Multilevel Legal Ontologies
Gianmaria Ajani1 , Guido Boella2 , Leonardo Lesmo2 , Marco Martin2 ,
Alessandro Mazzei2 , Daniele P. Radicioni2 , and Piercarlo Rossi3
1
3
Dipartimento di Scienze Giuridiche - Università di Torino
2
Dipartimento di Informatica - Università di Torino,
Dipartimento di Studi per l’Impresa e il Territorio, Università del Piemonte
Orientale
gianmaria.ajani@unito.it,
{guido,lesmo,mazzei,radicion}@di.unito.it,
notmart@gmail.com,
piercarlo.rossi@eco.unipmn.it
June 25, 2009
Abstract. In order to manage the conceptual representation of European law we have proposed the Legal Taxonomy Syllabus (LTS) and
the related methodology. In this paper we consider further legal issues
that emerged during the testing and use of the LTS, and how we take
them into account in the new release of the system. In particular, we
consider the problem of representing interpretation of terms besides the
definitions occurring in the directives, the problem of normative change
and the process of planning legal reforms of European law. In particular,
we show how to include into the Legal Taxonomy Syllabus the Acquis
Principles - which have been sketched by scholars in European Private
Law from the so-called Acquis communautaire -, how to take the temporal dimension into account in ontologies, and how to apply techniques
of natural language processing to the legal texts uploaded in the LTS.
1
Introduction
European Union Directives (EUDs) are sets of norms that have to be implemented by the national legislations and translated into the language of each
Member State. The general problem of multilingualism in European legislation
has recently been addressed by using linguistic and ontological methodologies
and tools, e.g. [13,10,17,33,34]. The management of EUD is particularly complex,
since the implementation of a EUD does not correspond to a straight transposition into a national law. By converse, managing this kind of complexity with
appropriate tools can facilitate the comparison and harmonization of national
legislation [7]. For instance, LOIS Project aims at extending EuroWordnet with
legal information, thus adopting a similar approach to multilinguism, with the
goal of connecting legal ontology to a higher level ontology [33].
In previous work we proposed the Legal Taxonomy Syllabus4 (LTS), a methodology and a tool to build multilingual conceptual dictionaries aimed at representing and analysing terminologies and concepts from EUDs [2,1,3,4]: the main
goal of the LTS concerns with the access of human experts to the EU documents.
LTS is based on the distinction between terms and concepts. The latter ones are
arranged into ontologies that are organised in levels. In [4] only two levels were
defined: the European level –containing only one ontology deriving from EUDs
annotations–, and the national level –hosting the distinct ontologies deriving
from the legislations of EU member states.
While annotating the EUDs, testing and using the system, some more requirements emerged from users expert in law, demanding for a more sophisticated approach along with further developmental efforts: first, we noted that it
is frequent the case of concepts which are the result of a doctrinal interpretation
process rather than of the definition in directives. If, on the one hand, the definitions in directives and their relation with the actual text are required by legal
scholars to have a precise model of European law, the layman is more interested
in the concepts which results from the doctrinal interpretation. Second, laws are
typical objects evolving through time. A big open issue to cope with in building
legal frameworks both at the European and at the national level is the normative change [26,9]. Concepts in the legal ontologies should not only represent the
consolidated legal text, but should also keep trace of the evolution of meaning.
Finally, besides the actual directives, the European Union aims to harmonize
law by reformulating terminology in a more coherent way. The European Commission provide in various way common principles, terminology, and rules for
law to address gaps, conflicts, and ambiguities emerging from the application of
European law, and this effort should be taken into account in the LTS as well.
Thus, in this paper we address the following research questions:
– How to consider not only the terms defined in the directives but also the
interpretation process of legal scholars in the LTS? How to better integrate
concepts and the text of EUDs in the LTS?
– How to extend the ontology with a temporal dimension to be able to represent normative change? How to allow users to search also for past meanings
of terms and the modified norms introducing them?
– How to extend the levels of the LTS from European and national to new
ones representing the possible reforms of European law? How to represent
the relation between the existing European law and the planned revisions?
We answer the first question by introducing concepts called abstract, in that
they are not related to a single directive, which should be conveniently recognized
as a grouping of concepts. The users will be allowed to navigate the ontology at
4
LTS is a dictionary of Consumer Law, which has been carried out within the broader
scope of the Uniform Terminology Project, http://www.uniformterminology.
unito.it [28]. The implemented system can be found at the URL: http://www.
eulawtaxonomy.org.
different levels of details depending on their goals. Moreover, we use natural language processing techniques to facilitate the management of legal text associated
to concepts.
We answer the second question by introducing time into the ontology and by
allowing to have new concepts replacing the old ones while keeping the latter in
the system as well.
We answer the third question by introducing new levels in the LTS and show
how can they manage a set of principles (namely, the Acquis Principles) which
is gaining in popularity to the ends of ameliorating the quality of EUDs national
implementations by Member States.
Note that our answers concern the methodology underlying LTS, but all the
features that will be described have a direct software implementation in the LTS
tool. The paper is structured as follows. In Section 2 we introduce the new legal
requirements we take into account in the revision of the LTS. In Section 3 we
first summarize the LTS (Sec. 3.1) and then we explain how the new system
satisfies the additional requirements previously outlined (Sections 3.2 to 3.5).
