Content uploaded by Martin Doerr
Author content
All content in this area was uploaded by Martin Doerr on Feb 25, 2014
Content may be subject to copyright.
1
The CIDOC CRM – an Ontological Approach to
Semantic Interoperability of Metadata
M. Doerr
Institute of Computer Science,
Foundation for Research and Technology – Hellas,
Science and Technology Park of Crete,
Vassilika Vouton, P.O. Box 1385, GR 711 10, Heraklion, Crete, Greece
martin@ics.forth.gr
Abstract: This paper presents the methodology that has been successfully employed over the past 7 years by an
interdisciplinary team to create the CIDOC Conceptual Reference Model (CRM), a high-level ontology to enable
information integration for cultural heritage data and their correlation with library and archive information. The CIDOC
CRM is now in the process to become an ISO standard. This paper justifies in detail the methodology and design by
functional requirements and gives examples of its contents. The CIDOC CRM analyses the common conceptualizations
behind data and metadata structures to support data transformation, mediation and merging. It is argued that such ontologies
are property-centric, in contrast to terminological systems, and should be built with different methodologies. It is
demonstrated that ontological and epistemological arguments are equally important for an effective design, in particular
when dealing with knowledge from the past in any domain. It is assumed that the presented methodology and the upper level
of the ontology are applicable in a far wider domain.
1 Introduction
The creation of the World Wide Web has had a profound impact on the ease with which information can be
distributed and presented. This has spurred an increasing interest from professionals, the general public, and
consequently politicians to make publicly available the tremendous wealth of information kept in museums,
archives and libraries - the so-called “memory organisations”. Quite naturally, their development has focused on
presentation, such as websites and interfaces to their local databases. Now with more and more information
becoming available, there is an increasing demand for targeted global search, comparative studies, data transfer
and data migration between heterogeneous sources of cultural contents. This requires interoperability not only at
the encoding level - a task solved well by XML for instance - but also at the more complex semantics level,
where lie the characteristics of the domain.
Formal methods are very helpful to deal effectively with the large amounts of information coming together on
the Internet. Information about cultural heritage poses particular challenges for formal handling – rather not, as
often assumed, because it is ill defined, but because of its high diversity, and the intrinsic incompleteness of
information about the past. So far most attempts for semantic interoperability have concentrated on the
development and standardization of a shared core data structure (e.g. the CIDOC Relational Model) and a
terminology system. In cases a common data structure seemed to be impossible, at least a common metadata
schema as “finding aids” have been attempted, the most prominent example being the Dublin Core Element Set.
On the terminology side, the Library of Congress Subject Headings and the Art & Architecture Thesaurus are
characteristic examples of standards in the US and beyond, for which equivalents in several other languages
have been created.
In the meanwhile, the reality of semantic interoperability is getting frustrating. In the cultural area alone, dozens
of “standard” and hundreds of proprietary metadata and data structures exist, as well as hundreds of terminology
systems. Core systems like the Dublin Core represent a common denominator by far too small to fulfill advanced
requirements. Overstretching its already limited semantics in order to capture complex contents leads to further
loss of meaning (“metadata pidgin”[1]), even though most of the contents encoded in the various structures seem
to be pretty comprehensive in common sense terms, and are often inter-related. We make the hypothesis that
much of the diversity of data and metadata structures is due to the fact that they are designed for data capturing –
as a guide for good practice of what should be documented, and to optimise coding and storage costs for a
specific application – far more than for interpreting data. Necessarily, these data structures are relatively flat (in
order to suggest a workflow of entering data to the user) and full of application-specific hidden constants and
simplifications.
2
Since 1996, we have taken part in the development of the CIDOC Conceptual Reference Model (CRM) [2],
[3],[4], an attempt of the CIDOC Committee of the International Council of Museums (ICOM) to achieve
semantic interoperability for museum data. Work has started in the beginning on a more intuitive base, from a
knowledge representation model [5],[6], based on the consensus of a varying team of different domain experts
and based on strict intellectual principles. It has got a wide acceptance in CIDOC and by other relevant
stakeholders in the domain and in September, 2000, the CIDOC CRM was successfully submitted to ISO TC46
as new work item. It is now registered as ISO/CD 21127 and is expected to become an ISO standard in 2003. It
is now in a very stable form, and contains 80 classes and 130 properties, both arranged in multiple isA
hierarchies. In the meanwhile, several applications [7] and comparison with related work improves our
theoretical understanding of the work done and still ongoing.
Instead of seeking a common schema as a prescription for data capturing, which would be supposed to ensure
semantic compatibility of the produced data, we have followed with the CIDOC CRM what Bergamaschi et.al
[8] call a semantic approach to integrated access. Being convinced that collecting information is already done
well by the existing data structures, possible improvements not withstanding, we aim simply for read-only
integration, as e.g. in the DWQ project [9][10]. This comprises data migration, data merging (materialized data
integration as in data warehouse applications), and virtual integration via query mediation [11]. Recently more
and more projects and theoreticians support the use of formal ontologies as common conceptual schema for
information integration [12], [13], [8],[16], [14],
• to provide a conceptual basis for understanding and analyzing existing (meta)data structures and
instances;
• to give guidance to communities beginning to examine and develop descriptive vocabularies;
• to develop a conceptual basis for automated mapping amongst metadata structures and their
instances[14].
Its seems that the semantics behind a large set of diverse (meta)data structures from a domain with many
subdisciplines can be expressed by a coherent formal ontology based on the common conceptualisations of the
respective domain experts, whereas the data entry structures themselves often seem to resist merging. We have
followed a pragmatic approach to separate a kind of top-level ontology [12], which represents knowledge
extracted from schemata and data structures, from pure terminology. This was partially done in order to keep the
basic ontology in a manageable size. The semantics of the data structures are richer in n-ary relationships
(attributes, properties etc.) than in fine distinctions between classes, whereas thesauri are just the opposite – they
build rich isA (BT/NT) hierarchies but typically employ only one (“RT”) relationship for any other internal
conceptual relation. We turned this observation into a rule: classes were introduced in the CRM only for the
domains or ranges of the relevant relationships, such that any other ontological refinement of the classes can be
done as additional “terminological distinction” without interfering with the system of relationships (see also
[2]). Such a conceptual model seems to cover the ontological top-level automatically, and provides an
integrating framework for the often isolated hierarchies found in terminological systems (Fig. 2). We call such a
model a “property centric ontology” to stress the specific character and functionality.
This paper presents the CIDOC CRM from a methodological point of view. It relates the intended scope and
functionality to the ontological principles that governed its design, presents its key concepts and positions the
model with respect to relevant related work. We expect this methodology to be applicable to other domains,
even though there is no experience yet to support this claim. The significant contribution of this work are
considerations about the specific nature of cultural – historical knowledge and reasoning, which aims at the
reconstruction of possible past worlds from loosely correlated records rather than at control and prediction of
systems, as in engineering knowledge. “Historical” must be understood in the widest sense, be it cultural,
political, archaeological, medical records, managerial records of enterprises, records of scientific experiments or
criminalistic data.
