® InfoJur.ccj.ufsc.br
Ontologies in the Design of Legal
Knowledge Systems; Towards a Library of Legal Domain Ontologies
Pepijn R.S. Visser and Trevor J.M. Bench-Capon LIAL - Legal
Informatics at Liverpool University of Liverpool - Department of Computer Science PO Box
147, Liverpool, L69 7ZF, United Kingdom 0151 - 794 3792 {pepijn, tbc}@csc.liv.ac.uk
http://www.csc.liv.ac.uk/~lial/ Abstract
Legal ontologies are useful in the design of knowledge systems because
they are reusable. A library of such ontologies could greatly enhance the development of
legal knowledge systems. In this article we address the creation of such a library. In
particular, we discuss four legal ontologies and investigate how the ontologies can be
indexed and represented in a library.
1. Introduction
What are the building blocks of legal knowledge? Legal philosophers have pondered over
this question for a long time. Research in Artificial Intelligence and Law has also run up
against this question, albeit from a different angle. Arguably, the creation of a legal
knowledge representation requires a conceptualisation of the building blocks of legal
knowledge. Eventually, these building blocks will form the basis for operational legal
knowledge systems.
An important reason for producing ontologies is that they form reusable building blocks
for the design of (legal) knowledge systems. Many development methodologies for knowledge
systems, such as CommonKADS (Breuker and Van de Velde, 1994), recognise the role of
ontologies. Although literature is showing efforts to create ontology libraries (Farquhar et
al., 1997) there are no such efforts in the legal domain. In this article we address
the creation of libraries of legal (domain) ontologies. Crucial to the success of such a
library is that its users are aware of what the library contains. This is less trivial
than it seems. For books there are well established methods of classification, typically
based on subject, and the physical organisation of the library uses this. Since the title,
author and subject of a book provides a reasonably good idea of the contents, users can
browse the library on these items. The problems with an legal ontology library are
different: while there will be far fewer items in the library, there are no well
understood principles of organisation. In this article we suggest some features of, and
relations between, ontologies, which could serve as the basis for structuring the contents
of a library of legal ontologies.
This article is structured as follows. In section 2 we briefly discuss four legal
ontologies which can be used in the method. In section 3 we address the development of a
library of legal ontologies and in particular we discuss the indexing mechanisms of the
library. In section 4 we draw conclusions.
2. Four Legal Ontologies
Ontologies are reusable building blocks for knowledge system design (Visser et al.,
1997). However, AI and Law research so far has not produced very many legal ontologies for
this purpose. Only a small number of explicit conceptualisations of the legal domain is
available (for an overview on legal ontologies for system design, see Visser and Winkels,
1997). Below, we discuss four well known legal ontologies that could be used in the design
of a LKS: (a) McCarty's LLD, (b) Stamper's NORMA, (c) Valente's
Functional Ontology of Law and (d) the Frame-based Ontology of Van Kralingen and
Visser.
(a) LLD
McCarty (1989) has proposed a language for legal discourse (lld). He considered the
language to be a first step towards a general applicable representation language for legal
knowledge. Although lld itself is a representational language and not an ontology it
clearly reveals a generic conceptualisation of the legal domain. The basic components of
lld are atomic formulae and rules. Atomic formulae are merely predicate relations
used to express factual assertions. A distinction is made between count terms (to
express tangible objects, such as houses, and persons) and mass terms (to express
intangible objects, such as cash, and stock). Rules are formed by connecting atomic
formulae with logical connectives. They have a left-hand side which is an atomic formula,
and a right-hand side which is a compound expression. The compound expression determines
the type of rule involved. There are five types of rules: (1) horn clauses, (2) 'horn
clauses' with embedded implications, (3) 'horn clauses' with embedded negations, (4)
default rules, and (5) prototype-and-deformations. Together atomic formulae and rules
allow the creation first-order expressions. Modalities are stated as second-order
expressions. The following modalities are supported: time, events and actions, and deontic
expressions (McCarty, 1989). To express temporal statements lld recognises states. A state
essentially is the (temporal) reification of a predicate relation. Predicate relations can
be reified both with points of time, as well as with intervals (two points of time).
