Methodological Perspectives for Legal Ontologies Building:

an Interdisciplinary Experience

Danièle Bourcier / Mélanie Dulong de Rosnay / Jacky Legrand
CERSA – CNRS
University of Paris 2 / CERSA – CNRS
University of Paris 2, Medialive / University of Paris 2

Abstract

Legal ontologies design raises knowledge management but also jurisprudence issues. To which extent is it possible to establish common concerns between technical means and theoretical constraints?

An interdisciplinary network gathering lawyers, knowledge engineers, cognitive scientists and linguists has been set up to bridge a gap between various approaches. Working groups have been created: Usage and research on fundamentals (legal theory), European Community legal concepts (cognitive approach), Text-based ontologies (applicative approach).

Extensive collaboration allowed participants to experiment, share and evaluate concepts, tools and methods. The groups finally agreed that the cognitive approach is the accurate level to handle most methodological differences.

After a brief presentation of the network methodology, we give two convergent results of our working groups on the feasibility of legal ontologies. First, the definition of a meta legal language handling fundamentals (acts, actors, conditions…) contributes to a shared understanding of legal knowledge between disciplines and jurisdictions. Second, ontological alignment concerns not only knowledge management but is also the core of the judge’s activity.

Keywords

Ontologies, legal information systems, knowledge management, methodology, cognition

1. Introduction

Legal ontologies cover two types of approaches. On the one hand, generic ontologies trying to reflect theories of law produce upper-level ontologies specific to the legal domain but hardly applicable by the practitioners. They are usually built from a conceptual point of view and follow a top-down process from concepts to texts. On the other hand, applicative ontologies answering to professional needs often miss the specificity of legal knowledge but are efficient for information retrieval. They are extracted bottom-up from very large textual corpus.

These two types of ontologies are built by different scientific and professional communities. Generic ontologies are rather made by legal academics for heuristic purposes and represent another way to describe jurisprudence, while applicative ontologies are produced by knowledge management engineers to reduce legal language complexity and favor storage and access to legal documents. It seemed interesting to combine best practices from these various communities experiences. In this paper, we describe a research and development project launched by the Department of Communication and Information Science and Technology of French National Center for Scientific Research (CNRS) which aimed at supporting the research on methodological benefits of an interdisciplinary action for legal ontologies building and issues. After a brief presentation of the group framework, we will define some guidelines obtained from results.

2. Methodology

The Specific Action "Legal ontologies and legal language" proceeded over one year, gathering 25 active researchers from several laboratories (law,legal science, philosophy, logic, linguistics, computer science, cognitive science, knowledge management, semantic web) and institutions (Constitutional Court, local authorities). Working groups have been dedicated to various aspects of the ontological issues: Search and use of fundamentals (legal theory), European Community legal concepts (cognitive approach), and Text-based ontologies (applicative approach). Results have been presented and evaluated by three international experts at a closing symposium[i] at the end of 2004 in Paris.

A theoretical subgroup undertook preliminary work to evaluate the difficulty and the interest of the construction of ontologies. , the search and use of fundamentals of the law, and a general reflection on legal ontologies purposes and raison d’être. Is law a specific ontological domain? How can jurisprudential scholars and legal knowledge engineers enrich each other?

Another subgroup worked on European community law Directives. European Directives illustrates the need to have a common representation of concepts in order to translate accurately shared requirements into national legislation during the transposition process. The multilingual aspect of EC Directives transposition offers perspectives in terms of translation applications, comparative legal terminology studies, and the evaluation of compliance, correspondence and harmonization between national laws[ii], as it was described by a member of the group who presented his prior work[iii]. What does ontology software bring to legal knowledge understanding?

The working method included multiple face to face meetings with members lectures, presentations and debates and electronic exchanges. A collaborative virtual workspace[iv] (wiki) will be settled to continue the experiences exchange. Prospective goals investigation has been preferred to a state of the art mining and interpretation.

Substantial tests have been undertaken to illustrate some hypothesis of the participants.