2 The Problem
2.1 The Necessity of Data Structure Diversity
3
Let us regard here data structures for long-term storing of data, as database schemata, but also tagging schemes
like SGML/XML DTD, RDF Schema, and data structures designed as fill-in forms guiding users to a complete
and consistent documentation, be it as primary data or as metadata about another information source. These
structures are always a compromise between the complexity of the information one would like to make
accessible by formal queries, the complexity the user can handle, the complexity of the system the user can
afford to implement or to pay for, and the cost to learn those structures and fill them with contents. As most
applications run in a relatively uniform environment – a library, a museum of modern art, a historical archive of
administrational records, a paleontological museum etc., much of the complexity of the one application is
negligible for another. This allows for a variety of simplifications, which are required to create efficient
applications. For example, documentation in modern art needs no other dating than Julian dates, whereas for
archaeology dating is a process of multiple measurements, evaluation of sources, inferring and justification,
including different dating systems. It makes no sense to ask the modern art curator about carbon 14
measurements, nor to reserve dozens of special fields and storage space never used. Nevertheless, and this is the
crucial point, the notion of date and dating of both experts are completely compatible; there is no difference in
conceptualisation, at least from a scientific point of view. The complexity typical for archaeology hardly ever
occurs in modern art, however. Similarly, the documentation of a historical building implies a complex history
with various phases and persons implied, whereas paintings are typically created in a relatively compact process.
Consequently, art documentation schemes like AMICO [15] do not capture multiple creators for multiple phases
of objects, in contrast to the architectural descriptions of the Greek National Archive of Monuments [16],[17].
The rare exceptions to such simplifications are normally handled by free text comments and are not used as
reasons to change the schema.
Another criterion of simplification appears in the “finding aids”. Subtle differences in association, like John and
Mary collaborating in the design of one building, but John overseeing the design and Mary the construction of
another building, cannot be captured by a scheme listing all involved persons without their individual roles. The
question is, how much noise will the replacement of the query “John and Mary designing” by “John and Mary
involved” create – probably not much. This is the justification for the use of “flat” metadata records like Dublin
Core, that sum up relevant persons and relevant dates etc. without interrelations. These simplifications actually
violate our conceptualisation, the sources retrieved on this basis can only be sorted out by reading them. How
long this will work is a question of scale. If we have the resources and the requirements for more complex
searches and processing, we must find a way to “recover” the common conceptualisation behind these
simplifications, if at all possible, or improve our data structures (see section 4.4).
2.2 The Yalta Conference – a demonstration case
Let us regard an artificial, but realistic demonstration case about information objects related to the Yalta
Conference in February, 1945. This was the event officially designating the end of WWII. One can hardly find
a better documented event in history. We owe the association to the photo below to the Microsoft Encarta
Encyclopedia 2000. The text and the TGN record we found on the Internet. The titles are as we have found
them. We have created the demonstration metadata below from the information we found associated with the
objects:
The State Department of the United States holds a copy of the Yalta Agreement. One paragraph begins, “The
following declaration has been approved: The Premier of the Union of Soviet Socialist Republics, the Prime
Minister of the United Kingdom and the President of the United States of America have consulted with each
other in the common interests of the people of their countries and those of liberated Europe. They jointly declare
their mutual agreement to concert …” [http://www.fordham.edu/halsall/mod/1945YALTA.html]. A Dublin Core
record may be:
Type: Text
Title: Protocol of Proceedings of Crimea Conference
Title.Subtitle: II. Declaration of Liberated Europe
Date: February 11, 1945.
Creator: The Premier of the Union of Soviet Socialist Republics
The Prime Minister of the United Kingdom
4
The President of the United States of America
Publisher: State Department
Subject: Postwar division of Europe and Japan
The Bettmann Archive in New York holds a world-famous photo of this event (Fig. 1). A Dublin Core record
for this photo might be:
Type: Image
Title: Allied Leaders at Yalta
Date: 1945
Publisher: United Press International (UPI)
Source: The Bettmann Archive
Copyright: Corbis
References: Churchill, Roosevelt, Stalin
Figure 1: Allied Leaders at Yalta
The striking point is that both metadata records have nothing more in common than “1945”, hardly a distinctive
attribute. An “integrating” piece of information comes from the Thesaurus of Geographic Names (TGN,
[http://www.getty.edu/research/tools/vocabulary/tgn/index.html]), which may be captured by the following
metadata:
TGN Id: 7012124
Names: Yalta (C,V), Jalta (C,V)
Types: inhabited place(C), city (C)
Position: Lat: 44 30 N,Long: 034 10 E
Hierarchy: Europe (continent) <– Ukrayina (nation) <– Krym (autonomous republic)
Note: Located on S shore of Crimean Peninsula; site of conference between Allied powers in
WW II in 1945; is a vacation resort noted for pleasant climate, & coastal &
mountain scenery; produces wine, canned fruit & tobacco products.
Source: TGN, Thesaurus of Geographic Names
The keyword “Crimea” can finally be found under the foreign names for “Krym”, i.e. via another record
(id=1003381). This example demonstrates a fundamental problem: In order to retrieve information related to
one specific subject, information from multiple sources must be integrated. Vocabulary and data structure
unification only does not solve the problem.
2.3 Requirements
One problem in this example is to be able to relate Crimea to Krym and then to Yalta, the Premier of the Union
of Soviet Socialist Republics to Joseph Stalin and to the Allied Leaders etc. A deeper problem is the fact that
the artifacts do not fit our question: People document persistent items like images, texts, and places, but our
question was about an event, here the Yalta Conference, something that is only indirectly preserved in those
items. The data structures express certain relationships between items, which may or may not be globally
identified. Other relations are hidden, like UPI taking pictures, which can be either guessed from the context or
must be recovered from secondary sources or from background knowledge (at these times, the press
photographers were not documented). Having argued above that data structures are full of simplifications and
hidden constants, we see that the main problem is to recover this information during data integration. This is
where ontologies are most valuable.
5
When we started work on the CIDOC CRM in 1996, the CIDOC working groups had virtually given up on
creating one standard data structure for all museums. We have assumed (and still assume) that a fairly small set
of good practice guides (e.g. mda SPECTRUM [18], CIDOC International Guidelines for Museum Object
Information [19]) and standard data structures already express well what museum professionals want and should
say about their objects in the various disciplines (an extended list can be found in [20]), albeit these can still be
improved. In particular, the necessary constraints to improve data integrity are typically applied locally at data
entry time already.
The global interoperability between disciplines is clearly needed for the following functions, after data has been
created:
• The mediation of global queries to local structures [11];
• the extraction of individual statements from larger units of documentation and
• their comparison for alternative opinions;
• the transformation of data for migration to other systems and
• for merging into more informative units (like data warehouses).
The CIDOC CRM working group wanted to provide one key element: the encoding of the key domain
conceptualizations by an interdisciplinary group in a form that enables the above functionality and is extensible
enough to ensure a long life-cycle and increasing coverage of details and disciplines. In addition, the ontology is
also thought as an intellectual guide in the requirements analysis and conceptual modeling phase of cultural
information systems as proposed by [12]. Parallel to the on-going work, more and more methodological
principles have been elaborated and applied on the basis of these fundamental requirements. We are now in the
position to give a consistent account of this methodology, that has been documented so far only in presentation
slides and minutes of the working group. (e.g. [21]).
3 About the CRM Methodology
The problems computer scientists and system implementers have in comprehending the logic of cultural
concepts seems to be equally notorious as the inability of the cultural professionals to communicate those to
computer scientists. The CIDOC CRM working group is therefore interdisciplinary, aiming at closing that gap.
People with background in museology, history of arts, archaeology, natural history, physics, computer science,
philosophy and others where involved. We have achieved a functional compromise between the complexity of
the conceptualisations and the complexity of formalism the participants would appreciate. Therefore the AI
reader may miss in this work some obvious formalization, which is due to future work by the appropriate
specialists. On the other side we could convey in a series of targeted seminars more KR principles to non-
experts than in any other related standardization work.
