=Paper=
{{Paper
|id=Vol-321/paper-8
|storemode=property
|title=Semantic Spaces and Multilingualism in the Law: The Challenge of Legal Knowledge Management
|pdfUrl=https://ceur-ws.org/Vol-321/paper8.pdf
|volume=Vol-321
|dblpUrl=https://dblp.org/rec/conf/icail/Liebwald07
}}
==Semantic Spaces and Multilingualism in the Law: The Challenge of Legal Knowledge Management==
https://ceur-ws.org/Vol-321/paper8.pdf
Semantic Spaces and Multilingualism in the Law:
The Challenge of Legal Knowledge Management
Doris Liebwald
Vienna Center for Computers and Law, Austria
d@liebwald.com
Abstract. It is the concern of the author to arrange cogitations and experiences
she gained by collaborating in relevant international project works, by conducting
scientific studies regarding legal knowledge representation and by teaching legal
information retrieval. The main focus is the demonstration of problems of commu-
nication within and between humans and legal information systems, which are often
hidden, overlooked or ignored. The author uses the concept "semantic spaces" to de-
scribe and explain semantic related difficulties detected in legal knowledge bases and
data retrieval. Realization of these various semantic spaces might help further work
in this area. Emphasis is also placed on the problem of multilingualism and diversity
of legal cultures in EU legislation. The practical examples of the EU tools N-Lex
and EUROVOC are used to illustrate the various situations, the current limits and
the specific requirements and information needs in multilingual and cross-national
legal information retrieval.
Keywords: Semantics, linguistics, information retrieval, knowledge presentation,
legal language, multilingualism, cross-national IR, diversity of legal cultures and
traditions, ontology, thesaurus, European Union law, national law, EUROVOC, N-
Lex.
1. Introduction
This paper deals with machine processible semantics. Of particular
concern are the following questions. Firstly, if it is possible, how can
the normative and the real world be represented in machine executable
language? Secondly, what problems must be resolved in order to accom-
plish this task? Numerous previous attempts to represent the meaning
of legal concepts and the knowledge related to those concepts, especially
in regards to coping with the step from simple string matching to an
interpretation and comprehension of semantics, have proven tedious and
labor-intensive.
What are the goals for these efforts? The goals can be divided into
two categories: on one side is the user-oriented editing of legal in-
formation to provide experts as well as laymen easier access to legal
documents; on the other is the implementation of machine-processible
representation of legal norms to create systems which are capable of
132 Doris Liebwald
applying legal rules or supporting humans in the application of the
rules1 .
Communication takes places within and between a wide range of
different semantic spaces2 . In the area of law it is important to consider
the various perceptions and information needs of the large array of
people involved in the process. Examples are that of a judge, whose
focus is on case-solving, the application-oriented approach of an admin-
istrative officer, the systematical point of view of a legislative drafter,
or that of the persons subject to the law, i.e. the layperson. Even in
the domain of “legal informatics”, different semantic spaces exist and
cause communication errors between legal and computer experts. While
the computer scientist uses the syntax and semantics of a programming
language, the lawyer considers the treatment of legal conceptualities,
which are not easy for the legal expert to formulate in a computer
sensitive way.
Another important element is intelligibility. Intelligibility is not in-
herent in the text; it is rather a process of understanding – a construc-
tive, mental activity. Background knowledge and the intention of the
reader as well as the design, composition and characteristics of the text
play an important role. When reading text, common sense knowledge,
factual knowledge, and the individual semantic spaces of the reader
are activated. Therefore reading comprehension, which is a knowledge
dependant mental representation, goes beyond what is explicitly com-
municated by the text. Is it possible to represent textual knowledge
and implicit human knowledge with machines? It is indeed possible,
however, only partially.
Ideally, semantic editing begins at the origin, such as in the prepa-
ration phase of a document, e.g. the draft bill. In this way knowledge
about the realization of a law (e.g. explanatory notes, expert reports,
opinions, etc.) and other metadata may be correctly related at source.
Further applications may reuse and exploit this knowledge base to
support further legislation (also amendments, impact assessments, fol-
low up costs, etc.), execution of the enacted law, and decision making
processes as well as information retrieval and document and knowledge
management in general.
1
Prominent examples are automatic contracting in e-commerce and rule-based
systems for public administration or large insurance companies.
2
The author introduces the concept “semantic space” to point up the different
interpretations and spaces of meanings attached to a specific phenomenon or con-
cept. The more similar different semantic spaces are, the easier communication will
take place. People and/or machines not sharing a common or at least very similar
semantic space run the risk of more or less obvious communication errors. See the
in-depth analysis in Liebwald (2007): Semantische Räume als Strukturhintergrund
der Rechtsetzung (“Semantic Spaces as Structural Patterns of Legislation”).
Semantic Spaces and Multilingualism in the Law 133
1.1. Terms and Semantics
A term obtains conceptual content by the semantics assigned to it.
Semantics3 refers to the meanings attached to words, expressions and
sentences, and are not part of the syntax; semantics comes from the
“outside” and is constructed by individual mental models. Semantics
are needed to turn terms into contextual concepts4 . Term relates to the
exterior, concept to the semantic content.