Conclusions end the paper.
2
Multilingual and Multilevel Ontologies for European
Directives
In this Section we start by briefly summarizing the motivations which lead to the
development of the LTS (Section 2.1), and then we introduce new requirements
which have been raised by legal experts using the LTS (Sections 2.2 to 2.4).
2.1
Terminological and conceptual misalignment
Comparative Law has identified two key points in dealing with EUD, which
make more difficult dealing with the polysemy of legal terms: we call them the
terminological and conceptual misalignments.
In the case of EUD (usually adopted for harmonising the laws in the Member
States), the terminological matter is complicated by the need to implement them
in the national legislations. In order to have a precise transposition in a national
law, a Directive may be subject to further interpretation. A single concept in a
particular language can be expressed in a number of different ways in a EUD and
in the national law implementing it. As a consequence we have a terminological
misalignment. For example, the concept corresponding to the word reasonably
in English, is translated into Italian as ragionevolmente in the EUD, and as con
ordinaria diligenza into the transposition law.
In the EUD transposition laws a further problem arises from the different national legal doctrines. A legal concept expressed in an EUD may not be present
in a national legal system. In this case we can talk about a conceptual misalignment. To make sense for the national lawyers’ expectancies, the European legal
terms have not only to be translated into a sound national terminology, but
they also need to be correctly detected when their meanings refer to EU legal
concepts or when their meanings are similar to concepts which are known in the
Member states. Consequently, the transposition of European law in the parochial
legal framework of each Member state can lead to a set of distinct national legal
doctrines, that are all different from the European one. In the case of consumer
contracts (like those concluded by the means of distance communication as in
Directive 97/7/EC, Art. 4.2), a related example of this phenomenon concerns the
notion of providing in a clear and comprehensible manner some elements of the
contract by the professionals to the consumers represents a specification of the
information duties which are a pivotal principle of EU law. Despite the pairs of
translation in the language versions of EU Directives (e.g., klar und verständlich
in German - clear and comprehensible in English - chiaro e comprensibile in
Italian), each legal term, when transposed in the national legal orders, is influenced by the conceptual filters of the lawyers’ domestic legal thinking. So, klar
und verständlich in the German system is considered by the German commentators referring to three different legal concepts: 1) the print or the writing of the
information must be clear and legible (Gestaltung der Information), 2) the information must be intelligible by the consumer (Formulierung der Information), 3)
the language of the information must be the national of consumer (Sprache der
Information). In Italy, the judiciary tend to control more the formal features of
the concepts 1 and 3, and less concept 2, while in England the main role has been
played by the concept 2, though considered as plain style of language (not legal
technical jargon) thanks to the historical influences of plain English movement
in that country.
Note that this kind of problems identified in comparative law has a direct
correspondence in the ontology theory. In particular Klein [21] has remarked
that two particular forms of ontology mismatch, are terminological and conceptualization ontological mismatch which straightforwardly correspond to our
definitions of misalignments.
2.2
Concepts Abstraction
The LTS system relies on the concept of unitary-meaning or umeaning: such
atomic concepts can be derived from excerpts of the text of legal norms, such
as European directives or national laws, and are arranged into two separate
categories of umeanings, as described in [4].
EUDs provide rigorous definitions of some terms, such as the definition of
the Italian term consumatore (consumer ), in the Italian version of the EUD
93/13/EEC, Art. 2 is:
[. . . ](b) “consumatore”: qualsiasi persona fisica che, nei contratti oggetto
della presente direttiva, agisce per fini che non rientrano nel quadro della
sua attività professionale; [...]
[. . . ](b) “consumer”: means any natural person who, in contracts covered
by this Directive, is acting for purposes which are outside his professional
activity; [. . . ] (our literal translation)
However, two facts must be pointed out. Different EUDs might affect different
aspects of the legislation: thus the definition of a term in a EUD only applies
to a specific context. Furthermore, EUDs could be written at different points in
time, and they can introduce diverging definitions. Let us consider the definition
of consumatore, as it appears in the Italian version of the EUD 2002/65/EC,
Art. 1:
[. . . ](d) “consumatore”: qualunque persona fisica che, nei contratti a distanza, agisca per fini che non rientrano nel quadro della propria attività
commerciale o professionale; [. . . ]
[. . . ](d) “consumer”: means any natural person who, in distance contracts covered by this Directive, is acting for purposes which are outside
his business or professional activity; [. . . ] (our literal translation)
We remark that in contrast with English, in Italian the second definition of consumatore is broader than the first one, since the term professionale (professional )
does not include commerciale (business). This divergence of term definitions can
often occur, since EUDs have usually a sectorial specific target. In this way, EUDs
covering different sectors can provide different definitions, and as many views on
the same concept. Lawyers and legislators started to put together highly sectorial concepts into more abstract concepts with broader meaning, in order to
describe (complex) entities, such as the consumatore in all of its aspects.