Changes in states are realised by events. Events are either elementary (viz. a
state-change) or complex (viz. elementary events connected by the operations of
disjunction, sequential and parallel composition, and universal and existential
quantification applied to the elementary events). An action is the relation between an
actor and an event. With regards to deontic statements, lld supports four modal operators:
permitted (P), forbidden (F), obligatory (O), and enabled (E). Modalities, such as time
and permissions, are stated as second-order expressions. For more details on lld we refer
to McCarty (1989; 1993).
(b) NORMA
Stamper has criticised the use of traditional logics for the representation of (legal)
knowledge because they suffer from some important semantic problems (Stamper, 1991).
Briefly stated, traditional logics rely on symbolic representations that have only a very
weak connection to the real-world concepts they intend to denote. In particular, symbolic
representations rely (according to Stamper - invalid) on notions such as truth,
individuality, and identity. Accordingly, expressing legal knowledge in the form of rules
is an over simplification of what legal knowledge is about. To overcome these problems
Stamper argues that there is need to escape from the frame of reference of classical logic
(Stamper, 1991, p.229). Building on his LEGOL work (see Stamper, 1980), he proposed the
norma formalism (Stamper, 1991). norma, which means 'logic of norms and affordances', is
based on two main philosophical assumptions: (1) there is no knowledge without a knower,
and (2) the knowledge of a knower depends on his behaviour (Stamper, 1996). Using norma
(henceforth: nor) the entities in the world are described by their behaviour rather than
by assigning them an individuality and truth values. The main ontological concepts are (a)
agents, (b) behavioural invariants, and (c) realisations. An agent is
an organism standing at the centre of reality. It gains knowledge, regulates, and modifies
the world by means of actions. For its actions the agent takes responsibility. The concept
of an agent (a) can be extended to include groups, teams, companies, social agents
or even nation states. A behavioural invariant (b) is a description (e.g.,
using verbs, nouns, or adjectives) of a 'situation' whose features remain invariant. Here,
a situation loosely denotes some knowledge of the world, such as an object (e.g., a
cup, a piano) or a state of affairs (e.g., walking, paying). The realisation of a
situation - a realisation (c)- is specified as the combination of (1) an agent and
(2) a behavioural invariant, shortly written as Ax (the situation, denoted by
behavioural invariant x, that is realised by agent A). An example of
a realisation Ax is John walks. Different kinds of realisations are
recognised, for instance, Ax* (denoting the ability of A to realise x),
Ax@ (denoting the authority of A to realise x), Ax+ (denoting A
starts to realise x), Ax- (denoting A finishes the realisation of
x), and Ax# (denotes that x can be divided into individuals, cf.
classes and objects). By combining behavioural invariants composite realisations can be
made. We here mention the most important composite realisations: Axy (denoting that
A cannot realise y without first realising x), A.x.y (denoting
that x is a part of A and y is a part of x), A(x while y),
A(x orwhile y), A(x whilenot y), A(x whenever y) (denotes that x
is realised whenever y is realised); A(x then y) (denotes that if x
is realised then y is be realised), A "Bx" (denoting that an agent A
can tell another agent B to bring about x, for instance by commanding or
suggesting), and A(a:b:c) d) (denoting that a, b, and c are
instances of d). For a more extensive discussion we refer to Stamper (1991; 1996)(1).
(c) The Functional Ontology of Law
Valente's ontology of law (1995) - LFU - is based on a functional
perspective of the legal system. The legal system is considered an instrument to change or
influence society in specific directions, determined by social goals. Its main function is
reacting to social behaviour. This main function can be decomposed into six primitive
functions, each corresponding with a category of primitive legal knowledge in lfu.
Accordingly, lfu distinguishes six categories of legal knowledge: (a) normative knowledge,
(b) world knowledge, (c) responsibility knowledge, (d) reactive knowledge, (e) meta-legal
knowledge, and (f) creative knowledge.