Given the limited resources of a project that had no funding at all until recently, and the interdisciplinary
character of the group, the concern has always been to concentrate the resources on the most effective task for
such a group: to achieve consensus about the ontological commitment of a set of formally defined core concepts
of the domain in a way that can guide implementers and computer scientists, and can later be refined by domain
specialists. Therefore, many good contributions (e.g. modelling beliefs) were excluded just because they could
be taken over either by specialists in a later stage or because they could be dealt with separately. Several of them
are mentioned in this paper. Also, strict neutrality with respect to commercial interest groups is the declared
policy of CIDOC. All this contributed to create a construct of high intellectual quality and coherence. Starting
from an initial formulation of the scope of the CIDOC CRM [21], we have independently developed intellectual
principles similar to those in [13].
The CIDOC CRM is a formal conceptual model, however CIDOC CRM instantiation data (factual knowledge)
is allowed to be contradictory. Historical data - any description made in the past about the past, be it scientific,
medical or cultural – is normally unique and cannot be verified, falsified nor completed in an absolute sense. In
history, any conflict resolution of contradictory records is nothing more than yet another opinion. So, for an
ontology capable of supporting the collection of knowledge from historical data, ontological principles about
how we perceive and express things must be taken into account, and epistemological principles about how
knowledge can be acquired must be respected. There can be huge differences in the credibility of propositions.
Typically claims about the existence of material particulars (e.g. El Greco, Mona Lisa, Niniveh) are by far more
stable than the reported relationships and attributes. A bit more frequent than absolute doubts about existence of
6
material particulars are doubts of whether two individuals have actually been one (two pieces of pottery may
have been from the same pot, or two different pots), or similarly if one has been two (see e.g. the Union List of
Artist Names, [23]). Without going into more detail here, we want to
• provide an ontologically correct conceptual model that is
• compatible with the granularity of knowledge we typically get from our sources, and
• allows for compiling gracefully contradictory information.
3.1 Integration of Context-Free Propositions
In the following we mean by conceptual model or ontology a description of categorical knowledge about
“possible states of affairs” rather than about one state of affairs [12], and regard both as a special kind of
knowledge base [24]. We prefer the term “conceptual model” when talking about actual instantiation and
constructs dictated more by the representation formalism than the intended meaning. Categorical knowledge
may come from the analysis of data structures, hidden constants or terminology used in the data. We have the
vision of a global semantic network model, a fusion of relevant knowledge from all museum sources,
abstracted from their context of creation and units of documentation under a common conceptual model. The
network should, however, not replace the qualities of good scholarly text. Rather it should maintain links to
related primary textual sources to enable their discovery under relevant criteria.
Actors Events Objects
Integrated
factual knowledge
Thesauri
extent
CRM entities
Ontology
extension
Sources
and
metadata
(XML/RDF)
Background
knowledge /
Authorities
CIDOC
CRM
Figure 2: An information integration architecture
Figure 2 shows a possible architecture integrating a property-centric top ontology (here the CIDOC CRM),
which provides the semantics for properties of subordinate terminological systems and an integrated factual
knowledge layer constructed from source data, metadata and background knowledge like the TGN and other
authorities.
The CIDOC CRM plays the role of an “enterprise model” [10], in the following called “common model”. We
assume for all sources the existence of a conceptual model (“source model”), and that source data can be
expressed without loss of meaning in terms of a source model, which is based on the same formalism as the
enterprise model. The source model may be restricted to the semantics falling within the scope of the common
model. As a representation formalism we have selected the TELOS data model [25],[26] without its assertional
language. TELOS, as many other knowledge representation languages, decomposes knowledge into elementary
propositions – declarations of individuals, classes and binary relations.
The properties of TELOS relevant for the purpose of this paper are similar to those of RDF, RDFS [27]. As RDF
(and OWL) are now on the way to becoming standards for the applications we target, we shall use here the
terminology of RDFS as it may be more familiar than that of TELOS, and talk about classes and properties. As
our primary interest is ontological, we intend to edit the CRM in various representations, but the primary source
7
for the CRM is a complete implementation in TELOS on the SIS knowledge management system [28]. Logical
assertions are omitted because they can be added in a later stage, once the ontological commitment of the
primitive classes, properties and isA relations are set up satisfactorily.
The process of instantiating the common model with factual knowledge can be broken into 2 steps: (1) The
creation of global identifiers for the instances of classes (individuals) taken from an interpretation of the source
data and their classification in the common model. “Global” is understood with respect to the declared scope of
the application. (2) The instantiation of the properties (roles, relationships) of the common model with relations
that connect those individuals and that are compatible with the intended meaning of the source data. The
mechanisms of creating the global identifiers themselves are out of the scope of the CRM work. Relevant for the
design of the ontology are the following properties we would like the represented factual knowledge to satisfy:
1. Context-free interpretation: The ontological commitment of each proposition should be interpretable
without any other contextual data. This is achieved on one side by the global identification of individuals;
on the other side it requires appropriate design of the ontology. E.g. an instance of a property
“creator_birth_date” with domain “man-made object” cannot be interpreted without another property
“creator”; a proposition “Martin.has role: buyer” cannot be understood without a sales event etc.. These
would be bad models. The advantage of context-free propositions as intermediate step for data
transformation and merging should be obvious: The global identifiers are the “fix points” around which
directly related information can be compiled without other processing.
2. Alternative views: The model should be able to capture multiple alternative propositions about any fact,
e.g. alternative birth dates for people whose actual birth date is debatable. This is mandatory for historical
data. The example demonstrates also that this a design principle: The birth event must be explicitly modeled
to render the intended meaning by context free propositions. The compilation of alternative propositions at
well-defined points is a great help for subsequent reasoning. One of the more expressive examples of
reasoning about historical contradictions is the Union List of Artist Names (ULAN) [23], which has tried to
consolidate life data of more than 100.000 artists by compiling all alternative data and expert opinions,
opinions on opinions etc.
3. Appropriate granularity: The model should make hidden concepts explicit to allow for extensions. If we
want to integrate documentation about works of an artist with a report about his birth, the usual properties
“birth_date”,”birth_place” are inappropriate. The hidden, intermediate concept “birth” should have been
made explicit beforehand. Under this view, instantiation of a property “birth_date” is no more an
elementary proposition. Indeed, as shown later, this notion of “elementary propositions” is not completely
application independent for property instances (binary relations).
4. An ontology should require the minimal ontological commitment sufficient to support the intended
knowledge sharing activities. This overlaps and competes with principle 2. and 3. above, however, we
strictly avoid underspecification for that purpose, like the Dublin Core concept of “resource”.
Finally it should be noted that virtually all metadata structures violate the above principles for reasons referred
to in section 2.1.
3.2 Monotonicity
From the epistemological side, the addition of knowledge that is not in contradiction to existing knowledge is
needed on both the categorical and the factual level, otherwise the integration of facts as they come in over time
becomes a non-scalable task. Maintaining monotonicity is an important practical consideration in the design of
ontologies, because the notion of what is in contradiction and what is not is grounded in the domain
conceptualisation. For our purposes, the Open World Assumption is mandatory, because our knowledge changes
and this must be taken into account. We view the impact of monotonicity on ontology design in three ways:
classification, attribution (properties), and modelling constructs. We regard all those changes in a conceptual
model as monotonic that do not invalidate previous instances of it.
3.2.1 Classification and Specialization
8
No complements: We have chosen to avoid defining any classes to be the complement of others, as such a class
would change meaning with each new subclass found. During all our work, we could not find fundamental
cultural concepts for which the complement is obvious. Even “male” is not clearly the complement of “female”
(there are hermaphrodites, sexless, what else?). We don’t know however, if this always holds. “Siblings” B
i,
B
j
are understood as not mutually exclusive, if not explicitly stated otherwise.
