work-2-item relationship,, care must be
taken in the design of displays to pull in enough infor-
mation about the two or more works so as not to confuse
the user.
9. How do we express the arrangement of elements that
have a definite order? I am having trouble imagining how
to encode the ordering of data elements that make up a
larger element, such as the pieces of a personal name.
This is really a desire to control the display of those atom-
ized elements so that they make sense to human beings
rather than just to machines. Could one define a property
such as natural language order of forename, surname, middle
name, patronymic, matronymic and/or clan name of a person
given that the ideal order of these elements might vary
from one person to another? Could one define proper-
ties such as sorting element 1, sorting element 2, sorting
element 3, etc., and assign them to the various pieces that
will be assembled to make a particular heading for an
entity, such as an LCSH heading for a historical period?
(Depending on the answer to the question in item 11,
it may or may not be possible to assign a property to a
property in this fashion.) Are there standard sorting rules
we need to be aware of (in Unicode, for example)? Are
there other RDF techniques available to deal with sorting
and arrangement?
Bruce D’Arcus suggests that, instead of coding the
name parts, it would be more useful to designate sort-
Name properties;19 might it not be necessary to designate
a sortName property for each variant name, as well,
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 67
for cases in which variants need to appear in sorted
displays? And wouldn’t these sortName properties com-
plicate maintenance over time as preferred and variant
names changed?
10. How do we link related data elements in such a way that
effective indexing and displays are possible? Some examples:
number and kind of instrument (e.g., music written for
two oboes and three guitars); multiple publishers, fre-
quencies, subtitles, editors, etc., with date spans for a
serial title change (or will it be necessary to create a new
manifestation for every single change in subtitle, pub-
lisher name, place of publication, etc?). The assumption
seems to be that there will be no repeatable data ele-
ments. Based on my somewhat limited experience with
RDF, it appears that there are record equivalents (every
data element—property or relationship—pertaining to
a particular entity with a URI), but there are no field or
subfield equivalents that allow the sublinking of related
pieces of data about an entity. Indeed, Rob Styles goes
so far as to argue that ultimately there is no notion of a
“record” in RDF.20 It is possible that blank nodes might
be able to fill in for fields and subfields in some cases for
grouping data, but there are dangers involved in their
use.21 To a cataloger, it looks as though the plan is for RDF
data to float around loose without any requirement that
there be a method for pulling it together into coherent
displays designed for human beings.
11. Can a property have a property in RDF? As an exam-
ple of where it might be useful to define a property of a
property, Robert Maxwell suggests that date of publication
is really an attribute (property) of the published by rela-
tionship (another property).22 Another example: In my
model, a variant title for a serial is a property. Can that
property itself have the property type of variant title to
encompass things like spine title, key title, etc.? Another
example appeared in item 9, in which it is suggested that
it might be desirable to assign sort-element properties to
the various elements of a name property.
12. How do we document record display decisions? There
is no way to record display decisions in RDF itself; it is
completely display-neutral. We could not safely commit
to a particular RDF–based data model until a significant
amount of sample bibliographic data had been created
and open-source indexing and display software had been
designed and user-tested on that data. It may be that we
will need to supplement RDF with some other encoding
mechanism that allows us to record display decisions
along with the data. Current cataloging rules are about
display as much as they are about content designation.
ISBD concerns the order in which the elements should
be displayed to humans. The cataloging objectives con-
cern display to users of such entity groups as the works
of an author, the editions of a work, and the works on a
subject.
13. Can all bibliographic data be reduced to either a class
or a property with a finite list of values? Another way to put
this is to ask if all that catalogers do could be reduced to
a set of pull-down menus. Cataloging is the art of writing
discursive prose as much as it is the ability to select the
correct value for a particular data element. We must deal
with ambiguous data (presented by Joe Blow could mean
that Joe created the entire work, produced it, distributed
it, sponsored it, or merely funded it). We must sometimes
record information without knowing its exact meaning.
We must deal with situations that have not been antici-
pated in advance. It is not possible to list every possible
kind of data and every possible value for each type of
Figure 2. examples of part–whole relationships. how might these
be best expressed in rdF?
issued-with relationship
A copy of Charlie Chaplin’s 1917 film The Immigrant
can be found on a videodisc compilation called Charlie
Chaplin, The Early Years along with two other Chaplin
films. This compilation was published and collected
by many different libraries and media centers. If a
user wants to view this copy of The Immigrant, he or
she will first have to locate Charlie Chaplin, The Early
Years, then look for the desired film at the beginning
of the first videodisc in the set. The issued-with rela-
tionship between The Immigrant and the other two
films on Charlie Chaplin, The Early Years is currently
expressed in the bibliographic record by means of a
“with” note:
First on Charlie Chaplin, the early years, v. 1 (62
min.) with: The count – Easy Street.