Normative knowledge (a) is characterised as knowledge that
defines a standard of social behaviour. It thereby prescribes behaviour of the people in
society. The standard is defined by issuing individual norms, expressing what ought to be
the case. In lfu world knowledge is legal knowledge that describes the world that
is being regulated. It delineates the possible behaviour of (people, and institutions) in
society, and thereby it provides a framework to define what behaviour ought (and ought
not) to be performed. It can be considered to be an interface between the commonsense
knowledge of people in society and the normative knowledge. Within the world knowledge
Valente distinguishes (b.1) definitional knowledge, and (b.2) causal knowledge. The
definitional knowledge is the static part, it consists of definitions of (b.1.1) legal
concepts (e.g., agents, objects), (b.1.2) legal relations (e.g., legal
qualifications of actions), (b.1.3) a case (viz. the problem case under
investigation), (b.1.4) circumstances (viz. the grounded facts, or, building blocks
of a case), (b.1.5) generic cases (viz. typical generic legal cases), and (b.1.6)
conditions (viz. the building blocks of the generic legal cases). Together these
constructs provide a vocabulary which can be used to describe the relevant aspects of the
world under a specific perspective taken by the legislator. The causal knowledge (b.2) is
the dynamic part, describing the behaviour of people in society in terms of the
definitional knowledge. Responsibility knowledge (c) is legal knowledge that either
extends (assigns), or restricts the responsibility of an agent for its behaviour. Its
function is to provide the legal means to reject the common idea that someone is only
responsible for what one causes. Reactive knowledge (d) is legal knowledge that
specifies which reaction should be taken (and how) if an agent violates a primary norm. Meta-legal
knowledge (e) is legal knowledge about legal knowledge, or, legal knowledge that
refers to other legal knowledge. There are four sub categories of meta-legal knowledge:
(e1) norm data, (e2) ordering norms, (e3) normative default, and (e4) validity knowledge.
Norm data (e1) includes information about norms, such as their scope of application, their
type, their place in the norm hierarchy, their power origin, their promulgation, and the
norm goal. Ordering norms (e2) are norms that determine how to solve conflicts. Creative
knowledge is legal knowledge that allows the creation of previously non-existent legal
entities. For more details on this ontology we refer to Valente (1995).
(d) FBO: Van Kralingen and Visser's Frame-based Ontology
Van Kralingen (1995) and Visser (1995) argue that robust (conceptual
and formal) ontologies of the legal domain are necessities for reducing the
task-dependency of legal knowledge specifications. Although there are some minor
differences between the (conceptual) ontology as defined by Van Kralingen, and the
(formal) ontology as defined by Visser, their similarities allow us to treat them as one
ontology. The main ontological distinction in fbo concerns the generic legal ontology and
the statute-specific ontology. The distinction is based on the observation that some parts
of an ontology are reusable across different legal subdomains.
The generic legal ontology (GLO), in contrast to the
statute-specific ontology, is the generic and reusable part of the ontology. It divides
legal knowledge over three distinct entities: norms, acts and concept descriptions. For
each of these entities the ontology defines a template (also referred to as frame
structure) that lists all attributes relevant for the entity. Norms are the general
rules, standards and principles of behaviour that subjects of law are enjoined to comply
with. In the ontology a norm comprises the following eight elements: (1) a norm identifier
(used as a point of reference for the norm), (2) a norm type (either norm of conduct or
norm of competence), (3) a promulgation (the source of the norm), (4) the scope (the range
of application of the norm), (5) the conditions of application (the circumstances under
which the norm is applicable), (6) the norm subject (the person or persons to whom the
norm is addressed), (7) the legal modality (either ought, ought not, may or can), and (8)
the act identifier (used as a reference to a separate act description). Acts
represent the dynamic aspects which effect changes in the state of the world. Within the
category of acts two distinctions are made. The first distinction is between events
and processes. Events represent an instantaneous change between two states, while
processes have duration. The second distinction is between institutional acts and physical
acts. The former type of acts are considered legal (institutional) versions of the
(physical) acts that occur in the real world (more precisely: an institutional act is a
legal qualification of a physical act). Note that these two distinctions result in four