Preservation of classification: If an individual is once correctly classified according to a certain state of
knowledge, additional non-contradictory knowledge in the sense of the experts’ conceptualization should not
invalidate its instantiation of this class, but may add an additional class. The use of multiple instantiation, e.g. to
classify a willful destruction event with E7Intentional Activity and E6 Destruction, is essential to the CRM and
supports preservation of existing classifications. Staying in this example: An event may be first recognized as a
destruction. The willfulness of the event may be recognized at a later stage by other evidence. Or vice versa.
Intentional activity does not imply destruction, nor does destruction imply an intentional activity. The creation of
a class “Willful Destruction” does not offer any additional understanding.
An example of non-monotonic change of classification are the large Minoan terracotta vessels in Crete that
Evans took for bath tubs – due to their striking similarity with modern ones. After enough had been found with
bones in them, they were recognized as sarcophaguses. Had he classified them as container-like - the property he
could really recognize - the additional knowledge would not have invalidated the previous classification. This
argument is epistemological. It may come into conflict with ontological arguments, however we can design our
ontology in a general enough way to support it. The principle presented here has worked well in the way we
have used it, though obviously can not always hold sway.
3.2.2 Attribution
Whereas object-oriented design has provided us with an understanding of extension via specialization, the
extension of the granularity of attribution seems to be rarely regarded. By that we mean the replacement of one
property by a chain of properties and intermediate entities. The inverse operation, to reduce a path to a single
property corresponds to the join operation in Relational Algebra and is well-defined and well understood.
As pointed out in [2], this variable indirection or granularity of attribution is another major source of
incompatibility between semantically overlapping descriptions. Such property paths are potentially infinite. One
system may refer to the condition of an object as an assessment of the outcome of a number of measurements
carried out by a number of people over a period of time. A 'poorer' system may not even refer to the assessment
date and diagnosis, but simply register a term such as 'good' 'bad, or 'indifferent'. Such differences may be
entirely justified by the intended use of the information in a given context. We have encountered numerous
cases where radical differences in the granularity of information are justified by the intended purpose of the
documentation.
In such cases, the CIDOC CRM models two paths, a direct and an indirect one, and characterises the "poorer",
direct property as a short cut of the intermediate entity it “bypasses”. The resulting CRM model thus appears to
be redundant [2]. The idea is that collected factual knowledge would instantiate either the one or the other path.
In order to be monotonic, a model must foresee a disciplined way to increase the indirection in data paths
without losing the relationship to the coarser information. The intuitive short cut constructs introduced in the
interdisciplinary CIDOC working group should be formalized in the future. In particular we are as yet unsure
under which conditions reasoning described in section 4 is preserved by extending attribution paths.
3.2.3 Alternative Models
Finally, the monotonicity that can be achieved in practice may vary depending on modeling alternatives chosen.
Our practical experience has not yet given us much guidance here, however we can present some examples
instead:
Avoiding unconfirmed states: Many phenomena in history can be perceived as a chain of states and state
transition events. There are sound logical theories dealing with such systems. This view is the basis of the ABC
model [14], aiming like the CIDOC CRM to capture cultural contents. From an ontological point of view, the
transitions can be produced out of the descriptions of the states and vice-versa. From an epistemological point of
view, there is a huge difference: (1) If the information is incomplete, states and transitions cannot be
transformed into each other. (2) States are difficult to be observed. That a property was valid over an interval of
time and neither before nor after needs continuous complete observation. One can observe more easily a status,
i.e. the validity of some properties at a point in time, or a transition event.
9
Under these considerations, the CIDOC CRM gives preference to modeling e.g. ownership changes rather than
ownership states. It would result in a non-monotonic model to construct a set of states from any list of events, be
they directly observed or not, as in the examples given by [14], because information about additional events may
require deletion of existing states. The CIDOC CRM cannot claim to deal with the issue completely, mainly
because it tries to restrict itself to the semantics found in a definite set of data structures. We so far propose to
transform even a true (rare) observation of a state into transition events for normalization, which results in a
slight loss of information. Nevertheless, the issue of introducing more elaborate models of states is under
ongoing discussion.
View-neutrality: This principle has been described in detail in [2]. E.g. museums register accession
(acquisition) and deaccession events. A transfer from one museum to another is an accession event for the one
museum and a deaccession event for the other. Classification as “deaccession” or “accession” may be regarded
as non-monotonic, if one allows for the respective change of context. In the CIDOC CRM we replace these
notions by symmetric ones, like Acquisition, Change of Custody.
3.3 Global Coverage
When producing a standard, some attribute of validity is sought. With an extensible model in an open domain it
is a priori difficult to say what a model covers, and if it has reached any definable stage of maturity. The
approach proposed by Calvanese et.al [10] and others is open-ended. The enterprise model is incrementally
improved to comprise more and more source model semantics, and we have basically followed the same
procedure. In the process of taking more and more data structures into the scope of the CIDOC CRM however
we have observed that the upper level becomes very stable, and new data structures typically introduce
specializations “covered” by the model rather than “horizontal” extensions. This observation allows two things:
(1) The definition of a compatibility attribute; (2) the definition of a standard.
We have designed CIDOC-CRM as a common model that contains or covers the intended meaning of all data
structures used to encode “information required for the scientific documentation of cultural heritage
collections”, under certain semantic restrictions defined in [20]. For that purpose, the CRM group maintains a
list a representative data structures [20], for which the coverage will be identified, some of which are actually
contained in the CRM (e.g. [29], [30], [31]). With this claim, the CIDOC CRM is proposed as a standard
reference model for the description of cultural heritage collections including the necessary concepts to
communicate with library and archives contents. Note that by data structures we mean any database schema or
formal document structure, be it Relational, object-oriented, XML-DTD or even an RDF Schema, used to
describe primary data or metadata.
3.4 Designing a Manageable Unit
The creation of a standard ontology with limited resources in a reasonable time frame needs strict rules to
partition the total of work one could do into functionally complete and manageable units. Such restrictions have
been applied to (1) what meanings the contents should cover,(2) the modelling constructs and (3) the explicit
rules formulation. In 1997 we identified the following intellectual aspects suitable to restrict the ontology
without hampering its utility:
1. The conceptual framework (viewpoints) of the intended users: scholars, professionals in cultural heritage
management, educators.
2. The activities intended to be supported : scientific documentation, research and the exchange of information
with libraries and archives relevant to the documentation of cultural heritage collections.
3. The kinds of objects targeted at : objects in museums, libraries and archives.
4. The level of detail and precision required to provide an adequate level of quality of service.
5. Considerations of the necessary and manageable technical complexity.
By these criteria an intended scope has been formulated [20]. It excludes e.g. data only relevant for the internal
management of a museum and not relevant for the exchange of knowledge between organisations. Still, these
definitions are fairly fuzzy in practice. Therefore a practical scope is defined based on the semantics that can be
identified in a list of existing data structures and are necessary for their coherent interpretation. This list is
updated as the progress of work allows.
10
3.5 A property-driven design process
We have stressed properties (i.e. relations) in our methodology and ontology, and have required that, for the
most part, classes are required to be either the domain or range of some. This is motivated by the fact that
traditional museum data structures basically do the same. Fine granularity terminology is kept as variables in
data fields. It seems not to be as relevant to the propositions data structures render as the properties themselves.
This point may deserve further study.