Concepts are used to characterize and to distinguish phenomena,
whereby, dependent from the observer, each phenomenon may feature
different semantic spaces. Individuals use different notions and diffe-
rent pictures of reality. Therefore, the intention of the author of this
paper may differ to the interpretation of the paper by the reader. A
certain notion of reality is not necessarily true or false, rather, it may
be considered as more or less appropriate or functional. But even one
particular observer may, dependent on the respective context, interpret
one particular phenomena in different ways. For instance, the mother
may relate the term warmth to love and security. The physicist, however,
may offer a definition on the transfer of thermal energy. In the case that
the mother and the physicist meet in a cold room, they will attach the
same or at least a very similar meaning to the term warmth. Can the
same be said about the term warming?
The meaning assigned to a phrase or sentence and therefore the
interpretation and understanding of (technical) language may not only
significantly differ between diverse organizations or expert circles, but
also between the individuals participating. Furthermore, semantics are
dynamic: they may change, e.g. due to new experiences or knowledge,
changes in reality, progress.
In a joint semantic space, a space of mutual understanding, the se-
mantics must follow a logic, which is shared by all members of this
space.5 Semantic spaces can be defined as networks of concepts that are
used to describe the world as well as for behavioral orientation of the
individuals acting in these spaces. Therefore semantic interoperability
3
From semantikos (significant meaning), Greek; derives from sema, semeion
(sign).
4
According to ISO 1087 a concept is defined as “a unit of thought constituted
through abstraction on the basis of properties common to a set of objects”; this defi-
nition is accompanied by the note “concepts are not bound to particular languages;
they are, however, influenced by the social or cultural background.”
5
Compare Uschold’s definition of ontology: “An ontology is a shared understand-
ing of some domain of interest.” Uschold/Gruninger (1996): Ontologies: Principles,
Methods and Application (1996).
134 Doris Liebwald
is of particular importance. Semantic interoperability6 exists where the
accurate meaning of information is understood and interpreted in the
same way by all individuals and applications involved. All the actors
must share the same model of what the data represents. The necessary
linkage of several semantic networks of concepts necessitates a network
of semantic spaces.
1.2. Legal Concepts7
A legal practitioner applies conceptual thinking and legal structural
knowledge that she or he gained over long-term training. The com-
plexity of law demands an abstract, differentiating, economical, and
functional technical language (“legal language”) which is able to repre-
sent the structures and meanings in law. The law is not just a collection
of mechanical if/then-rules; based on the same facts and on the same
legal rules, legal experts may indicate contradictory solutions. A correct
syllogism may be overruled by social conventions, principles or extra-
ordinary circumstances. Although where explicit knowledge exits, some
legal problems may not be resolved simply and legal decisions will not
always be predictable; in other cases the legal expert may be confronted
with controversial facts.
Law is based on text and language and language is dependent on
interpretation. Even if the lawmaker is anxious to reach maximal pre-
cision in legal texts and concepts, she or he will never reach absolute
precision, because language itself is often ambiguous. Similarly, vague
legal concepts can be considered an answer to missing accuracy of real-
ity. Furthermore there exists some deliberate vagueness of legal concepts
(or perhaps even deliberate incomprehensibility of legal texts). Reasons
for this could be to cover future, not yet predictable circumstances or
to cover at least all typical cases, to leave space for more specific rules,
judicial discretion and interpretation, or just because more “accurate”
political consent is missing.
Where legal rules are implemented in informatics systems, classical
logic of jurisprudence and symbolic logic of informatics encounter one
another. Open legal concepts, inherent dynamics of law, system models
and syntactic ambiguities prove to be extremely problematic, whereby
vague concepts seem to be the largest obstacle to overcome.
6
Galinski follows the semiotic triad and cuts more accurately into a syntac-
tic, pragmatic and conceptual level of semantic interoperability. Galinski (2006):
Wozu Normen? Wozu semantische Interoperabilität? (“Why Norms? Why Semantic
Interoperability?”).
7
For a competent and comprehensive scholarly piece see Bydlinski (1991):
Juristische Methodenlehre und Rechtsbegriff2 (“Legal Methodology and Nomen
Juris”).
Semantic Spaces and Multilingualism in the Law 135
1.3. Semantic Spaces in Law
The legal language cannot be considered as one semantic space, but
rather a network of semantic spaces. Therefore it is not sufficient to
only differentiate between legal experts and laypersons, since even bet-
ween and within various groups of legal experts, the concepts, document
types, styles of writing and parlance may vary.8 Each field of law forms
its own specific concepts and structures, which all show significant
differences in their semantics. This is also true within legislation, ad-
ministration, justice and doctrine. In some cases, when a draft bill,
the enacted law and subsequent amendments are compared, there is a
substantial shift in semantics; in other cases, judges’ interpretations of
a constant legal rule may change9 . Where legal experts interact with
other experts, the differences in the semantics of jargon may also have
an effect, e.g. in reports, opinions, studies, comments. In such groups
hidden misunderstandings are “pre-programmed”. Divergent semantic
spaces of different national legal systems or of national legal languages
in comparison to the EU legal language are, however, more obvious.
Nevertheless, the identification and expression of the subtle differences
of similar concepts that arise from various national legal traditions is a
sophisticated process.
1.4. The Problem of Multilingualism and Cultural
Diversity in EU-Law
The EU currently embraces 27 Member States and has 23 official lan-
guages10 . Legislation and documents of major public importance or in-
terest are produced in all official languages, but most of the institutions’
work is available in French and/or English only. Communication with
the EU and its institutions by governments, civil servants, businesses
and citizens may take place in any of the official languages.