different types of acts. All acts are assumed to have the following thirteen elements: (1)
the act identifier (used as a point of reference for the act), (2) a promulgation (the
source of the act description), (3) the scope (the range of application of the act
description), (4) the agent (an individual, a set of individuals, an aggregate or a
conglomerate), (5) the act type (both basic acts, and acts that have been specified
elsewhere can be used), (6) the modality of means (material objects used in the act or sub
acts; e.g., a gun), (7) the modality of manner (the way in which objects have been
used or sub acts have been performed) (e.g., aggressively), (8) the temporal
aspects (an absolute time specification; e.g., on the first of August, on Sundays,
at night, etc, but not: during a fire, after the King dies, etc), (9) the spatial aspects
(a specification of the location where the act takes place; e.g., in the
Netherlands, in Leiden, on a train), (9) the circumstantial aspects (a description of the
circumstances under which the act takes place; e.g., during a war), (10) the cause
of the action (a specification of the reason(s) to perform the action, e.g.,
revenge), (11) the aim of the action (the goal visualised by the agent; e.g., with
a view to unlawfully appropriate an object), (12) the intentionality of an action (the
state of mind of the agent; e.g., voluntary), and (13) the final state (the results
and consequences of an action; e.g., the death of the victim). Concept
descriptions deal with the meanings of the concepts found in the domain. They may be
definitions or deeming provisions and can be used to determine definitively the meaning of
a notion, either by, as in the case of the former, providing necessary and sufficient
conditions, or, as in the case of the latter, establishing a legal fiction. Another type
of concept is the factor, which may either establish a sufficient condition, or indicate
some contribution to the applicability of the concept, as discussed above. Finally there
are meta concepts which are provisions governing the application of other provisions.
Concept descriptions comprise the following seven elements: (1) the concept to be
described, (2) the concept type (definition, deeming provision, factor, or meta), (3) the
priority (the weight assigned to a factor), (4) the promulgation (the source of the
concept description), (5) the scope (the range of application of the concept description),
(6) the conditions under which a concept is applicable, and (7) an enumeration of
instances of the concept.
The statute-specific ontology consists of predicate relations
that are used to complement the terminology for norms, acts and concept descriptions. The
statute-specific ontology cannot be reused for other legal subdomains, and should always
be created for each legal sub domain under consideration. The statute-specific ontology
states the vocabulary with which the knowledge base is constructed. A more elaborate
discussion of the legal and the statute-specific ontology can be found in Van Kralingen
(1995) and Visser (1995), which gives a statute-specific ontology for the Dutch
Unemployment Benefits Act (DUBA). An ontolingua specification of the legal ontology is
given by Visser and Bench-Capon (1996).
3. The Ontology Library
The ontologies discussed in the previous section differ substantially
in the way they conceptualise the legal domain even though all four ontologies are
intended to support the construction of legal knowledge or information systems (LKS).
Visser and Bench-Capon (1998) - after a comparison of these ontologies - argue that no
ontology is necessarily better than another. Which ontology is most adequate depends on
the specific application being developed. The indexing of the ontologies in our library
should allow for the selection of the ontology that best suits the intended application.
Otherwise stated, the ontology indexing mechanism should reveal the differences between
the ontologies. In this section we address the structure and indexing mechanism of the
legal ontology library. In section 3.1 we discuss related work on ontology indexing
schemes. In section 3.2 we address ontology relations between ontologies in the library.
Then, in section 3.3 we propose a set of questions that can be used to index the library.
3.1 Ontology Indexing Schemes: Related Work
Any library indexing mechanism is based on distinguishable features of
ontologies. Indexing mechanisms differ in the features that are distinguished and in the
way they are expressed. To create a library of legal ontologies, and thus, to determine
what ontology features we will distinguish we first make an inventory of related work on
this matter. Below, we list the ontology features distinguished in five different research
efforts.
Gruber (1995) formulates five design criteria for ontologies in the
context of knowledge sharing and interoperation among programs. The ontology features he
distinguishes are: (1) clarity, (2) coherence, (3) extendibility, (4)
encoding bias, and (5) ontological commitment.