Our property-centric approach has led to an empirical design process that we proposed to the CIDOC
Documentation Standards Working Group in 1996. It was based on modeling experience with semantic network
applications [5][38], and successfully applied to create the first version of the CIDOC CRM from the CIDOC
Relational Model. Since then it has been loosely followed by the CRM group, but it has not yet been verified by
independent groups. The driving force behind our process and ontology are the properties rather than the
classes. This is contrary to the well-established Booch, Rational Rose [35] and other o-o design methodologies,
yet we have found it far more productive for our purposes.
In the past, the CIDOC CRM had been extended several times. Some of these extensions are explicitly
documented [41]. During extension, more general domains or ranges were sometimes assigned to preexisting
properties. Such a change is monotonic as most of the last extensions have been. This observed behavior
confirms the utility of the presented methodology, in particular of seeking minimal domains and ranges within
the scope of the model.
4 About the CIDOC CRM contents
The CIDOC CRM contains classes and logical of groups of properties [43]. Those groups have to do with
notions of participation, parthood and structure, location, assessment and identification, purpose, motivation, use
etc. These properties have put Temporal Entities and with it events in a central place, as symbolically shown in
Fig.3.
Figure 3: A qualitative metaschema of the CIDOC CRM
All property paths to dates go through temporal entities. Property paths to places that bypass temporal entities
are understood as short cuts of temporal entities. Similarly, Actors are thought to relate to material and
immaterial things (Physical Stuff, Conceptual Objects) only via temporal entities. Any instance of a class may
be identified by Appellations, the names, labels, titles or whatever used in the historical context. We model the
relation to names and its ambiguity as part of the historical knowledge acquisition process. This should not be
confused with database identifiers in implementations of the Model, which are not part of the ontology. All class
participate in
Actors
Types
Conceptual Objects
Physical Stuff
Temporal Entities
A
p
p
e
l
l
a
t
i
o
n
s
affect or refer to
are referred / refine
r
e
f
e
r
t
o
/
i
d
e
n
t
i
f
y
have location
at
w
i
t
h
i
n
Places
Time-Spans
11
instances can be classified in more detail by Types, for the additional terminological distinction, as described
above. Frequently Types serve as the range of properties which refer in general to things of a certain kind, like
"a dress made for a wedding" in contrast to the "dress made for my wedding". We present here some prominent
logical groups of CRM properties.
4.1 Participation and Spatiotemporal Reasoning
As pointed out in [38], [2], [14], [44] and motivated by examples in this paper, the explicit modelling of events
leads to models of cultural contents which can be better integrated. The participation or presence of several
non-temporal entities in an event e1 allows for a most important conclusion: They have been in the same time-
interval and in the same space, even without knowledge of the particular time or space. They must have existed
at that time. They have not been somewhere else at that time (with electronic communication, the space volume
in which events occur can become very large, e.g. Earth to Moon). Culturally seen, the participants may have
influenced each other, or, in the case of people, exchanged information. The events e0
i
of creation of each
participant i have happened before or at the time of e1. The events e2
i
of destruction (or vanishing) of each
participant have happened after or at the time of e1. These are nothing more than the well known termini
postquem and termini antequem of chronological reasoning in historical research. Often this knowledge is more
reliable than sequencing based on explicit date information. Therefore we try carefully to preserve such
knowledge if it is primary (i.e. referred as such in a historical record or based on physical evidence).
The property P11 had participants denotes active or passive involvement of Actors, whereas P12 occurred in
the presence of ranges from objects just being there (e.g. a desk where a treaty was signed) to use of tools,
weapons, consumption of raw products, being produced. Specialization clarifies the more concrete senses
modelled in the CIDOC CRM. Table 1 shows the full subproperty hierarchies as indented lists, each dash
denoting another specialization level. By such generalization the normally implicit properties that enable
temporal ordering of events become explicit and can be used in rules independent from further extension of the
model.
Pid Property Name Domain Range
P11 had participants (participated in) E5 Event E39 Actor
P14 - carried out by (performed) E7 Activity E39 Actor
P22 - - transferred title to (acquired title of) E8 Acquisition E39 Actor
P23 - - transferred title from (surrendered title of) E8 Acquisition E39 Actor
P28 - - custody surrendered by (surrendered custody) E10 Transfer of Custody E39 Actor
P29 - - custody received by (received custody) E10 Transfer of Custody E39 Actor
P95 - - has formed (was formed by) E66 Formation E74 Group
P96 - by mother (gave birth) E67 Birth E21 Person
P98 - brought into life (was born) E67 Birth E21 Person
P99 - dissolved (was dissolved by) E68 Dissolution E74 Group
P100 - was death of (died in) E69 Death E21 Person
P12 occurred in the presence of (was present at) E5 Event E70 Stuff
P13 - destroyed (was destroyed by) E6 Destruction E19 Physical Object
P16 - used object (was used for) E7 Activity E19 Physical Object
P24 - transferred title of (changed ownership by) E8 Acquisition E19 Physical Object
P25 - moved (moved by) E9 Move E19 Physical Object
P30 - transferred custody of (custody changed by) E10 Transfer of Custody E19 Physical Object
P31 - has modified (was modified by) E11 Modification E24 Physical Man-Made Stuff
P108 - - has produced (was produced by) E12 Production E24 Physical Man-Made Stuff
P34 - concerned (was assessed by) E14 Condition Assessment E18 Physical Stuff
P36 - registered (was registered by) E15 Identifier Assignment E19 Physical Object
P39 - measured (was measured) E16 Measurement E18 Physical Stuff
P94 - has created (was created by) E65 Conceptual Creation E28 Conceptual Object
Table 1: The CIDOC CRM property hierarchies P11 and P12.
The next notion relevant in this context are the properties brought into existence, took out of existence limiting
the existence of things which have a persistent existence, i.e. which can be identified at different, separate times,
as in the sentence: “I have seen him again after two years”. These properties and their specializations connect
the world lines of things with their terminating events. Even those events can be useful for temporal reasoning
without explicit time: via participation of other things in the same event one can derive further termini. As we
perceive events as continuous processes with non-zero extent and infinitely divisible, we argue that each item
12
participates partially in its creation. Therefore the respective specializations like has created etc. appear in both
hierarchies:
Pid Property Name Domain Range
P92 brought into existence (was brought into existence
by)
E63 Beginning of
Existence
E77 Existence
P94 - has created (was created by) E65 Conceptual Creation E28 Conceptual Object
P95 - has formed (was formed by) E66 Formation E74 Group
P98 - brought into life (was born) E67 Birth E21 Person
P108 - has produced (was produced by) E12 Production E24 Physical Man-Made Stuff
P93 took out of existence (was taken out of existence by) E64 End of Existence E77 Existence
P13 - destroyed (was destroyed by) E6 Destruction E19 Physical Object
P99 - dissolved (was dissolved by) E68 Dissolution E74 Group
P100 - was death of (died in) E69 Death E21 Person
Table 2: The CIDOC CRM property hierarchies P92 and P93.
The properties in tables 1 and 2 characterize the semantics of data structures in the cultural area. Fig. 4 shows an
example of instantiating some of these properties, the legendary meeting of Pope Leo the Great with Attila the
Hun, in Mantua. Even if the three dates may be wrong, the 4 deductions are true if the meeting has happened at
all. Each death date constrains the meeting and both birth dates, the meeting date constrains both death and birth
dates, etc. A maximum life-span assumed, any date constrains all others. Note that the CRM does not recognize
points in time, only time-intervals. The deductions are not part of the model. They do not contribute to the
compilation and integration of the primary data. They can be done by any other system at any other time.