Especially in regards to legal texts, multilingualism and diversity
in legal culture pose intractable situations. Of course, EU legislation is
translated into 23 languages, but the EU legal language and the specific
8
Consider also e.g. the different semantic spaces of a public appointed/sworn
expert, an eye-witness, the victim, the offender, the attorneys, the judge, the jury,
the media, a person who has the power of pardon, etc.
9
See e.g. Warta (2005): Zauberworte – Verwandlungen des Gleichheitsgrund-
satzes in der Judikatur des österreichischen Verfassungsgerichtshofes (“Magic Words
– Metamorphoses of the Principle of Equality in the Legal Practice of the Austrian
Constitutional Court”).
10
Some languages spoken in Member States (e.g. Catalan, Welsh, Basque, Breton,
Sardinian) don’t have the official EU language status. English, French and German
are the three strongest languages within the EU.
136 Doris Liebwald
concepts chosen do not correspond with the national legal language
and concepts of the respective Member State to a very high degree.11
27 Member States interpret the same legal text, each influenced by
its own political system, legal tradition, legal language and concepts,
and overall legal view. Member States are required to implement EU
legislation into their existing framework of national legislation, and
these frameworks are not congruent with one another to varying degrees.
Within the EU most countries belong to the civil law tradition, with the
exceptions of Ireland and the United Kingdom. In some countries, the
“Länder”, or states, have minor legislative importance, but this is not
true in all countries, e.g. Germany, Austria, and Belgium. Even where
the same language is used (e.g. Austria, Germany), the legal systems, its
structures, hierarchies and legal terminology differ. Therefore, e.g. one
particular EU Directive12 may be implemented in more than 2713 dif-
ferent ways. Furthermore the national law of the Member States is not
translated into the official languages of the EU. Thus, it is very difficult
for the EU institutions to watch, compare and correct implementation
measures, and it is also very difficult for governments, businesses and
citizens to locate relevant cross-national legal information.14
11
Lesmo et al. give a descriptive example by using the concept “in clear and
comprehensible manner ” taken from the Directive on Distance Contracts 97/7/EC.
The authors compare the conditions a distance seller has to fulfill to provide a
distance contract in clear and comprehensible manner under the U.K. (“clear and
comprehensible”), German (“klar und verständlich”) and the Italian (“chiaro e com-
prensible”) legal system. Finally they point out that the main foci (form or the
writing of the information must be clear and legible; information must be intelligible
by the consumer; language of the information must be that of the consumer) set to
identify a “clear and comprehensible manner” vary in all cases. See Lesmo et al.
(2005): The next EUR-Lex: What should be done for the needs of lawyers belonging
to different national legal systems?
12
Most of EU legislation is made in the form of Directives. Contrary to EU Reg-
ulations, Directives are only binding on the Member States (not directly applicable
to citizens) and usually leave some leeway as to the exact rules to be adopted.
13
On the federal and the state level.
14
The problem is not reduced to legislation. Schacherreiter analyzed two written
statements on a decision of the European Court of Justice, one of a German, one of
an English expert. Their conclusions are absolutely contrary: while the German
expert (civil law) considers the findings of the ECJ indicatory and general ap-
plicable, the English expert (common law) cannot detect a new general rule, he
rather considers the ruling of the ECJ an exception of the general rule, justified
by very specific circumstances and facts. Schacherreiter (2006): Legal culture und
europäische Harmonisierung (“Legal Culture and European Harmonization”).
Semantic Spaces and Multilingualism in the Law 137
2. “Up to Date” Approaches: XML and Ontologies
Considering all of the semantic spaces, the relationships between se-
mantic spaces and between concepts, and the inconsistency of natural
language itself, is it now possible to put the legal and the corresponding
real world knowledge into the machine? It is perhaps impossible or at
least infeasible to make the machine automatically determine the exact
meaning of legal text, but it is feasible to create machine-processible
specifications of the semantics, at least to some extent. An overview
of current approaches addressing these problems reveals two predomi-
nant keywords: XML and ontologies, most frequently connected to the
concepts “Semantic-Web” or “Web 2.0” 15 .
2.1. The Extendable Markup Language XML
The Markup Language XML has proven to be very helpful to structure
legal texts and to allocate meta-data. With regards to further automatic
processing it is a significant advantage to acquire the main features of a
document already in its preparatory phase. Moreover, XML allows for
logic notation, automated linkage and simplified visualization. Yet, it is
primarily tied to syntax and proves less suitable to represent semantics.
The level of semantics assigned to a document depends on how XML
is applied. XML is normally used to tag the implicit semantics of the
document structure only, and the tags are freely interchangeable and
do not carry the actual meaning of the document’s content. Often,
errors are cause because legal texts are drafted in complex MS word
templates incorporating many macros, and then converted into XML
files. Therefore each new element, e.g. the marking of legal definitions,
the representation of relations between different level instruments’ or
the denotation of roles would complicate the drafting of a document
and inevitably go beyond the scope of the drafter. Furthermore law is
dynamic – hence standards must enable subsequent changes.
The full potential XML offers has surely not yet been exploited,
but there are other, perhaps more appropriate technologies available.