Visser and Bench-Capon (1998) build on the work of Gruber and others
and present a taxonomic structure of ontology-comparison criteria. The ontology features
they distinguish are: (1) epistemological adequacy (1.1 epistemological clarity;
1.2 epistemological intuitiveness; 1.3 epistemological relevance; 1.4 epistemological
completeness; 1.5 discriminative power), (2) operationality (2.1 encoding
bias; 2.2 coherence; 2.3 computationality), (3) reusability (3.1 task-and-method
reusability; 3.2 domain reusability).
Farquhar et al. (1997) describe the Ontolingua Ontology Library.
They distinguish four type of ontology relations (1) inclusion, (2) restriction,
(3) polymorphic refinement and (4) circularity. Currently, the indexing of
ontologies in the library is done via full text and context-sensitive search facilities.
Also, the ontology library has an inclusion lattice which shows the inclusion relations
between the different ontologies.
Fridman-Noy and Hafner (1997) have compared 10 large ontology projects.
In their survey they used the following (groups of) features to distinguish the
ontologies: (1) general, (2) design process, (3) taxonomy, (4) internal
concept structure and relations between concepts, (5) axioms, (6) inference
mechanism, (7) applications and (8) contributions.
Van Heijst et al (1997) address the issues in creating an
ontology library. Ontologies are indexed using dimensions task / method dependency and
domain dependency. These dimensions are used to partition the library into two
regions: a core library and a peripheral library. The former contains
ontologies that are generic with respect to task / method and domain, and the latter
contains ontologies that are both domain and task / method specific. Other library
construction issues are the language in which ontologies are stated (inclusive
whether it supports higher-order expressions), the modularity of the
ontologies as building reusable blocks, and the option to allow alternative (and
possibly inconsistent) ontologies.
It should be noted that none of the features mentioned above is
currently measurable in a unique way. Scaling the ontologies on ontology features remains
a subjective task (Visser and Bench-Capon, 1998). Some of the features are not directly
usable for our ontology library. For instance, Gruber's 'minimal ontological commitment'
criterion. In the context of a library this criteria is not relevant, as is distinguishing
the ontological-commitment feature of ontologies. We note that there seem to be two broad
categories of features, namely (1) intra-ontology features, and (2) inter-ontology
features. Gruber, Fridman & Hafner, Van Heijst et al., and Visser and
Bench-Capon seem to emphasise the former type of ontology features, Farquhar et al.
seem to emphasise the latter type of features. Here, we deem both type of features to be
useful for our ontology indexing mechanism. Since the intra-ontology features have been
addressed extensively in the work mentioned above, in the next section we concentrate on
the inter-ontology features.
3.2 Inter-Ontology Features
One approach to find ontologies in the ontology library is to exploit
the relations between different ontologies in the library. One ontology might, for
instance, be an extension of another ontology in the library (this is the case with the statute-specific
ontology of Van Kralingen and Visser, which is an extension of the generic legal
ontology). In this section we address how ontologies can be characterised by their
relation to other ontologies. In particular, we look into the question (I) how component
ontologies can be identified and (II) what relations can exist between them(2).
(I) identify component ontologies
We assumed that an ontology consists of a set of definitions of
classes, relations, instances, functions, and axioms. Identifying component ontologies
implies isolating multiple sets of ontology definitions. Below, we list five principles
that can be used to define multiple, but related, ontologies.
1. Domain partitioning: An obvious way to identify multiple ontologies is to partition the domain itself in logical units, each representing another fragment of the domain. The Ontolingua ontology library, for instance, holds a collection ontologies for sets, numbers, lists, etc. (Farquhar et al., 1997).
2. Alternative domain views: Multiple ontologies can be created by allowing different - possibly inconsistent - views of the same fragment of the domain (Van Heijst et al., 1997). This principle covers the polymorphic refinement operator as defined by Farquhar et al. (1997, p.713).
3. Abstraction: Multiple ontologies can be defined by allowing both abstract and detailed ontologies. For instance, one can define generic legal concepts (e.g., norm, modality, definition, act, law) and some statute-specific legal concepts (e.g., definition-of-employee, forbidden, kill, penal-law). A top-level ontology (Sowa, 1995; Guarino, 1997a) is an example of an abstract ontology.