Pope Leo I Attila
Attila
meeting
Leo I
carried out by
(performed)
carried out by
(performed)
Birth of
Leo I
Birth of
Attila
Death of
Leo I
Death of
Attila
*
h
a
s
t
i
m
e
-
s
p
a
n
(
i
s
t
i
m
e
-
s
p
a
n
o
f
)
*
h
a
s
t
i
m
e
-
s
p
a
n
(
i
s
t
i
m
e
-
s
p
a
n
o
f
)
w
a
s
d
e
a
t
h
o
f
(d
i
e
d
i
n
)
w
a
s
d
e
a
t
h
o
f
(
d
i
e
d
i
n
)
b
r
o
u
g
h
t
i
n
t
o
l
i
f
e
(
w
a
s
b
o
r
n
)
b
r
o
u
g
h
t
i
n
t
o
l
i
f
e
(
w
a
s
b
o
r
n
)
*
has tim e-span
(is time- span of)
at most
within
at most
within
at most
within
AD453
AD461
AD452
before
before
before
before
Deduction:
before
Figure 4: Pope Leo I meeting Attila
Note that any extension of the Model with another property that implies participation, e.g. “was injured in”,
would not be captured by the above reasoning in some implementations, unless it is an explicitly declared
subproperty of P11. As subproperties are not supported by OMG models, it is not possible to implement such a
feature in a simple way that is not affected by extension. Note also that the preservation of such a reasoning
capability puts further constraints on “compatible extensions”, which need more exploration.
4.2 Properties of Locating
The question “where is it” can be answered in natural language by relation to two different kinds of entities:
geometric areas or objects. Examples of areas are: in France, in Athens, 39N 124E. Points given by spatial
coordinates are typically understood as the centre of a wider, extended area. Objects can be in the proper sense
(“bona fide objects”,[45]), as: on Queen Elizabeth (the ship), in my suitcase, at home, or they can be landscape
13
and other features (“fiat objects”,[45]), as: on mount St Helens, at the Rhine river. Following the CIDOC CRM,
geometric areas (E53 Place) can only be defined relative to larger objects, including the surface of earth. Those
objects in turn may be located at different times at different places (relative to a larger object). The cultural
interest is in the relation to other things and not to an abstract absolute space. Absolute coordinates seem to
make no sense when the reference objects move. As historical information is incomplete and sparse, and many
reference objects move, normalization of place information in cultural databases to absolute coordinates should
not replace the primary information, which is typically relative.
Any direct relation of an object to a place is seen as result of a move or a construction in situ, as with buildings.
This view is a result of a longer discussion. The notion “place” is ambiguous in English, and gives rise to
endless confusions in database design. In particular we take the position, that there is no image of a place, as it
is not a material entity.
Address
Place Name
Spatial Coordinates
Section Definition Physical Object
Place Appellation
Place
consists of
(forms part of)
defines section of
(has section definition)
i
s
l
o
c
a
t
e
d
o
n
o
r
w
i
t
h
i
n
(
h
a
s
s
e
c
t
i
o
n
)
i
d
e
n
t
i
f
i
e
s
(
i
s
i
d
e
n
t
i
f
i
e
d
b
y
)
h
a
s
f
o
r
m
e
r
o
r
c
u
r
r
e
n
t
l
o
c
a
t
i
o
n
(
i
s
f
o
r
m
e
r
o
r
c
ur
r
e
nt
l
o
c
a
t
i
o
n
o
f
)
Move
moved to
(occupied)
moved from
(vacated)
moved
(moved by)
Production
has produced
(was produced by)
happened at
(witnessed)
Figure 5: Properties of locating items
Places are identified by proper names or names referring to topological characteristics of object types, so-called
“segments” [46] or E46 Section Definition in the CRM, like bow, head, neck, bottom etc.
Addresses in general need not be places, their function is often that of a contact point for some person or
organisation (P76 has contact points (provides access to)), be they physical letterboxes or P.O. boxes. Fig. 5
shows the part of the CIDOC CRM dealing with location. The property P88 consists of (forms part of) from
Place to Place is the normal part-of relation for areas. There are no minimal nor maximal area elements.
4.3 Notions of influence
The knowledge of what influenced or motivated a human activity and in turn the persistent things that have
come upon us are culturally most relevant. We have not yet developed a systematic understanding of the
different forms of influence and their mutual relations. Some are more physical, like using a mould or a tool.
The influence of a mould on a produced object is strong and can often be verified on the object afterwards. The
influence of a hammer is less specific. Similarly, making a copy of a painting has a strong influence on the
product, copying the idea of a painting, a weak one. The latter is more an intellectual influence than a physical
one. Further, activities are influenced by other activities, like orders, or just by the emotions they raise. If a real
influence existed, a temporal sequence can be deduced. In contrast to “hard facts” as described in section 4.1,
the notions described here vary over a continuum of stronger and weaker influence, which can be verified more
or less easily afterwards. So far, the CRM contains the following properties of influence:
Pid Property Name Domain Range
P15 took into account (was taken into account by) E7 Activity E28 Conceptual Object
P33 - used specific technique (was used by) E11 Modification E29 Design or Procedure
P16 used object (was used for)
(mode of use : String)
E7 Activity E19 Physical Object
14
Pid Property Name Domain Range
P62 depicts object (is depicted by)
(mode of depiction : Type)
E24 Physical Man-Made Stuff E18 Physical Stuff
P63 depicts event (is depicted by)
(mode of depiction : Type)
E24 Physical Man-Made Stuff E5 Event
P65 shows visual item (is shown by) E24 Physical Man-Made Stuff E36 Visual Item
P67 refers to ( is referred to by)
(has type : Type)
E28 Conceptual Object E1 CRM Entity
P70 - documents (is documented in) E31 Document E1 CRM Entity
P17 was motivation for (motivated) E7 Activity E19 Physical Object
P18 motivated the creation of (was created for) E7 Activity E28 Conceptual Object
P20 had specific purpose (was purpose of) E7 Activity E7 Activity
The properties P15,P33,P16 describe plans, prototypes and physical tools (moulds, hammers etc.) that assisted
in or influenced an activity, and preexisted. These properties are used in particular in connection with
Modification, Production and Conceptual Creation to model not only the influence on the process but also on
the product, as with copies, prints etc.
The properties P62,P63,P65,P67,P70 describe an influence which can be manifested in the product without
knowledge of the process. They can be seen as short cuts of the respective activities. Intended depictions and
documentation of identifiable persons, objects, events, periods, ideas etc. play an extraordinary role in historical
studies. All range values of these properties must have existed before the respective process which manifested
them in the product.
The properties P17,P18,P20 describe an influence that originates in the activity itself, like orders, impressions,
or emotions. P20 in particular captures sequences of planned activities. E.g. in the sentence, "George of Kyriaze
orders a commemoration cross for donation to the Metropolitan Church of Ankara" [47], there is a had specific
purpose relation between the order and the donation. All these notions deserve deeper analysis. Only for P15-
P33 and P67-P70 could we establish subproperty hierarchies, an indication that the matter is relatively
unexplored. Nevertheless they can normally be objectified and play a basic role in historical (as well as
jurisdictional) reasoning. (at the time of publication of this paper, the properties of influence have been revised
and a final form was decided, see [50] and respective minutes of the Working Group on the CIDOC CRM home
page).