In recent years, in the Italian legislation EUDs are not being implemented
as single laws, but rather as groups of EUDs. The juridical concepts are defined
as the union of all the sectorial concepts provided by the individual EUDs, as a
result of the doctrinal interpretation process of directives.
We remark that these problems are common to all European languages. Consider, for instance the definition of consumer, in the English version of the EUD
1999/44/EC, Art. 1.2 is:
[. . . ] (a) consumer: shall mean any natural person who, in the contracts
covered by this Directive, is acting for purposes which are not related to
his trade, business or profession; [. . . ]
that has a different meaning with respect to the definition of consumer given in
the Council Directive 90/314/EEC, Art. 2.4:
[. . . ] “consumer” means the person who takes or agrees to take the package (‘the principal contractor’), or any person on whose behalf the principal contractor agrees to purchase the package (‘the other beneficiaries’)
or any person to whom the principal contractor or any of the other beneficiaries transfers the package (‘the transferee’) [. . . ]
The LTS should be able to represent both the more specific dimension related
to the definitions in EUDs and the more abstract one which results from the
doctrinal interpretation of European law. The LTS allows inserting the text
paragraphs where umeanings are defined. However, to gain better understanding
of legal concepts, it is often required to consider a broader fragment. For example,
in the case of consumer the definition is not enough, and it is necessary to
collect multiple paragraphs where consumer protection norms are presented and
discussed.
2.3
Normative Change
Another big open issue to cope with in building tools for describing legal frameworks both at the European and at the national level is the normative change [26].
One major problem, well-known in the literature, is the update of non-monotonic
ontologies and knowledge bases [9]. In other words, not necessarily ontologies and
knowledge bases have a structure constant through time (e.g., see [32]): concepts
and relations present in the ontology can become obsolete as new concepts and
relations are added. This is indeed the case of legal frameworks, that are continuously modified as new laws can modify paragraphs of old ones.
We can have two types of normative change: explicit change and implicit
change. In the first case the new norm explicitly states the abrogation of a
specific paragraph of an old law (for details on this line of investigation, please
refer to [11,8]). Alternatively, the newer law can state a concept in contradiction
to previous laws, but without mentioning them explicitly. In this case the concept
stated by the new law becomes the current one; also, the parts of the old laws
affected by changes (no longer updated) become obsolete.
2.4
Reforming European Law: toward a Common Frame of
Reference
In February 2003, the European Commission adopted a further communication
entitled “A More Coherent European Contract Law - An Action Plan” [14]. One
of the key measures proposed in the Action Plan was the elaboration of a Common Frame of Reference (CFR). According to the Action Plan, in which the
idea of a CFR was developed for the first time, a major aim of the forthcoming
CFR is to serve as a tool for the improvement of the EC law. The future CFR
was described in more detail in the Commission’s Communication on “European Contract Law and the Revision of the Acquis: The Way Forward” [15]. It
proposed that the CFR should provide fundamental principles of contract law,
definitions of the main relevant abstract legal terms and model rules of contract
law; its main purpose being to serve as a legislators’ toolbox.
In drafting the Action Plan the Commission emphasized that the CFR would
eliminate market inefficiencies arising from the diverse implementation of European directives, providing a solution to the non-uniform interpretation of European contract law due to vague terms and rules, now present in the existing
Acquis.
In particular, two issues arise from the vague terminology of EUDs. First,
directives adopt broadly defined legal concepts, therefore leaving too much freedom in their implementation to national legislators or judges. Second, directives
introduce legal concepts that are different from national legal concepts. Thus,
when judges face vague terms, they can either interpret them by referring to the
broad principles of the acquis communautaire, or they can refer to the particular goals of the directive in question. To respond to the Action Plan, in the last
few years, within the general framework of a “Network of Excellence” European
Project, a research group aiming at consolidating the existing EC law is working on the “Principles of the Existing EC Private Law” or “Acquis Principles”
(ACQP). These Principles will be discussed and compared with other outcomes
from different European research groups and, during a complex process of consultation with stakeholders under the direction of EC Commission, the CFR will
be set up. The Acquis Principles should provide a common terminology as well
as common principles to constitute a guideline for uniform implementation and
interpretation of European law [5,30]. One outcome of such project is the Acquis
Principles glossary, i.e., a set of interconnected terms and concepts.
The Acquis Principles have been sketched by scholars in European Private
Law from the so-called Acquis communautaire, the existing body of EU primary
and secondary legislation as well as the European Court of Justice decisions [5].
Nowadays such corpus contains some 80, 000 pages. Notwithstanding the importance of this existing body of settled laws, the Acquis is also a far wider notion,
encompassing an impressive set of principles and obligations, going far beyond
the internal market and including areas, such as agriculture, environment, energy
and transports.
In this paper we show how the multilevel architecture of LTS allows us to
relate the Acquis Principles with the legal concepts defined into the directives.
3
The Legal Taxonomy Syllabus
In this Section we first summarize the functionalities of the existing LTS [4], and
then we explain how it has been extended to cope with the new requirements
described in the previous Section.