It seems to be more useful to take XML as an ideal basis, on which
15
“The Semantic Web is an extension of the current Web in which information is
given well-defined meaning, better enabling computers and people to work in coop-
eration. It is based on the idea of having data on the Web defined and linked such
that it can be used for more effective discovery, automation, integration, and reuse
across various applications.” Hendler et al. (2002): Integrating Applications on the
Semantic Web, p. 676.
138 Doris Liebwald
further layers of semantic technologies may be established, similar to
the semantic web vision of Tim Berners-Lee16 .
Currently there are also some ambitious efforts to draft common
European legislative XML standards to be shared by all EU Member
States.17 However, those attempts will face similar problems that arose
within the N-Lex project described below. Existing legislative drafting
standards of the Member States correlate to national legislative proce-
dures. Differences in document structure, legal hierarchies etc. reflect
specific national needs and legal systems. A country-independent data
format will therefore either have to be restricted to a very simple com-
mon level or will otherwise not satisfy national requirements or even
constrain to national process models. Therefore a common standard
must be flexible enough to cover different national needs. Nevertheless,
unification on a low level will facilitate document and information ex-
change, and in areas with a high degree of European harmonization
such applications or shared tools may prove very useful. In the words of
Michael Uschold, “The more agreement there is, the less it is necessary
to have machine processable semantics.” 18
2.2. Legal Ontologies
Ontologies are knowledge models used to describe the meaning and
context of information. They allow an accurate definition of relevant
concepts and the representation of concept coherences, higher-level re-
lations as well as logic structures, and are used to specify semantics
in machine executable form (formal semantics). Their possible fields of
application in law are manifold and range from information retrieval
16
The source graphic of the oft-quoted layer model originates from
Koivunen/Miller (2002): W3C Semantic Web Activity.
17
See in particular the ONE-LEX project (ONtologies for European Laws in
EXecutable format, Prof. Sartor, European University Institute/Florence, http:
//castor.iue.it/), and Sartor (2005): The ONE-LEX project and the informa-
tional unification of the laws of Europe. A broader overview on the state of the art
is given by Biagioli et al. (eds.) (2007): Proceedings of the V Legislative XML Work-
shop (Florence 2006). See also the MetaLex Project (http://www.metalex.nl/)
and the LEXML (http://www.lexml.de/) initiative. A different approach takes the
new ESTRELLA project (European project for Standardized Transparent Repre-
sentations in order to Extend Legal Accessibility, University of Amsterdam et al.,
http://www.estrellaproject.org/). The main objective of ESTRELLA is to de-
velop a legal knowledge interchange format and to facilitate a market of interoperable
components for legal knowledge-based systems, allowing public administrations and
other users to freely choose among competing development environments, infer-
ence engines, and other tools. In the pilot applications, European and national tax
legislation of two European countries will be modeled.
18
Uschold (2003): Where are the Semantics in the Semantic Web?
Semantic Spaces and Multilingualism in the Law 139
across decision support systems to expert systems. Ontologies offer
the advantage of rendering semantics more precisely; concurrently the
nuances of relations allow for certain representation of ambiguity. The
use of ontologies overcomes linear hierarchical structures, allows the
integration of heterogeneous data sources and enables the step from
text documentation to content documentation.
There exist three main techniques for ontology engineering: statisti-
cal approaches which are less laborious but entail a certain ambiguity,
linguistic approaches whose reliability heavily depends on the appli-
cation area and which are not sufficient in the field of law, and man-
ual/intellectual methods, which – provided that there is a high degree of
enthusiasm and motivation in their engineering – offer the best results,
but are the most costly and time consuming. For most reasons it is
advisable to combine statistical, linguistic and manual methods. Since
statistical methods are more mature, subsequent manual adjustment
is less laborious. Linguistic tools may solve well-known problems like
synonymy, morphological changes of the word stem, compound words,
etc. Such tools exist in varying levels of quality, but are costly.19
Additionally, there exist several types of ontologies, which can be
roughly divided into meta-data ontologies, general ontologies to rep-
resent the world knowledge, specific domain ontologies, method- and
task-oriented ontologies, and finally representative ontologies, which
define only the frames of representation.
Methods are also diversified and range from WordNet-methods20 ,
which define concepts in natural language and go without a formal
language for the definition of semantics, to rule-based systems with a
high degree of formalization, e.g. Cyc21 , which uses millions of logic ax-
ioms, rules and other assertions to specify constraints on the individual
objects and classes. Linguistically motivated ontologies like WordNet
or in the legal field LOIS (Lexical Ontologies for legal Information
Sharing)22 are still primarily made for humans. The semantics are made
explicit in an informal manner, in natural language definitions. Direct
use of informally expressed semantics by machines is limited. For this
19
In the English language a simple word stemming may already resolve a large
part of the morphological problems. However, this does not apply to other languages.
20
See http://wordnet.princeton.edu/ and in particular Fellbaum (ed.) (1998):
WordNet: An Electronic Lexical Database.
21
See http://www.cyc.com/ (there is also a list of publications at http://www.
cyc.com/cyc/technology/pubs).
22
LOIS is a multilingual legal thesaurus with natural language definition of legal
terms based on the WordNet and the EuroWordNet (http://www.illc.uva.nl/
EuroWordNet/) technology. See the LOIS homepage at http://www.loisproject.
org/ and Schweighofer/Liebwald (2007): Advanced lexical ontologies and hybrid
knowledge based systems.