4.4 Applications
It would go beyond this paper to describe applications in detail. Several installations based on the CIDOC CRM
have already been made [48]. A recent test together with CIMI [49] aimed at demonstrating that the semantics
of heterogeneous museum records are preserved under the CIDOC CRM. Two examples were interesting: The
Clayton collection of the Natural History Museum in London and the Australian Museums On-line (AMOL)
initiative both use flat records in ACCESS databases. The Clayton collection describes a complex relation
between plant specimens, initial and current classification events and classification documents. These records
can be automatically transformed to CIDOC CRM instances because of the clear semantics of their fields. The
AMOL data were easily transformed to CIDOC CRM instances by hand, but not automatically, because their
fields are designed more for formatting the presentation. The examples demonstrate two things: data structures
(like the Clayton data) need not implement the complexity of an ontology for information integration in order to
be interpretable; however, an ontology can help to create interpretable data structures. More such tests will be
carried out in the near future.
5 Conclusion
In this paper we have presented an ontology for information integration in culture, and we have tried to justify
by the intended functionality a methodology and design. We assume that the applied methodology and the more
abstract levels of the model have a wider validity. The presented ontology is a result of ongoing work, and future
work will also address more advanced formalisations.
The CIDOC CRM has achieved a relatively high degree of maturity and completeness in capturing the
conceptualisations behind the data structures in its envisaged scope, as recent extensions of scope and data
transformation tests confirm. The purpose is information integration, but not the further reasoning like
15
reconstruction of a possible truth. It intends however to allow gathering all necessary information in a suitable
form for such further reasoning. It is sufficiently comprehensive for the domain expert, so that a broad
consensus on the correct ontological commitment could be achieved, and the ontology was accepted by the ISO
as a candidate international standard for cultural heritage information.
The methodology presented here has proven to be applicable in an interdisciplinary group, and our experience in
training non-experts in basic KR principles has been very encouraging. The complexity of the domain is
intriguing. Philosophical considerations and long discussions were necessary to clarify the role of the modelled
knowledge with respect to the working concepts of the domain experts. Without such clarifications, no
consensus on the relevant concepts could be achieved. This thinking was new for both sides, the computer
scientists and the domain experts, as it is not needed for either work in isolation. It was interesting to learn that
not all domain concepts are equally suited as basis for information integration.
The methodology presented here appears to be contrary to well-known o-o methodologies for designing the
controlling software of information systems. We wish to point out that there may be a qualitative difference,
even though some researchers take ontologies for software products (see for example the WebODE article in
this issue). Analysis of the semantics behind data structures for information integration is an ontological
problem. This paper tries to illustrate ontology from a point of view seldom taken: the relationships between
entities as a driving force for the logical structure, more than the nature of involved individuals. This seems to be
appropriate to analyse data structures in contrast to terminological systems. A coherent analysis of (non-unary)
properties is mandatory for information integration, even more than detailed entity analysis, in particular if one
separates the epistemological issue of correcting erroneous input data from the ontological issue of classifying
already correct information.
Historical knowledge, to our understanding independent of the specific domain, seems to reveal in this work a
character, which is quite distinct from engineering knowledge in a rather subtle way. Even though in our
conceptualisation of reality we do not distinguish between past, present and future, the way knowledge is
acquired, its quantity and quality, is completely different for the past. We argue therefore that the design of
conceptual models to capture the past must be governed far more by epistemological arguments than engineering
models. The nature of historical knowledge, the relation between reality, a perceived historical reality and the
form of knowledge we can acquire seems to be an interesting topic for further investigation.
The CIDOC CRM is envisaged to become an ISO standard in 2004. In parallel to the standardization work, we
intend to engage in more validation experiments and in research on the open theoretical and intellectual issues.
A general theory of extensibility for such an ontology under the preservation of certain reasoning capabilities
would be very helpful. As discussed in section 4.1, subproperties play a crucial role for that. From the point of
contents, the CIDOC CRM still touches only very fundamental concepts, and many extensions will be useful to
allow for more reasoning, like temporality of properties, phases of objects, a coherent model of influence,
modelling performing arts etc. We see also a need to clarify philosophical questions of foundational character
about the nature of the knowledge we describe.
Acknowledgement:
We wish to express our particular gratitude to Christopher Welty for the time he invested in editing this article,
substantially contributing to its current concise form and readability.
6 References
[1] T.Baker, "A Grammar of Dublin Core", D-Lib Magazine, 6(10), 2000.
[2] Martin Doerr and Nicholas Crofts "Electronic Esperanto: The Role of the Object Oriented CIDOC Reference Model",
Proc. of the ICHIM'99, Washington, DC, September 22-26, 1999.
[3] The CIDOC CRM Home Page, 2001, http://cidoc.ics.forth.gr/what_is_crm.html
[4] Nick Crofts, Ifigenia Dionissiadou, Martin Doerr, Matthew Stiff, "Definition of the CIDOC object-oriented Conceptual
Reference Model", Version 3.2.1, ISO Working Document ISO/TC46/SC4/WG9/N2, July 2001,
http://cidoc.ics.forth.gr/docs/cidoc_crm_version_3.2.1.rtf
16
[5] I.Dionissiadou, M.Doerr, "Mapping of material culture to a semantic network", in : Automating Museums in the
Americas and Beyond, Sourcebook, ICOM-MCN Joint Annual Meeting, August 28-September 3, 1994
[6] Panos Constantopoulos, "Cultural Documentation: The CLIO System", Technical Report FORTH-ICS/TR-115, 12 pages,
January 1994.
[7] Martin Doerr, "CIDOC Conceptual Reference Model, Correlation Test Project , Results", June 2001,
http://cidoc.ics.forth.gr/testproject_results.html
[8] S. Bergamaschi, S. Castano, S. De Capitani De Vimercati, S. Montanari and M. Vincini, "An Intelligent Approach to
Information Integration", International Conference on Formal Ontology in Information Systems (FOIS98), Trento 1998.
IOS-Press (Amsterdam)
[9] Foundations of Data Warehouse Quality (DWQ) European ESPRIT IV Long Term Research (LTR) Project 22469, 1996-
1999, http://www.dbnet.ece.ntua.gr/~dwq/
[10] D. Calvanese, G. De Giacomo, M. Lenzerini, D. Nardi, and R. Rosati, "Description Logic Framework for Information
Integration";, In Proc. of the 6th Int. Conf. on the Principles of Knowledge Representation and Reasoning (KR'98), 1998,
pages 2-13
[11] Gio Wiederhold, “Mediators in the Architecture of Future Information Systems”, in : IEEE Computer, March 1992.
[12] Guarino N. "Formal Ontology and Information Systems". In N. Guarino (ed.), Formal Ontology in Information Systems.
Proc. of the 1st International Conference, Trento, Italy, 6-8 June 1998. IOS Press
[13] Thomas R. Gruber, "Toward Principles for the Design of Ontologies Used for Knowledge Sharing", in: Formal
Ontology in Conceptual Analysis and Knowledge Representation, edited by Nicola Guarino, Roberto Poli, Kluwer, 1994
[14] Carl Lagoze, Jane Hunter, "The ABC Ontology and Model", DC-2001, International Conference on Dublin Core and
Metadata, Tokyo, October 2001, http://metadata.net/harmony/lagoze_hunter_dc2001.pdf
[15] The Art Museum Image Consortium AMICO, http://amico.org, AMICO data dictionary version 1.3,
http://amico.org/AMICOlibrary/dataDictionary.html
[16] Chryssoula Bekiari, Christina Gritzapi, Dimitrios Kalomoirakis "POLEMON: A Federated Database Management
System for the Documentation, Management and Promotion of Cultural Heritage", Proc. of the 26
th
Conference on
Computer Applications in Archaeology, Barchelona, March 24-28, 1998.