3.1
The basic LTS
The main assumptions of our methodology come from studies in comparative
law [28] and ontologies engineering [21]:
– Terms –lexical entries for legal information–, and concepts must be distinguished; for this purpose we use lightweight ontologies [18], i.e. simple
taxonomic structures of primitive or composite terms together with associated definitions. They are hardly axiomatized as the intended meaning of
the terms used by the community is more or less known in advance by all
members, and the ontology can be limited to those structural relationships
among terms that are considered as relevant [25].
– We distinguish the ontology implicitly defined by EUD, the EU level, from
the various national ontologies. Each one of these “particular” ontologies
belongs to the national level : i.e., each national legislation refers to a distinct
national legal ontology. We do not assume that the transposition of an EUD
Term-Ita-A
Term-Ger-A
Ger-5
EU-1
Ita-2
Ger-3
Ita-4
Fig. 1. Relationship between ontologies and terms. The thick arcs represent the interontology “association” link.
automatically introduces in a national ontology the same concepts that are
present at the EU level.
– Corresponding concepts at the EU level and at the national level can be
denoted by different terms in the same national language.
A standard way to properly manage large multilingual lexical databases is
to make a clear distinction among terms and their interlingual acceptions (or
axies) [31,24].
In the LTS project to properly manage terminological and conceptual misalignment, we distinguish the notion of legal term from the notion of legal concept
and we build a systematic classification based on this distinction. The basic idea
in our system is that the conceptual backbone consists in a taxonomy of concepts (ontology) to which the terms can refer in order to express their meaning.
One of the main points to keep in mind is that we do not assume the existence of a single taxonomy covering all languages. In fact, the different national
systems may organize the concepts in different ways. For instance, the term contract corresponds to different concepts in common law and civil law, where it
has the meaning of bargain and agreement, respectively [29]. In most complex
instances, there are no homologous between terms-concepts such as frutto civile
(legal fruit) and income, but respectively civil law and common law systems can
achieve functionally similar operational rules thanks to the functioning of the
entire taxonomy of national legal concepts [19]. Consequently, the LTS includes
different ontologies, one for each involved national language plus one for the
language of EU documents. Each language-specific ontology is related via a set
of association links to the EU concepts, as shown in Fig. 1.
Although this picture is conform to intuition, in the basic LTS it has been
implemented by taking two issues into account. First, it must be observed that
the various national ontologies have a reference language. This is not the case
for the EU ontology. For instance, a given term in English could refer either to
a concept in the UK ontology or to a concept in the EU ontology. In the first
case, the term is used for referring to a concept in the national UK legal system,
Cancellation
Consumer protection
Termination
Withdrawal
Conclusione del contratto
Difesa del consumatore
Diritto di recesso
Risoluzione
Recesso
EU-1
Ita-1
Eng-1
EU-2
Eng-2
Eng-3
purpose
purpose
Ita-3
purpose
is-a
Ita-2
purpose
Ita-4
concerns
is-a
Eng-4
Ita-5
concerns
Ita-6
Fig. 2. An example of interconnections among terms.
whilst in the second one, it is used to refer to a concept used in the European
directives. This is one of the main advantages of LTS. For example klar und
verständlich could refer both to concept Ger-379 (a concept in the German
Ontology) and to concept EU-882 (a concept in the European ontology). This is
the LTS solution for facing the possibility of a partial correspondence between
the meaning of a term in the national system and the meaning of the same term in
the translation of a EU directive. This feature enables the LTS to be more precise
about what “translation” means. It makes available a way for asserting that two
terms are the translation of each other, but just in case those terms have been
used in the translation of an EU directive: within LTS, we can talk about direct
EU-to-national translations of terms, and about implicit national-to-national
translations of terms. In other words, we distinguish between explicit and implicit
associations among concepts belonging to different levels. The former ones are
direct links that are explicitly used by legal experts to mark a relation between
concepts. The latter ones are indirect links: if we start from a concept at a given
national level, by following a direct link we reach another concept at European
level. Then, we will be able to see how that concept is mapped onto further
concepts at the various national levels.
The situation enforced in LTS is depicted in Fig. 1, where it is represented
that the Italian term Term-Ita-A and the German term Term-Ger-A have been
used as corresponding terms in the translation of an EU directive, as shown by
the fact that both of them refer to the same EU-concept EU-1. In the Italian
legal system, Term-Ita-A has the meaning Ita-2. In the German legal system,
Term-Ger-A has the meaning Ger-3. The EU translations of the directive is
correct insofar no terms exist in Italian and German that characterize precisely
the concept EU-1 in the two languages (i.e., the “associated” concepts Ita-4 and
Ger-5 have no corresponding legal terms). A practical example of such a situa-
tion is reported in Fig. 2, where we can see that the ontologies include different
types of arcs. Beyond the usual is-a (linking a category to its supercategory),
there are also the arcs purpose, which relate a concept to the legal principle motivating it, and concerns, which refer to a general relatedness. The dotted arcs
represent the reference from terms to concepts. Some terms have links both to a
National ontology and to the EU Ontology (in particular, withdrawal vs. recesso
and difesa del consumatore vs. consumer protection).