Bound-with relationship
The University of California, Los Angeles Film &
Television Archive has acquired a reel of 16 mm. film
from a collector who strung five Warner Bros. car-
toons together on a single reel of film. We can assume
that no other archive, library, or media collection will
have this particular compilation of cartoons, so the
relationship between the five cartoons is purely local
in nature. However, any user at the Film & Television
Archive who wishes to view one of these cartoons
will have to request a viewing appointment for the
entire reel and then find the desired cartoon among
the other four on the reel. The bound-with relation-
ship among these cartoons is currently expressed in a
holdings record by means of a “with” note:
Fourth on reel with: Daffy doodles – Tweety Pie –
I love to singa – Along Flirtation Walk.
68 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
data up front before any data is gathered. It will always
be necessary to provide a plain-text escape hatch. The
bibliographic world is a complex, constantly changing
world filled with ambiguity.
n What are the next steps?
In a sense, this paper is a first crude attempt at locating
unmapped territory that has not yet been explored. If we
were to decide as a community that it would be valu-
able to move our shared cataloging activities onto the
Semantic Web, we would have a lot of work ahead of us.
If some of the RDF problems described above are insolu-
ble, we may need to work with Semantic Web developers
to create a more sophisticated version of RDF that can
handle the transitivity and complex linking required by
our data. We will also need to encourage a very complex
existing community to evolve institutional structures that
would enable a more efficient use of the Internet for the
sharing of cataloging and other metadata creation. This is
not just a technological problem, but also a political one.
In the meantime, the experiment continues. Let the think-
ing and learning begin!
References and notes
1. “Notation3, or N3 as it is more commonly known, is
a shorthand non–XML serialization of Resource Description
Framework models, designed with human-readability in mind:
N3 is much more compact and readable than XML RDF nota-
tion. The format is being developed by Tim Berners-Lee and oth-
ers from the Semantic Web community.” Wikipedia, “Notation
3,” http://en.wikipedia.org/wiki/Notation_3 (accessed Feb. 19,
2009).
2. FRBR Review Group, www.ifla.org/VII/s13/wgfrbr/;
FRBR Review Group, FRANAR (Working Group on Functional
Requirements and Numbering of Authority Records), www
.ifla.org/VII/d4/wg-franar.htm; FRBR Review Group, FRSAR
(Working Group, Functional Requirements for Subject Authority
Records), www.ifla.org/VII/s29/wgfrsar.htm; FRBRoo, FRBR
Review Group, Working Group on FRBR/CRM Dialogue, www
.ifla.org/VII/s13/wgfrbr/FRBR-CRMdialogue_wg.htm.
3. Library of Congress, Response to On the Record: Report
of the Library of Congress Working Group on the Future of Bib-
liographic Control (Washington, D.C.: Library of Congress, 2008):
24, 39, 40, www.loc.gov/bibliographic-future/news/LCWGRpt
Response_DM_053008.pdf (accessed Mar. 25, 2009).
4. Ibid., 39.
5. Ibid., 41.
6. Dublin Core Metadata Initiative, DCMI/RDA Task
Group Wiki, http://www.dublincore.org/dcmirdataskgroup/
(accessed Mar. 25, 2009).
7. Mikael Nilsson, Andy Powell, Pete Johnston, and
Ambjorn Naeve, Expressing Dublin Core Metadata Using the
Resource Description Framework (RDF), http://dublincore.org/
documents/2008/01/14/dc-rdf/ (accessed Mar. 25, 2009).
8. See for example table 6.3 in FRBR, which maps to mani-
festation every kind of data that pertains to expression change
with the exception of language change. IFLA Study Group on
the Functional Requirements for Bibliographic Records, Func-
tional Requirements for Bibliographic Records (Munich: K. G. Saur,
1998): 95, http://www.ifla.org/VII/s13/frbr/frbr.pdf (accessed
Mar. 4, 2009).
9. Roy Tennant, “MARC Must Die,” Library Journal 127, no.
17 (Oct. 15, 2002): 26.
10. W3C, SKOS Simple Knowledge Organization System Refer-
ence, W3C Working Draft 29 August 2008, http://www.w3.org/
TR/skos-reference/ (accessed Mar. 25, 2009).