4. Primary ontologies versus Secondary ontologies: Multiple ontologies can be defined by allowing both primary ontologies and secondary ontologies. The distinction being that a primary ontology defines the basic concepts and relations in a domain (e.g., apple, pear) and the secondary ontology adds a dimension to the concepts and relations (e.g., rotten, red) by distinguishing additional features (e.g., Borst et al., 1996).
5. Terminological, Information and Knowledge modelling
ontologies: The difference between these three types of ontologies is defined by 'the
amount and type of structure in an ontology' (Van Heijst, 1995; Van Heijst et al., 1997).
Terminological ontologies define a lexicon, Information ontologies define record
structures of databases, and knowledge modelling ontologies specify knowledge (with a
richer internal structure than information ontologies)(3).
(II) defining ontology relations
The principles above enable us to define multiple sets of ontology
definitions. We have not yet addressed how these ontology definitions are related. Below,
we list three kinds of relations that might be defined between component ontologies. The
kind of relations that can be defined between component ontologies is greatly influenced
by the way these component ontologies are identified.
a. Subset / Superset relation. Ontology O1 is a subset of ontology O2 if all definitions in O1 are contained in O2 (O2 is the superset of O1). Strictly speaking, this kind of relation cannot occur with any of the above mentioned principles since all principles create ontologies without overlap (a primary and a secondary ontology, for instance, do not overlap). If we relax the definitions of the partitioning principles so as to allow overlap between the ontologies we can use this relation in combination with principles domain partitioning, abstraction, primary versus secondary ontologies, and terminological / information / knowledge modelling ontologies.
b. Extension Relation. Ontology O2 is an extension of ontology O1 if all definitions in ontology O1 are available in ontology O2. The difference with subset /superset relations is that the definitions contained in O1 are themselves not contained in ontology O2. This kind of relation can be used in combination with domain partitioning, abstraction, primary versus secondary ontologies, and terminological / information / knowledge modelling ontologies.
c. Restriction. Ontology O2 is a restriction of ontology O1 if all definitions in ontology O1 are available in ontology O2 except for those that are redefined (cf. Farquhar et al. 1997). This kind of relation can be used in combination with alternative domain views, abstraction, primary and secondary ontologies, and terminology / information / knowledge modelling ontologies.
d. Mapping Relation. O1 is mapped onto O2
if some expression in O1 is linked to an expression in O2 where both
expressions are (assumed to be) semantically equivalent or similar. This kind of relation
can only be used in combination with alternative domain views since equivalent or similar
expressions cannot occur using the other partitioning principles.
Both the principles to define multiple ontologies and the ontology
relations can be exploited as indexing mechanisms. In the next section we propose a set of
questions that can be used to index the ontologies in the library.
3.3 Indexing the Ontologies in the Library
The indexing mechanism of the legal ontology library consists of two
groups of features: (I) the intra ontology features, and (II) the inter-ontology features.
Below, we present a set of questions for each group.
Intra Ontology Features
Supply name of the ontology, its author(s), their affiliation, relevant publications, relevant URLs, and the date at which it is designed?
What tools were used to design the ontology?
What is the purpose of the ontology?
Does the ontology make method / task-specific commitments? If so, to what method(s) / task(s).
Does the ontology make specific commitments towards legal subdomains (is the ontology reusable throughout all legal subdomains)? If so, state the restrictions and commitments?
Does the ontology make any commitments towards representational formalisms (which)?
What language is used to specify the ontology?
What are the most fundamental distinctions in the ontology?
How many concepts are defined (how many classes, relations, functions, instances and axioms)?
Has the ontology been used in practical information-system applications (prototype, operational)?
Sketch the internal structure of the ontology: (a) it is only a set of terms, (b) there is some structure, or (c) it has a high degree of structuring?
What organisation principle is used (a) none, (b) hierarchy, (c) graph.
Inter Ontology Features
Does the ontology include all definitions of another ontology in the library (which)?
Are all definitions in the ontology included in another ontology in the library (which)?
Does the ontology assume all definitions of another ontology in the library to be known (which)?