140 Doris Liebwald
reason the semantics must be hardwired into application software to
make the ontology usable for machines. But even in applications like
Cyc, automated inference to process the semantics at runtime is limited.
Cyc does not dynamically discern what content means; the meanings
of terms and how to use them are hard coded by humans.
The use of ontologies for the formalization of the law is, however, not
a new approach.23 Today there are new implementation technologies
available, which have given rise to numerous proposals and projects in
this area24 . The chosen approaches and methods are manifold, but a
“universal valid code of practice” on how to engineer a legal ontology
does not exist. Nevertheless, some critical points can be isolated.
The law is dynamic and consists of dissimilar, variable semantic spaces.
Therefore ontologies need to be flexible and dynamic and must describe
processes instead of static models. The formalization of implicit know-
ledge proved to be especially difficult. Application-oriented, specific
domain ontologies (networks of meanings) are feasible at this stage.
However, the cross-linking of different domain models and the intercon-
nection of the concept spaces of world knowledge (the world model)25
and legal knowledge (the domain models) are still substantial problems.
Within a network of semantic spaces, overlapping, conflicting or even
contradictory conceptualizations must be resolvable. Findings of related
research, especially in the areas of forensic linguistics, comparative law
and automatic text analysis, which could bridge the gap between con-
ceptualization und stored information, seem to have been put aside in
the fervor of ontology engineering but should be incorporated as much
as possible.
Furthermore it can be clearly stated that ontologies are extremely cost
and labor-intensive; they demand expert knowledge and a high level of
consistency. The quality of the conceptualizations and their relation-
23
See e.g. Hohfeld (1917): Fundamental Legal Conceptions as Applied in Judicial
Reasoning; but also the findings of Hart, Kelsen, etc.
24
An overview is given by Schweighofer/Liebwald (2007): Advanced lexical on-
tologies and hybrid knowledge based systems. More detailed is Benjamins’ et al.
(eds.) (2005): Law and the Semantic Web.
25
The underlying problem may already exist in the modeling of the “neces-
sary” world knowledge, in the “facts” (e.g. consideration of evidence, reconstruc-
tion/finding of facts, etc.). It cannot be ignored that models are always abstract
– part of reality is lost in models. A nice example can be taken from the TRACS
project. The TRACS prototype was developed about 1990 to check the consistency
and completeness of a new (Dutch) traffic regulation. It drew inter alia the conclusion
that a tram running on the tram-lane commits a traffic violation. This was due to
the fact that Art. 10.1 stated that all vehicles except those mentioned in Articles 5
to 8 should use the drive-lanes. However, the tram is not mentioned in these articles,
and the tram-lane is not a drive-lane. See Breuker et al. (2005): Use and Reuse of
Legal Ontologies in knowledge engineering and Information Management.
Semantic Spaces and Multilingualism in the Law 141
ships are of utmost importance and cannot merely be replaced by a
high number of low-quality concepts. The undertaking will be more
worthwhile if ontologies allow for reuse. Again, this requires high qual-
ity, common design and compatible technologies. Nevertheless, ontology
developers should always consider the specific needs of the intended
application area(s) and user group(s).
Finally one must consider that a determination, standardization, or
terminology normalization which is too strong or too stringent may
also emerge as sort of “semantic shackle” which compromises diversity
of language(s) and constrains further development. When dealing with
European legal texts, merely reducing languages, legal systems and legal
traditions to the highest common denominator will not contribute to a
better mutual understanding. It is of highest importance to factor in
national differences in legal language, concepts and structure. Contrary
to e.g. a biological taxonomy, a legal ontology is not language and
country independent.
3. Two Practical Examples: N-Lex and EUROVOC
Two practical examples, the new and experimental search-engine N-Lex
and the traditional EUROVOC-Thesaurus, shall demonstrate the cur-
rent problems within the EU caused by the diversity of legal traditions
and semantic spaces in the law.
3.1. The Experimental N-Lex Project26
N-Lex is an attempt of the European Publications Office to provide a
common gateway to national law of the EU Member States.27 It is an
experimental system, put online for test-use in April 2006. N-Lex allows
users to search the national legal databases of 22 Member States using
a single, uniform N-Lex search mask.
A user may choose the source country and then fill in one or more in-
put fields. This query put to N-Lex is forwarded unaltered to the search
form of the respective freely available national online-database. In the
next step, N-Lex presents the original result set or result document
in its main frame. For display of the search form and of some basic
information on the respective national database the user may choose
26
A more in-depth analysis is given by Liebwald (2007): Einheitsschnittstellen zu
Rechtssystemen am Beispiel von N-Lex (“Unified Interfaces to Legal Systems Using
the Example of N-Lex”).
27
N-Lex is maintained by the Office for Official Publications of the EC; the
application is publicly available at https://europa.eu.int/celexdev/natlex/.
142 Doris Liebwald
her/his preferred language. However, a user lacking sufficient language
abilities will not be able to formulate an adequate query or to assess
the retrieved documents.28
Information on the respective national databases (“country informa-
tion”) is poor, and the user is left with many questions regarding com-
pleteness, authenticity and timeliness of the content, the document
hierarchies and relationship of documents, or the technical function-
ing of the corresponding national systems – all of which influence the
appropriateness of a search.