11. The extract in figure 1 is taken from my complete
RDF model, which can be found at http://myee.bol.ucla.edu/
ycrschemardf.txt.
12. Mary W. Elings and Gunter Waibel, “Metadata for All:
Descriptive Standards and Metadata Sharing Across Libraries,
Archives and Museums,” First Monday 12, no. 3 (Mar. 5, 2007),
http://www.uic.edu/htbin/cgiwrap/bin/ojs/index.php/fm/
article/view/1628/1543 (accessed Mar. 25, 2009).
13. OCLC, A Holdings Primer: Principles and Standards for Local
Holdings Records, 2nd ed. (Dublin, Ohio: OCLC, 2008), 4, http://
www.oclc.org/us/en/support/documentation/localholdings/
primer/Holdings%20Primer%202008.pdf (accessed Mar. 25,
2009).
14. The Library of Congress Working Group, On the Record: Report
of the Library of Congress Working Group on the Future of Bibliographic
Control (Washington, D.C.: Library of Congress, 2008): 30, http://
www.loc.gov/bibliographic-future/news/lcwg-ontherecord
-jan08-final.pdf (accessed Mar. 25, 2009).
15. Talis, Sir Tim Berners-Lee Talks with Talis about the Seman-
tic Web: Transcript of an Interview Recorded on 7 February 2008,
http://talis-podcasts.s3.amazonaws.com/twt20080207_TimBL
.html (accessed Mar. 25, 2009).
16. Bruce D’Arcus, e-mail to author, Mar. 18, 2008.
17. Ibid.
18. Rob Styles, e-mail to author, Mar. 25, 2008.
19. Bruce D’Arcus, e-mail to author, Mar. 18, 2008.
20. Rob Styles, e-mail to author, Mar. 25, 2008.
21. W3C, “Section 2.3, Structured Property Values and Blank
Nodes,” in RDF Primer: W3C Recommendation 10 February 2004,
http://www.w3.org/TR/rdf-primer/#structuredproperties
(accessed Mar. 25, 2009).
22. Robert Maxwell, FRBR: A Guide for the Perplexed (Chicago:
ALA, 2008).
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 69
Entities/classes in rDa, FrBr, FraD compared to Yee Cataloging rules (YCr)
RDA, FRBR, and FRAD YCR
Group 1: Work Work
Group 1: Expression Expression
Surrogate
Group 1: Manifestation Manifestation
Title-manifestation
Serial title
Group 1: Item Item
Group 2: Person Person
Fictitious character
Performing animal
Group 2: Corporate body Corporate body
Corporate subdivision
Place as jurisdictional corporate body
Conference or other event as corporate body creator
Jurisdictional corporate subdivision
Family (RDA and FRAD only)
Group 3: Concept Concept
Group 3: Object Object
Group 3: Event Event or historical period as subject
Group 3: Place Place as geographic area
Discipline
Genre/form
Name
Identifier
Controlled access point
Rules (FRAD only)
Agency (FRAD only)
APPENDIx. Entity/class and attribute/property comparisons
70 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
attributes/properties in FrBr compared to FraD
Model
Entity FRBR FRAD
Work title of the work
form of work
date of the work
other distinguishing characteristics
intended termination
intended audience
context for the work
medium of performance (musical work)
numeric designation (musical work)
key (musical work)
coordinates (cartographic work)
equinox (cartographic work)
form of work
date of the work
medium of performance
subject of the work
numeric designation
key
place of origin of the work
original language of the work
history
other distinguishing characteristic
Expression title of the expression
form of expression
date of expression
language of expression
other distinguishing characteristics
extensibility of expression
revisability of expression
extent of the expression
summarization of content
context for the expression
critical response to the expression
use restrictions on the expression
sequencing pattern (serial)
expected regularity of issue (serial)
expected frequency of issue (serial)
type of score (musical notation)
medium of performance (musical notation or recorded sound)
scale (cartographic image/object)
projection (cartographic image/object)
presentation technique (cartographic image/object)
representation of relief (cartographic image/object)
geodetic, grid, and vertical measurement (cartographic image/
object)
recording technique (remote sensing image)
special characteristic (remote sensing image)
technique (graphic or projected image)
form of expression
date of expression
language of expression
technique
other distinguishing characteristic
Surrogate
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 71
Model
Entity FRBR FRAD
Manifestation title of the manifestation
statement of responsibility
edition/issue designation
place of publication/distribution
publisher/distributor
date of publication/distribution
fabricator/manufacturer
series statement
form of carrier
extent of the carrier
physical medium
capture mode
dimensions of the carrier
manifestation identifier
source for acquisition/access authorization
terms of availability
access restrictions on the manifestation
typeface (printed book)
type size (printed book)
foliation (hand-printed book)
collation (hand-printed book)
publication status (serial)
numbering (serial)
playing speed (sound recording)
groove width (sound recording)
kind of cutting (sound recording)
tape configuration (sound recording)
kind of sound (sound recording)
special reproduction characteristic (sound recording)
colour (image)
reduction ratio (microform)
polarity (microform or visual projection)
generation (microform or visual projection)
presentation format (visual projection)
system requirements (electronic resource)
file characteristics (electronic resource)
mode of access (remote access electronic resource)
access address (remote access electronic resource)
edition/issue designation
place of publication/distribution
publisher/distributor
date of publication/distribution
form of carrier
numbering
Title-manifestation
Serial title
Item item identifier
fingerprint
provenance of the item
marks/inscriptions
exhibition history
condition of the item
treatment history
scheduled treatment
access restrictions on the item
location of item
attributes/properties in FrBr compared to FraD (cont.)