Are all definitions in the ontology assumed to be known in another ontology in the library (which)?
Does the ontology provide an alternative view on the legal domain than another ontology in the library (which)?
Is the ontology an specialisation of another ontology in the library (which)?
Is the ontology an abstraction of another ontology in the library (which)?
Is the ontology a restriction on another ontology (which)?
Can the ontology be mapped onto another ontology in the library? If so,
what is the nature of this mapping (a) method to domain, (b) domain to method, (c) method
to method, (d) domain to domain?
The idea to distinguish inter-ontology features next to the
intra-ontology features allows us to build a lattice of ontologies showing how the
ontologies relate to one another. If, while browsing through the library, one dislikes the
commitments of a particular ontology then the inter-ontology links will lead to similar
ontologies in the library which make different commitments. Alternatively, we could depict
the ontologies according to method and domain specificity, when the core ontologies will
tend to lie towards the origin. This is illustrated in figure 1 (cf. Van Heijst et
al., 1997).
4. Conclusions
If we try to relate our indexing questions to the four legal ontologies
discussed in section 2 the result is at first sight a bit disappointing. Almost all the
answers are no; the only exceptions being whether the ontologies give different views of
the domain, and, if we include the statute specific ontology for the DUBA as a different
ontology from the generic ontology, the former can be seen as an extension of the latter.
This is readily explained, however, if we consider Figure 1. From this we can see that all
the four ontologies we discussed earlier are intended to be core ontologies. As such, they
record rather fundamental decisions about how the domain is to be conceptualised, and
represent fundamentally different conceptualisations. If we had, for example, more
examples of statute specific ontologies developed from LFU, we would find considerable
inter-ontology relations with the statute specific ontology of the DUBA, particularly if
they were in similar areas of law, such as equal employment law. Our indexing questions
are most appropriate for rather detailed ontologies.
At the level of core ontologies, the users of the library will want to
select an ontology which embodies the fundamental design choices that they find congenial.
This really means that they have to look at the highest level distinctions; for example
the norm / concept / event distinction of FBO or the distinction into six kinds of
knowledge of LFU. What the users need to do therefore, is to be able to browse and
preferably compare, the competing ontologies starting with the top-level concepts. This
will at once alert them to the fundamental design choices to be made, and enable them to
decide which one is most in accord with their own conceptualisation. Armed with this
knowledge, they will then be able to select a more refined ontology which conforms to this
conceptualisation, using the other indexing questions. We observe that to construct a
software tool to support this activity it is desirable that the language in which the
ontologies are stated is formal and standardised.
Acknowledgements
The authors want to thank Radboud Winkels, Dean Jones and Robert van
Kralingen for their contribution.
References
Bench-Capon, T.J.M. (1991). Knowledge-Based Systems and Legal Applications, APIC series, No. 36, Academic Press, London, United Kingdom.
Bench-Capon, T.J.M., and F.P. Coenen (1992). Isomorphism and Legal Knowledge Based Systems, Artificial Intelligence and Law, Vol. 1, No. 1, pp.65-86.
Borst, P., J. Benjamin, B. Wielinga, and H. Akkermans (1996). An Application of Ontology Construction, ECAI '96 Workshop on Ontological Engineering, Budapest, Hungary.
Breuker, J.A., and W. van de Velde (1994). CommonKADS Library for Expertise Modelling, Reusable Problem Solving Components, J.A. Breuker, and W. van de Velde (eds.), IOS Press, Amsterdam, the Netherlands.
Farquhar, A., R. Fikes, and J. Rice (1997). The Ontolingua Server: A Tool for Collaborative Ontology Construction, Int. Journal of Human-Computer Studies, 46, pp.707-727.
Fridman Noy, N. and C.D. Hafner (1997). The State of the Art in Ontology Design, A Survey and Comparative Review, AI Magazine, 4,pp.53-74.
Gruber, T.R. (1995). Toward principles for the design of Ontologies used for Knowledge Sharing, Int. Journal of Human-Computer Studies, 43, pp.907-928.