The search mask offers only a few input fields,29 some of which may
or may not function depending on the country selected. The input field
for document numbers is not available for many countries, and where it
is active, the necessary input format is not clear. In the advice section
the N-Lex help-entry recommends against using the date/time span
field, “as it is . . . liable to produce zero responses”. In fact, the results
retrieved by using the date-field are not comprehensible, at least re-
garding searches for Austrian legal documents. Document numbers and
date/time-span are, however, very important and in many cases even
essential criteria for the identification of legal documents.
The original search forms of the national legal databases offer more
sophisticated search fields and search functions. Their search masks are
not only adapted to the national legal systems, the national document
structure, and the national language(s), but also to the features and
abilities of the respective technical system. Even search functions most
typically used for legal information retrieval may have been imple-
mented in very different ways. In the national systems digital infor-
mation is available in different formats and comes along with coun-
try specific meta-data. A simple, unified search mask covering all of
these different legal and technical systems nullifies many of the coun-
try specific functionalities, meta-data, and textual information (e.g.
“Länderrecht”). Regarding the N-Lex system, the general principle that
authenticity is directly related to the closeness to the original source is
true.
In its current state, N-Lex displays many deficiencies and considers
differences in the national legal and technical systems inadequately.30
However, it is still in the experimental phase and some points of criti-
cisms might become obsolete at a later time. At the very least, it offers
28
Regarding the implementation of the EUROVOC thesaurus see the next section.
29
Full text search, search in document titles, document type, document number,
date of document.
30
The EULEGIS (European Legal Information in a Structured Form, 1999-2001)
research project already identified those problems. An overview on the EULEGIS
reports is available at http://www.it.jyu.fi/raske/publications.html.
Semantic Spaces and Multilingualism in the Law 143
a single access point and has provided a first test that can be studied
and improved upon.
3.2. The EUROVOC-Thesaurus
The multilingual und multidisciplinary EUROVOC-Thesaurus was orig-
inally built for processing the documentary information of the EU in-
stitutions.31 It offers a controlled set of vocabulary covering 21 wide-
ranging fields and more than 20 languages.32 EUROVOC, however,
contains “European” concepts, with a certain emphasis on the Euro-
pean legal language and parliamentary activities. It is an effective tool
to index (European) documentary resources and to retrieve documents
indexed by this means, but its general usability in information retrieval,
especially in full text retrieval, is limited. The following two examples
shall illustrate those restrictions.
The EUROVOC-Thesaurus, which is used in various applications, is
also used in EUR-Lex, the gateway to EU law33 . Legislative documents
in EUR-Lex34 are indexed according to EUROVOC, and the simple-
search form allows a keyword search restricted to those EUROVOC
descriptors. However, EUR-Lex contains a huge amount of documents35
and only the upper levels of EUROVOC may be selected from the classi-
fication schema provided. Therefore, the use of EUROVOC descriptors
usually results in a set of a few hundred, sometimes even of a few
thousand documents. Of course, the system allows the user to refine
the search by adding additional keywords, by selecting the document
type or the date/time span. Alternatively the user can use the trial and
error method and enter various EUROVOC descriptors from a deeper
level. Admittedly, most of these choices assume additional knowledge
about the document or about EUROVOC, which might not be available
at this stage, or simply do not reduce the amount of result documents
to a manageable number.
The implementation of the EUROVOC-Thesaurus in the experimental
N-Lex application demonstrates the limits of such thesauri more obvi-
31
EUROVOC is maintained by the Office for Official Publications of the EC and
available at http://eurovoc.europa.eu.
32
EUROVOC consists of more than 6000 concepts with a maximum depth of 8
levels. The 21 fields of the first level split up into some 130 micro-thesauri.
33
EUR-Lex is also maintained by the Publications Office and available at http:
//eur-lex.europa.eu/.
34
The ECJ case law is, however, not indexed by EUROVOC descriptors but by
the case law directory code.
35
According to the EUR-Lex FAQ (point 2.2.) “it includes some 400000 references
in several languages, 1400000 texts in total. An average of 15000 documents are
added each year.”
144 Doris Liebwald
ously. Within N-Lex, EUROVOC is intended to support full text search
capabilities. National legal documents (legal documents produced by
the Member States) are generally not linked to EUROVOC descriptors.
Therefore a corresponding keyword search cannot be established. Users
may either search and select a suitable descriptor in the target language
or search for a descriptor in her/his preferred language and ask for the
translation into the target language.
Due to the fact that EUROVOC uses European terminology, it is not
convenient to search or to index national legal documents, even if those
texts are partially based on European input requirements. Each Member
State has its own legal tradition, legal system and legal terminology. A
national indexer would in many cases choose different descriptors based
on national legal traditions and interpret EUROVOC descriptors in a
different way. Additionally, EUROVOC descriptors do not necessarily
appear in the relevant national legal texts. On the contrary, more spe-
cific concepts, variants of concepts and specific national legal language
terms are used within national codes and case law. Additionally, EU-
ROVOC appears to be based to some extent on literal translations
not indicating the exact implied meaning. European as well as lit-
erally translated concepts usually don’t correspond to the terms and
phrases a national user of a legal database would use naturally.36 and
the German translation “persönliche Daten”. Even though all German-
speaking lawyers will understand this translation, the Austrian legal
language uses the concept “pesonenbezogene Daten”. Searching the Aus-
trian law with the search term “persönliche Daten” will retrieve result
documents, but not the relevant ones. It will mainly retrieve those texts
containing the terms “persönliche” and “Daten” beyond the meaning of
“peronsenbezogene Daten”.