72 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
Model
Entity FRBR FRAD
Person name of person
dates of person
title of person
other designation associated with the person
dates associated with the person
title of person
other designation associated with
the person
gender
place of birth
place of death
country
place of residence
affiliation
address
language of person
field of activity
profession/occupation
biography/history
Fictitious character
Performing animal
Corporate body name of the corporate body
number associated with the corporate body
place associated with the corporate body
date associated with the corporate body
other designation associated with the corporate body
place associated with the
corporate body
date associated with the corporate
body
other designation associated with
the corporate body
type of corporate body
language of the corporate body
address
field of activity
history
Corporate subdivision
Place as jurisdictional
corporate body
Conference or other
event as corporate
body creator
Jurisdictional corporate
subdivision
Family type of family
dates of family
places associated with family
history of family
Concept term for the concept type of concept
Object term for the object type of object
date of production
place of production
producer/fabricator
physical medium
Event term for the event date associated with the event
place associated with the event
attributes/properties in FrBr compared to FraD (cont.)
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 73
Model
Entity FRBR FRAD
Place term for the place coordinates
other geographical information
Discipline
Genre/form
Name type of name
scope of usage
dates of usage
language of name
script of name
transliteration scheme of name
Identifier type of identifier
identifier string
suffix
Controlled access point type of controlled access point
status of controlled access point
designated usage of controlled
access point
undifferentiated access point
language of base access point
script of base access point
script of cataloguing
transliteration scheme of base
access point
transliteration scheme of
cataloguing
source of controlled access point
base access point
addition
Rules citation for rules
rules identifier
Agency name of agency
agency identifier
location of agency
attributes/properties in FrBr compared to FraD (cont.)
74 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
attributes/properties in rDa compared to YCr
Model
Entity RDA YCR
Work title of the work
form of work
date of work
place of origin of work
medium of performance
numeric designation
key
signatory to a treaty, etc.