Guarino, N. (1997a). Some Organizing Principles for a Unified Top-Level Ontology, Revised version of a paper appeared at AAAI 1997 Spring Symposium on Ontological Engineering (LADSEB-CNR Int. Rep. 02/97, V3.0, Italy).
Guarino, N. (1997b). Understanding, Building and Using Ontologies. A Commentary to "Using Explicit Ontologies in KBS Development", by van Heijst, Schreiber, and Wielinga, International Journal of Human and Computer Studies, Vol. 46 n. 2/3, pp. 293-310.
Heijst, G. van (1995). The Role of Ontologies in Knowledge Engineering, PhD. Thesis, University of Amsterdam, The Netherlands.
Heijst, G. van, A. Th. Schreiber, and B.J. Wielinga (1997). Using Explicit Ontologies in KBS Development, Int. Journal of Human-Computer Studies, Vol. 45, pp.183-292.
Kralingen, R.W. van (1995). Frame-based conceptual models of statute law, Computer/Law Series, No.16, Kluwer Law International, The Hague, the Netherlands.
McCarty, L.T. (1989). A Language for Legal Discourse, I. Basic Features, Proceedings of the Second International Conference on Artificial Intelligence and Law, pp.180-189, Vancouver, Canada.
McCarty, L.T. (1993). OWNERSHIP: A Case Study in the Representation of Legal Concepts, Presented at a Conference in Celebration of the 25th Anniversary of the Istituto Documentazione Giuridica, Florence, Italy, 1993.
Sowa, J.F. (1995). Top-level ontological categories, Int. Journal of Human-Computer Studies, Vol 43, pp.669-685.
Stamper, R.K. (1980). LEGOL: Modelling Legal Rules by Computer, Computer Science and Law, Bryan Niblett (ed.), Cambridge University Press, Cambridge, United Kingdom.
Stamper, R.K. (1991). The Role of Semantics in Legal Expert Systems and Legal Reasoning, Ratio Juris, Vol. 4, No. 2, pp.219-244.
Stamper, R.K. (1996). Signs, Information, Norms and Systems, Signs at Work, B. Holmqvist, and P.B. Andersen (eds.), De Bruyter, Berlin, Germany.
Valente, A. (1995). Legal Knowledge Engineering; A Modelling Approach, University of Amsterdam, The Netherlands, IOS Press, Amsterdam, The Netherlands.
Visser, P.R.S. (1995). Knowledge Specification for Multiple Legal Tasks; A Case Study of the Interaction Problem in the Legal Domain, Computer / Law Series, No. 17, Kluwer Law International, The Hague, The Netherlands.
Visser, P.R.S., and T.J.M. Bench-Capon (1996). The Formal Specification of a Legal Ontology, Proceedings of the Ninth International Conference on Legal Knowledge-Based Systems (JURIX'96), Van Kralingen et al. (eds.), Tilburg, The Netherlands, pp.15-24.
Visser, P.R.S. (1997). Structuring Multiple Ontologies in Heterogeneous Distributed Architectures, Fellowship Report, Intelligent Business Systems, BT Laboratories, Martlesham Heath, Ipswich, UK.
Visser, P.R.S. and R.G.F. Winkels (1997). Proceedings of the First International Workshop on Legal Ontologies; LEGONT '97, University of Melbourne, Law School, Melbourne, Victoria, Australia.
Visser, P.R.S., R.W. van Kralingen and T.J.M. Bench-Capon (1997), A Method for the Development of Legal Knowledge Systems, Proceedings of the Sixth International Conference on Artificial Intelligence and Law (ICAIL'97), Melbourne, Australia.
Visser, P.R.S. and T.J.M. Bench-Capon (1998). A Comparison of Four Ontologies for the Design of Legal Knowledge Systems, Artificial Intelligence and Law (accepted for publication).
1. 1 A full appreciation of Stamper's theory requires a more extensive discussion than the - necessarily very brief - description presented in this article. We have attempted to compile Stamper's 1991 and his 1996 article although there are some notable differences between both articles. When confusion could arise, we have used his 1996 article.
2. 2 This section is based on a similar discussion in Visser (1997).
3. 3 For a critical review of these distinctions, see Guarino (1997b).