Once the N-Lex user has chosen a fitting EUROVOC descriptor,
the system sends the search question to the corresponding national
databases. Most of these national databases execute a simple string
search. Specific technical parlance, morphological changes, derivations,
compounds, synonyms, polysems, etc. are therefore not considered.37
Only those documents containing exactly the same character string are
36
E.g. the German concept “Datenschutz” is a prominent concept in German and
Austrian law, but is not covered by EUROVOC. Austrian and German lawyers will
connect a specific concept regarding the protection of personal data processed by
electronic means to the term “Datenschutz”. On the other hand, EUROVOC offers
the English concept “data-processing law” with the German translation “Datenrecht”.
What is “Datenrecht”? A test search concluded that “Datenrecht” is never used in
Austrian or German legislation. EUROVOC also offers the concept “personal data”
37
There are of course language and provider dependent differences (additional
functions offered by the corresponding national provider will influence/better the
result set).
Semantic Spaces and Multilingualism in the Law 145
sent back to the user.38
This accumulation of shortcomings produces incomplete result sets with
low-recall, low-precision or empty result sets. This is contrary to the
use of EUROVOC within EUR-Lex. Using EUROVOC in the way in
which it is implemented in N-Lex wrongly assumes that all agents use
exactly the same wording to state the same thing and that the same
terms always have the same meaning. The correct conclusion is not that
EUROVOC is generally a bad thesaurus of low quality but that it is
being used for purposes other than originally intended and has not been
adapted to such uses.
3.3. Excursus: The Semantic Spaces of the Persons
Subject to the Law
To make the law more easily accessible for the persons subject to the
law is an ambitious goal. The descriptions of the LOIS and the N-Lex
projects stress the target to enable easier access to legal information
for professional users as well as for laypersons. Both use the example
of a family migrating to another EU Member State and searching for
information regarding taxes, social insurance, childcare, etc. In fact
neither LOIS nor N-Lex solve or even support such questions, at least
at their current state. The semantic spaces of laypersons significantly
differ from the semantic spaces of the lawmaker or legal expert. Citizens
will use other concepts, other questions, and will have other information
needs.39 Usually they will not be able to retrieve relevant bills from legal
databases or be able to identify the relevant articles therein. They will
not understand the original text of a bill or the legal language, and they
will need some complementary explanations in their common language.
It is even more unlikely that citizens will understand the concepts, struc-
ture and language of a foreign legal system. It is not sufficient to enable
easier access to the law by offering a choice of life situations from which
a citizen may select, or by semantic translation of the common language
information request, and then the presentation of the original legal texts
to the citizen. This shall not, however, prevent establishing links to the
38
EUROVOC offers e.g. the English concept “protection of communications” and
the corresponding German concept “Brief-, Post- und Fernmeldegeheimnis”. This
string is never used in Austrian legislation, even though the concept does exist.
Searching the German law brings up 22 hits.
39
Significantly there is a “plain language guide to Eurojargon” available in 20
languages at the Europe-server (http://europa.eu/abc/eurojargon/index\_en.
htm). According to this site the guide was developed because euro-jargon can be
very confusing to the general public. The language guide and the attached glossary
contain in sum about 300 concepts and short descriptions, but the concepts are not
linked to further information and the descriptions do not solve real life questions.
146 Doris Liebwald
original legal sources where appropriate, but citizens primarily need
citizen-tailored texts and issue-related information. In addition, citi-
zens will appreciate supplementary information such as the responsible
departments, contact data or references to further appropriate services.
The approach to develop one combined system that serves experts and
citizens is perhaps too ambitious and idealistic; such a system runs the
risk of being a confusing compromise instead of serving both in the best
possible manner.
4. Conclusions
It is an interesting matter that since the classic “Handbook of Legal
Information Retrieval” edited by Jon Bing was published in 198440 ,
improvement in legal information retrieval has not seen any major ad-
vancement. Quite to the contrary, information overload and increased
demand for cross-national and cross-lingual legal information has ampli-
fied the basic problems. The handbook already points out many of the
shortcomings a lawyer typically has to struggle with when searching
for relevant legal documents. About 20 years later, authors such as
Luuk Mathijssen, Peter Wahlgren and Doris Liebwald41 as well as the
common user still struggle with the very same problems. Legal informa-
tion retrieval systems still do not represent legal structural knowledge,
user friendliness regarding search strategies and input formats is lack-
ing, and information about system functions and information content
(completeness) is often not sufficient. Also, continuity, representation
of time layers and consolidated versions are inadequate and different
user situations and information needs are disregarded. Finally, finding
the correct search terms is a game of chance, language approximation
is minimal and even simple linguistic tools are missing.
Nevertheless, current developments in new technologies supporting
communication in human/human, human/machine and machine/machine
relations are promising. A shift from simple full text and keyword search
to more sophisticated semantic querying appears to be within reach.
Hopefully, these technologies will be used to serve the fundamental
principles of accessibility and intelligibility of the law.
40
A revised version is freely available at http://www.lovdata.no/litt/hand/
hand-1991-0.html.
41
See Matthijssen (1999): Interfacing between Lawyers and Computers; Wahlgren
(1999): The Quest for Law; Liebwald (2003): Evaluierung juristischer Datenbanken
(“Evaluation of Legal Databases”) and Liebwald (2005): An Evaluation of “New EUR-
Lex”: All Tasks Achieved and All Problems Solved?