[17] Chryssoula Bekiari, Panos Constantopoulos & Theodosia Bitzou " DELTOS : A Documentation System for the
Antiquities and Preserved Buildings of Crete, Requirements Analysis", Technical Report FORTH-ICS/TR-60, October
1992. Available in Greek.
[18] "SPECTRUM: The UK Museum Documentation Standard,", second Edition, Museum Documentation Association
mda(Europe), Cambridge, United Kingdom,1997-1998, http://www.mda.org.uk/spectrum.htm
[19] "International Guidelines for Museum Object Information: The CIDOC Information Categories", published by CIDOC
in June 1995. http://www.cidoc.icom.org/guide/guide.htm
[20] Nicholas Crofts et.al, "CRM Scope Definition", Proposal of the Steering Committee of the CIDOC CRM SIG, July 7,
2001. http://cidoc.ics.forth.gr/crm_scope_definition.html
[21] Nicholas Crofts et.al., "Notes on the data modelling meeting in Crete July 1997", July 1997,
http://cidoc.ics.forth.gr/docs/notes_data_modelling_1997_crete.doc
[22] Fernández, M. "Overview of Methodologies for Building Ontologies". Workshop on Ontologies and Problem-Solving
Methods: Lessons Learned and Future Trends. (IJCAI99). August. 1999.
[23] J.M.Bower,M.Baca et al., "Union List of Artist Names - A User's Guide to the Authority Reference Tool", Version 1.0,
Getty Art Information Program, G.K.Hall, New York, 1994
[24] N. Guarino,P. Giaretta, "Ontologies and Knowledge Bases, Towards a Terminological Clarification", in N.J.I.Mars
(ed.), Towards Very Large Knowledge Bases, IOS Press 1995, Amsterdam.
[25] J. Mylopoulos, A. Borgida, M. Jarke, M. Koubarakis, "Telos: Representing Knowledge about Information Systems",
ACM Transactions on Information Systems, October 1990.
17
[26] Anastasia Analyti, Nicolas Spyratos & Panos Constantopoulos, "On the Semantics of a Semantic Network",
Fundamenta Informaticae 36 (1998), pp. 109-144, IOS Press.
[27] G. Karvounarakis, V. Christophides, D. Plexousakis, S.Alexaki, "Querying Querying RDF Descriptions for Community
Web Portals". In Proc. of the 17 France National Conf. on Databases BDA, 29 Octobre - 2 Novembre 2001, Agadir, Maroc.,
http://139.91.183.30:9090/RDF/publications/sigmod2000.html
[28] ICS FORTH, Information Systems Laboratory, Heraklion, Crete, Greece, "The Semantic Index System - SIS",
http://www.ics.forth.gr/proj/isst/Systems/sis.html
[29] M. Theodoridou, M. Doerr, Mapping of the Encoded Archival Description DTD Element Set to the CIDOC CRM ,
Technical Report FORTH-ICS/TR-289, June 2001, http://www.ics.forth.gr/proj/isst/Publications/paperlink/ead.pdf
[30] Martin Doerr, Mapping of the AMICO data dictionary to the CIDOC CRM , Technical Report FORTH-
ICS/TR-288, June 2001, http://cidoc.ics.forth.gr/docs/mappingamicotocrm.rtf
[31] Martin Doerr, Mapping of the Dublin Core Metadata Element Set to the CIDOC CRM , Technical
Report FORTH-ICS/TR-274, July 2000, http://cidoc.ics.forth.gr/docs/dc_to_crm_mapping.rtf
[32] S.R. Ranganathan, "A descriptive account of Colon Classification", Bangalore: Sarada Ranganathan Endowment for
Library Science. 1965.
[33] Martin Doerr et.al., "Notes on the transformation of the CIDOC relational data model", July 1996,
http://cidoc.ics.forth.gr/docs/notes_trans_cidoc.rtf
.
[34] Paul Feyerabend, "Three Dialogues on Knowledge", Basil Blackwell 1991
[35] T. Quatrani, Visual Modeling with Rational Rose and UML, Addison-Wesley,1998.
[36] John L. Schnase, "Semantic Data Modelling of Hypermedia Associations", in: ACM Transactions on Information
Systems, Vol.11,No.1, January 1993, p 45
[37] Michael Erdmann, Rudi Studer, “Ontologies as Conceptual Models for XML Documents”, research report, Institute
AIFB, University of Karlsruhe, 1999.
[38] Maria Christoforaki, Panos Constantopoulos, Martin Doerr, "Modelling occurences in cultural documentation",
Proc. of the III Convegno Internazionale di Archeologia e Informatica, Roma, November 22-25, 1995.
http://www.ics.forth.gr/proj/isst/Publications/paperlink/Model_occur_in_cultur_doc.ps.gz
[39] B. Tversky, Kathleen Hemenway, “Objects, Parts, and Categories”, in Journal of Experimental Psychology: General,
Vol.113,No2, June 1984, pp169-193.
[40] "EAD Tag Library for Version 1.0, Encoded Archival Description (EAD)Document Type Definition (DTD)", Version
1.0, Technical Document No. 2, June 1998. Published by the Society of American Archivists and theLibrary of Congress
(http://lcweb.loc.gov/ead/tglib/tlhome.html
)
[41] Martin Doerr (ed.), “Agios Pavlos Extensions - Add-ons for the Completion of the CIDOC CRM”, July 2000,
http://cidoc.ics.forth.gr/docs/agios_pavlos_extensions.rtf
[42] Nick Crofts, Ifigenia Dionissiadou, Martin Doerr, Matthew Stiff (ed.), “Definition of the CIDOC object-oriented
Conceptual Reference Model”, Version 3.2, July 2001, ISO/TC46/SC4/WG9/3,
http://cidoc.ics.forth.gr/docs/cidoc_crm_version_3.2.rtf
[43] Nick Crofts, Ifigenia Dionissiadou, Martin Doerr, Pat Reed (editors), "CIDOC Conceptual Reference Model -
Information groups", ICOM/CIDOC Documentation Standards Group, September 1998,
http://cidoc.ics.forth.gr/docs/info_groups.rtf
[44] INDECS Home Page: Interoperability of Data in E-Commerce Systems, http://www.indecs.org
[45] B. Smith, A. Varzi, “Fiat and Bona Fide Boundaries: Towards on Ontology of Spatially Extended Objects.”
International Conference COSIT'97, October 15-18, 1997, Proceedings. Lecture Notes in Computer Science, Vol. 1329,
Springer, 1997, pp103-119
[46] P.Gerstl, S.Pribbenow, “A conceptual theory of part – whole relations and its applications”, Data & Knowledge
Engineering 20 305-322, 1996, North Holland- Elsevier
18
[47] M. Doerr, I. Dionissiadou, “Data Example of the CIDOC Reference Model - Epitaphios GE34604 –”
October 2, 1998, http://cidoc.ics.forth.gr/docs/crm_example_1.doc
[48] N. Crofts "Implementing the CIDOC CRM with a relational database" in MCN Spectra. 24 (1) Spring, 1999
[49] John Perkins, "ABC/Harmony CIMI Collaboration Project", September 30th, 2000,
http://www.cimi.org/public_docs/Harmony_long_desc.html
[50] Nick Crofts, Martin Doerr, Tony Gill, Stephen Stead, Matthew Stiff (editors), Definition of the CIDOC object-
oriented Conceptual Reference Model, Version 3.4 , November 2002
http://zeus.ics.forth.gr/cidoc/docs/cidoc_crm_version_3.4.rtf