The last item above is especially relevant: note that this configuration of
arcs specifies that: 1) withdrawal and recesso have been used as equivalent terms
(concept EU-2) in some European Directives (e.g., Directive 90/314/EEC). 2) In
that context, the term involved an act having as purpose some kind of protection
of the consumer. 3) The terms used for referring to the latter are consumer
protection in English and difesa del consumatore in Italian. 4) In the British
legal system, however, not all withdrawals have this goal, but only a subtype
of them, to which the code refers to as cancellation (concept Eng-3). 5) In the
Italian legal system, the term diritto di recesso is ambiguous, since it can be used
with reference either to something concerning the risoluzione (concept Ita-4),
or to something concerning the recesso proper (concept Ita-3).
The LTS is a theoretical instrument as well as a software platform that is
at the present time yet working. The actual number of annotated terms and
concepts are provided in Tables 1 and 2, respectively. Terms were initially extracted from a corpus of 24 EC directives, and 2 EC regulations, reported in
Appendix 4. Occurrences of such entries were detected from national transposition laws of English, French, Spanish, Italian and German jurisdictions.
Table 1. Number of terms
Language National European
French
8
47
Italian
28
52
English
71
75
41
60
Spanish
German
66
98
total
214
332
Finally, it is possible to use the LTS to translate terms into different national
systems via the transposed concepts at the European level, i.e. by using the
implicit associations. For instance suppose that we want to translate the legal
term credito al consumo from Italian to German. In the LTS credito al consumo
is associated to the national umeaning Ita-175. We find that Ita-175 is the
transposition of the European umeaning EU-26 (contratto di credito). EU-26 is
associated to the German legal term Kreditvertrag at European level. Again,
we find that the national German transposition of EU-26 corresponds to the
Table 2. Number of concepts
Language National European
French
7
43
Italian
24
45
English
54
71
Spanish
34
56
German
52
75
total
171
290
EU-50
Consumer
INTERPRETED_AS
INTERPRETED_AS
EU-25
Description:
EU-28
Description:
...
...
References:
References:
93/13/EEC, Art. 1
02/65/EC, Art. 2
Fig. 3. Umeanings Eu-25 and Eu-28 are interpreted by the more abstract umeaning
Eu-50, the link between Eu-50 and the term “consumer” is implicit.
national umeaning Ger-32 that is associated with the national legal term Darlehensvertrag. Then, by using implicit links in the European ontology, we can
translate the Italian legal term credito al consumo into the German legal term
Darlehensvertrag.
3.2
Enhancing LTS with interpretation and abstraction
As described in Section 2.2, different pieces of legislations can bear different
definitions of terms. Having different detailed definitions is important during the
interpretation of very sectorial legal cases, but for the general case it is important
to have a view that abstracts from the peculiarities of specific domains.
In order to solve this problem we introduced a new kind of ontologic relation
called INTERPRETED AS : it is a non transitive relation where the more general
umeaning, that we will call group leader represents the abstracted concept that
groups the meaning of a number of more specific umeanings, that are the sectorial
umeanings defined in the individual EUDs or national laws (see Fig. 3).
We have also introduced a number of constraints and integrity checks to
ensure that the semantics of the grouping concept is respected and to improve
the usability of the system:
– Each umeaning can belong to a single group.
– A group leader cannot exist without group members.
– When the user searches into the umeaning database, more specific umeanings
are excluded from the results unless the user explicitly asks to show them,
i.e. only the group leaders are shown in the results.
The need to contextualize concepts to the EUDs defining them leads to the
need of more complex instruments to deal with the language of the norms. An
umeaning is defined by the legal texts themselves; this makes clear that the
creation of umeaning is a quite long task, because it requires from the user
searching and reading a very large number of documents.
In order to ease this process, we developed a database that contains the full
versions of the desired EUDs and national laws. In this way, the user can carry
out his task according to the following workflow.
– The user creates a new umeaning linked with the term he wants to define.
– He selects relevant citation from legal text; consequently, the browser is
redirected to a search page and the main term attached to the umeaning is
used as the default query.
– After choosing one of the search results, the full text of the legal document
is displayed, with the search terms highlighted.
– Finally, the user selects the text that will go in the citation with the mouse
and confirms the insertion in the references database.
3.3
LTS with normative change
When a new normative is approved and enacted it can define a number of new
umeanings; moreover it can happen that the same law can change a number
of old umeanings defined by old laws. In particular, these old umeaning can
become obsolete and no longer valid. We are aware of the difficulties concerning
the modelling of the time in artificial intelligence and in formal ontology too5 .
Anyway, in LTS we adopted a naive solution in order to manage the simpler
situation concerning t In the LTS it was necessary to delete all old umeanings,
causing the loss of all historic informations from the database, informations that
are quite valuable to better understand the evolving of the normative. This
problem was resolved by using the same solution adopted for the interpretation
and abstraction of the norms (Section 3.2), i.e. empowering LTS with a new
ontological relation called REPLACED BY.