other distinguishing characteristic of
the work
original language of the work
history of the work
identifier for the work
nature of the content
coverage of the content
coordinates of cartographic content
equinox
epoch
intended audience
system of organization
dissertation or theses information
key identifier for work
language-based identifier (preferred lexical label)
variant language-based identifier (alternate lexical label)
language-based identifier (preferred lexical label) for work
language-based identifier for work (preferred lexical label) identified
by PrincipalCreator in combination with uniform title
language-based identifier (preferred lexical label) for work identified
by title alone (uniform title)
supplied title for work
variant title for work
original language of work
responsibility for work
original publication statement of work
dates associated with work
original publication/release/broadcast date of work
copyright date of work
creation date of work
date of first recording of a work
date of first performance of a work
finding date of naturally occurring object
original publisher/distributor/broadcaster of work
places associated with work
original place of publication/distribution/broadcasting for work
country of origin of work
place of creation of work
place of first recording of work
place of first performance of work
finding place of naturally occurring object
original method of publication/distribution/broadcast of work
serial or integrating work original numeric and/or alphabetic
designations—beginning
serial or integrating work original chronological designations—
beginning
serial or integrating work original numeric and/or alphabetic
designations—ending
serial or integrating work original chronological designations—
ending
encoding of content of work
genre/form of content of work
original instrumentation of musical work
instrumentation of musical work—number of a particular instrument
instrumentation of musical work—type of instrument
original voice(s) of musical work
voice(s) of musical work—number of a particular type of voice
voice(s) of musical work—type of voice
original key of musical work
numeric designation of musical work
coordinates of cartographic work
equinox of cartographic work
original physical characteristics of work
original extent of work
original dimensions of work
mode of issuance of work
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 75
Model
Entity RDA YCR
Work (cont.) original aspect ratio of moving image work
original image format of moving image work
original base of work
original materials applied to base of work
work summary
work contents list
custodial history of work
creation of archival collection
censorship history of work
note about relationship(s) to other works
Expression content type
date of expression
language of expression
other distinguishing characteristic of
the expression
identifier for the expression
summarization of the content
place and date of capture
language of the content
form of notation
accessibility content
illustrative content
supplementary content
colour content
sound content
aspect ratio
format of notated music
medium of performance of musical
content
duration
performer, narrator, and/or presenter
artistic and/or technical credits
scale
projection of cartographic content
other details of cartographic content
awards
key identifier for expression
language-based identifier (preferred lexical label) for expression
variant title for expression
nature of modification of expression
expression title
expression statement of responsibility
edition statement
scale of cartographic expression
projection of cartographic expression
publication statement of expression
place of publication/distribution/release/broadcasting for expression
place of recording for expression
publisher/distributor/releaser/broadcaster for expression
publication/distribution/release/broadcast date for expression
copyright date for expression
date of recording for expression
numeric and/or alphabetic designations for serial expressions
chronological designations for serial expressions
performance date for expression
place of performance for expression
extent of expression
content of expression
language of expression text
language of expression captions
language of expression sound track
language of sung or spoken text of expression
language of expression subtitles
language of expression intertitles
language of summary or abstract of expression
instrumentation of musical expression
instrumentation of musical expression—number of a particular
instrument
instrumentation of musical expression—type of instrument
voice(s) of musical expression
voice(s) of musical expression—number of a particular type of voice
voice(s) of musical expression—type of voice
key of musical expression
appendages to the expression
expression series statement
mode of issuance for expression
notes about expression
Surrogate [under development]
attributes/properties in rDa compared to YCr (cont.)
76 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
Model
Entity RDA YCR
Manifestation title
statement of responsibility
edition statement
numbering of serials
production statement
publication statement
distribution statement
manufacture statement
copyright date
series statement
mode of issuance
frequency
identifier for the manifestation
note
media type
carrier type
base material
applied material
mount
production method
generation
layout
book format
font size
polarity
reduction ratio
sound characteristics
projection characteristics of motion
picture film
video characteristics
digital file characteristics
equipment and system requirements
terms of availability
key identifier for manifestation
publication statement of manifestation
place of publication/distribution/release/broadcast of manifestation
manifestation publisher/distributor/releaser/broadcaster
manifestation date of publication/distribution/release/broadcast
carrier edition statement
carrier piece count
carrier name
carrier broadcast standard
carrier recording type
carrier playing speed
carrier configuration of playback channels
process used to produce carrier
carrier dimensions
carrier base materials
carrier generation
carrier polarity
materials applied to carrier
carrier encoding format
intermediation tool requirements
system requirements
serial manifestation illustration statement
manifestation standard number
manifestation ISBN
manifestation ISSN
manifestation publisher number
manifestation universal product code
notes about manifestation
Title-
manifestation
key identifier for title-manifestation
variant title for title-manifestation
title-manifestation title
title-manifestation statement of responsibilities
title-manifestation edition statement
publication statement of title-manifestation
place of publication/distribution/release/broadcasting of title-
manifestation
publisher/distributor/releaser, broadcaster of title-manifestation
date of publication/distribution/release/broadcast of title-
manifestation
title-manifestation series
title-manifestation mode of issuance
notes about title-manifestation
title-manifestation standard number
attributes/properties in rDa compared to YCr (cont.)
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 77
Model
Entity RDA YCR
Serial title key identifier for serial title
variant title for serial title
title of serial title
serial title statement of responsibility
serial title edition statement
publication statement of serial title
place of publication/distribution/release/broadcast of serial title
publisher/distributor/releaser/broadcaster of serial title
date of publication/distribution/release/broadcast of serial title
serial title beginning numeric and/or alphabetic designations
serial title beginning chronological designations
serial title ending numeric and/or alphabetic designations
serial title ending chronological designations
serial title frequency
serial title mode of issuance
serial title illustration statement
notes about serial title
serial title ISSN-L
Item preferred citation
custodial history
immediate source of acquisition
identifier for the item
item-specific carrier characteristics
key identifier for item
item barcode
item location
item call number or accession number
item copy number
item provenance
item condition
item marks and inscriptions
item exhibition history
item treatment history
item scheduled treatment
item access restrictions
attributes/properties in rDa compared to YCr (cont.)