Semantic Spaces and Multilingualism in the Law 147
References
Benjamins, R. et al. (eds.): Law and the Semantic Web: Legal Ontologies,
Methodologies, Legal Information Retrieval, and Applications. Lecture Notes
in Computer Science Vol. 3369, Springer, Berlin et al. 2005
Biagioli, C. et al. (eds.): Proceedings of the V Legislative XML Workshop (Florence
2006). European Press Academic Publishing, Florence 2007
Bing, J. (ed.): Handbook of Legal Information Retrieval. Elsevier, New York 1984
Breuker, J.A. et al.: Use and Reuse of Legal Ontologies in knowledge engineering
and Information Management. In: Benjamins (eds.): Law and the Semantic Web.
Lecture Notes in Computer Science Vol. 3369, Springer, Berlin et al. 2005, 36-64
Bydlinski, F.: Juristische Methodenlehre und Rechtsbegriff2 (“Legal Methodology
and Nomen Juris”). Springer, Wien 1991
Fellbaum, C. (eds.): WordNet: An Electronic Lexical Database. MIT Press, MA
1998
Galinski, C.: Wozu Normen? Wozu semantische Interoperabilität? (“Why Norms?
Why Semantic Interoperability?”). In: Pellegrini/Blumauer (eds.): Semantic
Web: Wege zur vernetzten Wissensgesellschaft. Reihe X.media.press, Springer,
Berlin et al. 2006, 47-72
Hendler, J. et al.: Integrating Applications on the Semantic Web. Journal of the
Institute of Electrical Engineers of Japan Vol. 122(10), October 2002, 676-680
Hohfeld, W.N.: Fundamental Legal Conceptions as Applied in Judicial Reasoning.
The Yale Law Journal Vol. 26/8 (June 1917), 710-770
Koivunen, M.-R./Miller, E.: W3C Semantic Web Activity. In: Proc. of the Semantic
Web Kick-off Seminar (Helsinki 2001). HIIT Publications, Helsinki 2002/1, 27-43
(available at http://www.w3.org/2001/12/semweb-fin/w3csw)
Lesmo, L. et al.: The next EUR-Lex: What should be done for the needs of
lawyers belonging to different national legal systems? In: Proc. of the JU-
RIX EU-Info Workshop. Brussels 2005 (available at http://www.di.unito.it/
\ensuremath{\sim}guido/PS/jurixWorkshopPaper.pdf)
Liebwald, D.: An Evaluation of “New EUR-Lex”: All Tasks Achieved and All Prob-
lems Solved? MR-Int 3/2005 (European Media, IP & IT Law Review), Verlag
Medien & Recht, Vienna, 156-160
Liebwald, D.: Einheitsschnittstellen zu Rechtssystemen am Beispiel von N-Lex
(“Unified Interfaces to Legal Systems Using the Example of N-Lex”). In:
Schweighofer et al. (eds.): Aktuelle Fragen der Rechtsinformatik 2007. Boorberg,
Stuttgart et al. 2007 (in print).
Liebwald, D.: Evaluierung juristischer Datenbanken (“Evaluation of Legal Data-
bases”). Verlag Österreich, Vienna 2003.
Liebwald, D.: Semantische Räume als Strukturhintergrund der Rechtsetzung (“ Se-
mantic Spaces as Structural Patterns of Legislation”). In: Bildungsprotokolle der
Kärtner Verwaltungsakademie zu den Klagenfurter Legistik-Gesprächen 2006,
Klagenfurt 2007 (in print).
Matthijssen, L.: Interfacing between Lawyers and Computers: An Architecture for
Knowledge-based Interfaces to Legal Databases. Kluwer Law International, The
Hague et al. 1999
Sartor, G.: The ONE-LEX project and the informational unification of the laws of
Europe. In: Proc. of the Klagenfurter Legistikgespräche 2005. Bildungsprotokolle
Vol. 12, Kärntner Verwaltungsakademie, Klagenfurt 2006, 193-202.
148 Doris Liebwald
Schacherreiter, J.: Legal culture und europäische Harmonisierung (“Legal Culture
and European Harmonization”). Juridikum 2006/1, Verlag Österreich, Vienna,
17-21
Schweighofer, E./Liebwald, D.: Advanced lexical ontologies and hybrid knowledge
based systems: First steps to a dynamic legal electronic commentary. AI&Law
Journal Vol. 15 (February 2007), Springer International, NL
Uschold, M.: Where Are the Semantics in the Semantic Web? AI Magazine 24(3):
Fall 2003, AAAI Press, Menlo Park, 25-36
Uschold, M./Gruninger, M.: Ontologies: Principles, Methods and Application. The
Knowledge Engineering Review 11(2)/1996, Cambridge University Press, 93-115
Wahlgren, P.: The Quest for Law. Jure AB, Stockholm 1999.
Warta, P.: Zauberworte – Verwandlungen des Gleichheitsgrundsatzes in der
Judikatur des österreichischen Verfassungsgerichtshofes (“Magic Words – Meta-
morphoses of the Principle of Equality in the Legal Practice of the Austrian
Constitutional Court”). In: Schweighofer et al. (eds.): Aktuelle Fragen der
Rechtsinformatik 2005. Boorberg, Stuttgart et al. 2005, 576-583.