When the paragraph of an EUD defining an umeaning has been modified
by a new EUD, the new one defines a new umeaning that will replace the old
umeaning in the ontology. There will be a relation of type REPLACED BY
5
E.g. see [6] for a general survey and [26] for normative systems
!"#$%&'()*"+,-
8!9:7;!<6=>
!"./%&'()*"+,-
<2?,%&*,@?,+A,-B&C#&C//C
4567
4567
!"#$%&01)23&'()*"+,!D@31'1?3E&1)*,-?,F&-,32?1()*
7"?(+2?1'233E&G,),-2?,F&-,32?1()*
Fig. 4. An example of use of the REPLACED BY relation.
between the two umeanings, where the child umeaning is replaced by the more
general umeaning. Also in this case the new ontologic relation has some peculiar
characteristics that distinguishes it from the usual ontologic relations (Fig. 4):
– A REPLACED BY relation brings with it a new data field not present in
the other relations: the substitution date.
– When the user performs a search in the umeanings database the replaced
ones will not be shown, unless the user asks for a certain past date, thus
obtaining a snapshot of the legal ontology that was valid in that particular
moment.
– When a new umeaning replaces an old one all the ontologic relations where
the old umeaning appeared are automatically copied in the new umeaning. If
some of them are no longer valid with the new umeaning, manual intervention
from the user is required.
3.4
Enhancing LTS with NLP techniques
In order to speed up the annotation process LTS 2.0 implements a number of
facilities based on natural language processing. A key point into the definition
of a new concept is important to collect all the occurrences of the terms that
implement the concept. With this aim, we realized a new intelligent search procedure for terms based on i. inverted index, ii. stemming, iii. function words, iv.
document similarity.
The term search through the legal texts database is realized with the standard
inverted index technique. The inverted index has a quite simple structure: it is a
relational table that maps the terms into the documents containing them, along
with the position in the document where terms actually occur. The documents
belonging to the national and European levels are stored and indexed in separate
relational tables.
When the user searches for a term in the documents database, the search
is not performed upon the exact word, rather with their roots, so for instance
when performing the search on the term “contracts” also documents containing
only “contract” will be found, this seems to enhance the information retrieval
performance as shown in [22]. The root extraction technique is called stemming
and in the LTS we used a library written in the Snowball (http://snowball.
tartarus.org) language by Martin Porter, using the Porter’s algorithm [27].
Inverted index based on stems can be computationally hard on a large number of directives. Anyway the index size remains quite small if we exclude the so
called function words (e.g. prepositions, articles), i.e. words with a precise grammatical function but without a real semantic content. In addition, we exclude
too a number of common words, e.g. “mio” (“my”) which no improve the quality
of the search. By excluding these lists of word we hardly reduced the number of
indexes and then improve the performance of the system. For instance the index
of the EUD 90/314/EC decreases from 2336 records to 1152.6
By using the inverted index technique we obtain as side effect a statistical
distribution of the words among the documents. This statistic allow us to define
a document similarity measure that can be used into a number of tasks, e.g.
to choose the next Directive to annotate with respect to a number of prefixed
concepts/terms The computation of the similarity index between documents
relies on the so called cosine document similarity technique [23,20].
3.5
Representing a new perspective in LTS: the Acquis Principles
Level
One major feature of the LTS approach relies on distinguishing legal information
as belonging to different levels. At the current stage of development, the system
manages terms and meanings at both EU and national levels. The former one
is an ontology of legal concepts derived from the EUDs; the latter one includes
national legal ontologies coming from the various national legal systems. The
current approach has been devised to be general enough to account for heterogeneous legal sources (like, e.g., EUDs and “Decreti Legislativi” for European
and Italian national levels respectively), and flexible enough to be extended by
adding further levels. To add a new level into the system, we connect a new
legal ontology to an exiting one. The new level is linked via explicit associations
connecting a concept belonging to the new ontology and a concept belonging to
the existing level ontology. We are applying this procedure in order to define an
Acquis level to the LTS.
We introduce the Acquis level into the LTS by defining explicit associations
between Acquis Principles concepts and EU-level concepts. For example, in Fig. 5
we have that the concept EU-25 (corresponding to the English legal term creditor) present in a EUD is explicitly associated to the national legal concepts
6
We adopted the list of function words described in [?,?] (http://members.unine.
ch/jacques.savoy/clef/).
EU-25
AC-72
AC-72
EU-25
Ita-124
Spa-110
creditor
creditor
creditore
prestamista
implicit association
} explicit associations
European level
Acquis level
Spa-110
Ita-2
Ita-124
National level:
Italian
National level:
Spanish
Fig. 5. LTS augmented with the Acquis level. Thick lines indicate explicit associations;
thin lines indicate implicit associations.
Ita-124 (finanziatore) and Spa-110 (prestamista) for Italian and Spanish, respectively. We can add the term creditor from the Acquis Level by inserting
an explicit association between the Acquis legal concept AC-72 and the European legal concept EU-25. As a consequence, the concept AC-72 is implicitly
associated to the legal concepts Ita-124 and Spa-110. This fact has deep consequences on the way one can build systems for reasoning, that are allowed to
make paths passing through more than two levels, thereby offering new insights
(and ready-to-use associations between terms) to scholars in comparative law.
4
Conclusions
In this paper we discuss some new features that have recently been introduced in
the LTS, a tool for building multilingual conceptual dictionaries for the EU law.
The tool is based on lightweight ontologies to emphasize the distinction between
concepts and terms. Different ontologies are built at the EU level and for each
national language, to deal with polysemy and terminological and conceptual
misalignment.