78 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
Model
Entity RDA YCR
Person name of the person
preferred name for the person
variant name for the person
date associated with the person
title of the person
fuller form of name
other designation associated with the
person
gender
place of birth
place of death
country associated with the person
place of residence
address of the person
affiliation
language of the person
field of activity of the person
profession or occupation
biographical information
identifier for the person
key identifier for person
language-based identifier (preferred lexical label) for person
clan name of person
forename/given name/first name of person
matronymic of person
middle name of person
nickname of person
patronymic of person
surname/family name of person
natural language order of forename, surname, middle name,
patronymic, matronymic and/or clan name of person
affiliation of person
biography/history of person
date of birth of person
date of death of person
ethnicity of person
field of activity of person
gender of person
language of person
place of birth of person
place of death of person
place of residence of person
political affiliation of person
profession/occupation of person
religion of person
variant name for person
Fictitious
character
[under development]
Performing
animal
[under development]
Corporate body name of the corporate body
preferred name for the corporate body
variant name for the corporate body
place associated with the corporate
body
date associated with the corporate body
associated institution
other designation associated with the
corporate body
language of the corporate body
address of the corporate body
field of activity of the corporate body
corporate history
identifier for the corporate body
key identifier for corporate body
language-based identifier (preferred lexical label) for corporate
body
dates associated with corporate body
field of activity of corporate body
history of corporate body
language of corporate body
place associated with corporate body
type of corporate body
variant name for corporate body
Corporate
subdivision
[under development]
Place as
jurisdictional
corporate
body
[under development]
attributes/properties in rDa compared to YCr (cont.)
CaN BiBlioGraPHiC DaTa BE PuT DirECTlY oNTo THE sEMaNTiC wEB? | YEE 79
Model
Entity RDA YCR
Conference or
other event
as corporate
body creator
[under development]
Jurisdictional
corporate
subdivision
[under development]
Family name of the family
preferred name for the family
variant name for the family
type of family
date associated with the family
place associated with the family
prominent member of the family
hereditary title
family history
identifier for the family
Concept term for the concept
preferred term for the concept
variant term for the concept
type of concept
identifier for the concept
key identifier for concept
language-based identifier (preferred lexical label) for concept
qualifier for concept language-based identifier
variant name for concept
Object name of the object
preferred name for the object
variant name for the object
type of object
date of production
place of production
producer/fabricator
physical medium
identifier for the object
key identifier for object
language-based identifier (preferred lexical label) for object
qualifier for object language-based identifier
variant name for object
Event name of the event
preferred name for the event
variant name for the event
date associated with the event
place associated with the event
identifier for the event
key identifier for event or historical period as subject
language-based identifier (preferred lexical label) for event or
historical period as subject
beginning date for event or historical period as subject
ending date for event or historical period as subject
variant name for event or historical period as subject
Place name of the place
preferred name for the place
variant name for the place
coordinates
other geographical information
identifier for the place
key identifier for place as geographic area
language-based identifier (preferred lexical label) for place as
geographic area
qualifier for place as geographic area
variant name for place as geographic area
Discipline key identifier for discipline
language-based identifier (preferred lexical label) (name or
classification number or symbol) for discipline
translation of meaning of classification number or symbol for
discipline
attributes/properties in rDa compared to YCr (cont.)
80 iNForMaTioN TECHNoloGY aND liBrariEs | JuNE 2009
Model
Entity RDA YCR
Genre/form key identifier for genre/form
language-based identifier (preferred lexical label) for genre/form
variant name for genre/form
Name scope of usage
date of usage
Identifier
Controlled
access point
Rules
Agency
note: In rdA, the following attributes have not yet been assigned to a particular class or entity: extent, dimensions, terms of availability, contact information,
restrictions on access, restrictions on use, uniform resource locator, status of identification, source consulted, cataloguer’s note, status of identification, and
undifferentiated name indicator. Name is being treated as both a class and a property. Identifier and controlled access point are treated as properties rather than
classes in both rdA and Ycr.
attributes/properties in rDa compared to YCr (cont.)