Many attempts have been done to use ontologies in legal field (e.g. [10,13],
and there are two projects that are strictly related to our approach, i.e. LOIS
and DALOS projects. LOIS project7 [34], aims at extending EuroWordnet with
7
http://www.loisproject.org
legal information, then it adopt a similar approach to multilinguism and aims
at connect legal ontology to a higher level ontology. DALOS project8 [16], aims
at building an ontological-linguistic resource for the multilingual EU legislative
drafting,and for linguistically reliable national transpositions is based and on
the LOIS database. Whilst the final goal of LOIS is to support applications
concerning information extraction and the DALOS project is realized with the
aim to help the legislative process, the LTS we describe herein concerns with
the access of human experts to the EU documents. Moreover, the LOIS and
DALOS approach follows the top-down paradigm, i.e. it starts from an existing
core ontology (the root of the the ontological tree), whilst on basic assumption
of our approach is the bottom-up processing, i.e. starting from the annotation
of the concept/term occurrences in EUD (the leaves of the ontological tree).
The present paper illustrates how to distinguish between concepts as they are
defined in the text of the directives and the concepts representing the doctrinal
interpretation of the terms. Moreover, we point out how to deal with normative
change by introducing a temporal dimension in ontologies. Finally we show how
to add further levels of representation to the EU and national levels. In particular, we described a methodology enabling to insert into the LTS platform
a novel set of principles, called Acquis principles. The Acquis principles can be
encoded into a number of legal concepts (along with a terminology), and belong
to a new level of representation, called Acquis priciples level. These concepts
can be integrated into the LTS by new explicit (and implicit) associations, that
connect the Acquis principles level with European and national levels.
Two main problems arise in our approach: the first one is theoretical, and it
concerns the issue of evaluating the performance of system with more massive
data. We would like to show with some quantitative measure the theoretical
adequacy of LTS. Secondly, the amount of work needed to annotate the EUDs
with concepts, terms and their transpositions, is huge. Future work will involve
exploring ways to extend the LTS ontology and populating it at the various
levels by semi-automatic approaches [12].
References
1. G. Ajani, G. Boella, L. Lesmo, M. Martin, A. Mazzei, and P. Rossi. A development tool for multilingual ontology-based conceptual dictionaries. In Proc. of 5th
International Conference on Language Resources and Evaluation, LREC06, pages
1–6, Genoa, 2006.
2. G. Ajani, G. Boella, L. Lesmo, A. Mazzei, and P. Rossi. Multilingual conceptual
dictionaries based on ontologies. In Proc. of V Legislative XML Workshop, pages
1–14, Florence, 2006. European Press Academic Publishing.
3. G. Ajani, G. Boella, L. Lesmo, A. Mazzei, and P. Rossi. Multilingual Ontological
Analysis of European Directives. In Proceedings of the 45th Annual Meeting of the
Association for Computational Linguistics Companion Volume Proceedings of the
Demo and Poster Sessions, pages 21–24, Prague, Czech Republic, 2007. Association
for Computational Linguistics.
8
http://www.dalosproject.eu/
4. G. Ajani, G. Boella, L. Lesmo, A. Mazzei, and P. Rossi. Terminological and ontological analysis of european directives: multilinguism in law. In 11th Internation
Conference o Arificial Intelligence and Law (ICAIL), pages 43–48, 2007.
5. G. Ajani and H. Schulte-Nölke. The Action Plan on a More Coherent European
Contract Law: Response on Behalf of the Acquis Group, 2003.
6. J. F. Allen. Towards a general theory of action and time. Artificial Intelligence,
23(2):123–154, 1984.
7. A. Boer, T.M. van Engers, and R. Winkels. Using ontologies for comparing and
harmonizing legislation. In ICAIL, pages 60–69, 2003.
8. R. Brighi, L. Lesmo, A. Mazzei, M. Palmirani, and D.P. Radicioni. Towards Semantic Interpretation of Legal Modifications through Deep Syntactic Analysis. In
Jurix 2008: The 21st Annual Conference, Frontiers in Artificial Intelligence and
Applications. IOS Press, 2008.
9. M. Cadoli and F. M. Donini. A Survey on Knowledge Compilation. AI Communications, 10(3–4):137–150, 1997.
10. P. Casanovas, N. Casellas, C. Tempich, D. Vrandecic, and R. Benjamins. OPJK
modeling methodology. In Proceedings of the ICAIL Workshop: LOAIT 2005, 2005.
11. M. Cherubini, G. Giardiello, S. Marchi, S. Montemagni, P. Spinosa, and G. Venturi.
NLP-based metadata annotation of textual amendments. In Proc. of WORKSHOP
ON LEGISLATIVE XML 2008, JURIX 2008, 2008.
12. P. Cimiano. Ontology Learning and Population from Text: Algorithms, Evaluation
and Applications. Springer, 2006.
13. S. Després and S. Szulman. Merging of legal micro-ontologies from european directives. Journal of Artificial Intelligence and Law, February 2007.
14. European Commission. Communication from the Commission to the European
Parliament and the Council - A More Coherent European Contract Law - An
Action Plan. COM, 2003.
15. European Commission. European Contract Law and the revision of the acquis: the
way forward. Communication from the Commission, COM(2004)(651 final), 2004.
16. E. Francesconi and D. Tiscornia. Building semantic resources for legislative drafting: The dalos project. pages 56–70, 2008.
17. E. Giguet and P.S. Luquet. Multilingual lexical database generation from parallel
texts in 20 european languages with endogenous resources. In Proceedings of the
COLING/ACL 2006 Main Conference Poster Sessions, pages 271–278, Sydney,
Australia, 2006. Association for Computational Linguistics.
18. F. Giunchiglia and I. Zaihrayeu. Lightweight Ontologies. Technical Report DIT-07071, University of Trento, Department of Information and Communication Technology, 38050 Povo – Trento (Italy), Via Sommarive 14, 2007.
19. M. Graziadei. Tuttifrutti. In P. Birks and A. Pretto, editors, Themes in Comparative Law, pages –. Oxford University Press, 2004.
20. J. D. Isaacs and J. A. Aslam. Investigating measures for pairwise document similarity. Technical Report PCS-TR99-357, 1999.
21. M. Klein. Combining and relating ontologies: an analysis of problems and solutions.
In Workshop on Ontologies and Information Sharing, IJCAI’01, Seattle, USA,
2001.
22. R. J. Krovetz. Word sense disambiguation for large text databases. PhD thesis,
University of Massachusetts, 1995.
23. D. Lin. An information-theoretic definition of similarity. In ICML ’98: Proceedings
of the Fifteenth International Conference on Machine Learning, pages 296–304,
San Francisco, CA, USA, 1998. Morgan Kaufmann Publishers Inc.
24. V. Lyding, Elena Chiocchetti, G. Sérasset, and F. Brunet-Manquat. The LexALP information system: Term bank and corpus for multilingual legal terminology consolidated. In Proc. of the Wokshop on Multilingual Language Resources
and Interoperability, ACL06, pages 25–31, 2006.
25. D. Oberle, editor.
Semantic Management of Middleware.
Springer Science+Business and Media, 2005.
26. M. Palmirani and R. Brighi. Time model for managing the dynamic of normative
system. Electronic Government, pages 207–218, 2006.
27. M.F. Porter. An Algorithm for Suffix Stripping. Program, 3(14):130–137, 1980.
28. P. Rossi and C. Vogel. Terms and concepts; towards a syllabus for european private
law. European Review of Private Law (ERPL), 12(2):293–300, 2004.
29. R. Sacco. Contract. European Review of Private Law, 2:237–240, 1999.
30. R. Schulze. European Private Law and Existing EC Law. European Review of
Private Law (ERPL), 2005.
31. G. Sérasset. Interlingual lexical organization for multilingual lexical databases in
NADIA. In Proc. COLING94, pages 278–282, 1994.
32. The Gene Ontology Consortium. Gene Ontology: tool for the unification of biology.
Nature Genetics http: // genetics. nature. com , 25:25–29, 2000.
33. D. Tiscornia. The Lois Project: Lexical Ontologies for Legal Information . In
Proceedings of the V Legislative XML Workshop. European Press Academic Publishing, 2007.
34. P. Vossen, W. Peters, and J. Gonzalo. Towards a universal index of meaning. In
Proc. ACL-99 Siglex Workshop, 1999.
APPENDIX A: List of EC directives
Core directives
– 84/450/EEC concerning misleading advertising
– 85/374/EEC concerning liability for defective products
– 85/577/EEC to protect the consumer in respect of contracts negotiated away
from business premises
– 87/102/EEC concerning consumer credit
– 90/88 concerning consumer credit
– 90/314/EEC on package travel, package holidays and package tours
– 93/13/EEC on unfair terms in consumer contracts
– 94/47/EC on the protection of purchasers in respect of certain aspects of
contracts relating to the purchase of the right to use immovable properties
on a timeshare basis
– 97/7/EC on the protection of consumers in respect of distance contracts
– 97/55/EC concerning misleading advertising so as to include comparative
advertising
– 98/6 on consumer protection in the indication of the prices of products offered to consumers
– 98/7 concerning consumer credit
– 98/27/EC on injunctions for the protection of consumers’ interests
– 99/44/EC on certain aspects of the sale of consumer goods and associated
guarantees
– 2000/13/EC relating to labelling, presentation and advertising of foodstuff
– 2001/95 on general product safety
– 2002/65/EC concerning the distance marketing of consumer financial services
– Regulation 2006/2004/EC on co-operation between national authorities responsible for the enforcement of consumer protection laws
– Directive 2005/29/EC on Unfair Commercial Practices
Ancillary directives
76/768/EEC relating to cosmetic products
88/378/EEC on toy safety
89/552/EEC on TV broadcasting activities
96/74/EC on textile names
97/5/EC on cross border credit transfers
Recommendation 98/257 on the principles applicable to bodies responsible
for the out-of-court settlement of consumer disputes
– 2000/31/EC on electronic commerce
– Regulation 2560/2001/EC on cross-border payments in Euro
–
–
–
–
–
–