. The corpus chosen by the experts corresponds to the group’s lawyers field of competence: social law around the definition of worker[v] and copyright law around the definition of author, a part of the equivalent for copyright being “author’s law” in civil law countries.

This experience emphases the layer model of copyright international block of normativity: the 2001 Copyright Directive is supposed to implement the 1996 World Intellectual Property Organization Copyright Treaty.

3. Results

3.1 Resources sharing

Legal and software resources were shared among the experts, leading to the production of hypotheses and a comparative reflection between tools and practices. For instance, the parsing results of the two European Directives with Linguistic Craft Workbench and Ontology Craft Workbench tools[vi] revealed at their intersection the skeleton of a Directive beyond domains characteristics. The terms extracted from the Copyright Directive allowed a re-organization of the intuitive categories that had been built by the legal expert, leading to new mappings and a renewed legal analysis of the European text. Also, a UML-type diagram was produced reflecting the conceptual and relational restructuration that was allowed by the non-linear reading of the Directive.

3.2 Definition of a common language

The goal of the copyright ontology is to gather terminological material for access to copyrighted works automated management[vii]. At the intersection between cyberspace law and AI, digital rights expression languages (RELs) regulate the creation and the exchange of information and works on the web. This legal knowledge embedded in intelligent agents needs to be represented in human-readable applications and interfaces. The definition of a metadata system for works according to their copyright status aims at a common understanding for lawyers, users and technology applications designers of computed actions and legal conditions requested to share works. The elaboration of a meta Rights Expression Language based on a legal ontology is necessary to ensure semantic interoperability between humans and machines to describe works and represent contracts.

3.3 Ontologies alignment for applications interoperability

The further use of an ontology editor, the Differential Ontology Editor ( and its associated method[viii] for semantic differential definition of the ontological resource terms is helping the legal expert to built a knowledge base. Terms and classes of a copyright legal ontology are then being mapped to existing top-level and domain ontologies, toward alignment and semantic interoperability[ix]. Shared terms for alignment purposes confirm the existence of some legal fundamentals, concepts that are common to lawyers, knowledge engineers, legal philosophers and semantic web professionals. Entities and their attributes (quantity, quality) are Work, Agent, Act, Have, Transfer, Conditions (permission/prohibition, authorization/requirement), Time, Location…

3.4 Legal fundamentals, a construction based on points of view and ontology goals

The whole group agreed that the research of fundamentals is a prerequisite for ontologies building. However, it cannot be achieved without taking into account the various points of view on law: (see fig. 3.4.a)Any construction shall be located in a 3-dimensional space deriving from these settings:

legalpoint of view: legal norms are abstract, mandatory general rules, to be differentiated from technical and causal rules. Ontology building concern will be to describe the variables of the case;

, actionpoint of view: legal norms are legal actions performed by the competent authority. Ontology building concern will be to take into account the performative aspect of the legal speech act and the “performer” legal discretion[x];

, logicalpoint of view.: legal norms are obligations, permissions and prohibitions utterances. Ontology building concern will be to cope with logic formalisms and transformations.

Another 3-dimensional space with three oriented axes corresponds to the ontology goals perspective (see fig. 3.4.b).

- more or less linguisticapproach: results should be seen as lexicographic ontology or computational modeling of the meaning;

- more or less generalizingapproach: legal ontology must cope with various levels of details. Search of upper-level concepts (i.e obligation) can be opposed to a descriptive ontology (i.e to decide obligations attached to an object);

- more or less decision-makingapproach: ontology has to characterize properties (essential, accidental...) as "decision-making entities". There is a re-conceptualization for practical applications.

3.5 A walk among theoretical contributions converging to legal fundamentals

As we can see on the picture fig 3.5, the Working Group member’s contributions are eclectic but converging. Looking like a Chinese Encyclopedia, lectures included studies of historical Legal Knowledge Systems, of Leibniz, of road signs, of cybernetics and law, of vagueness, of Civil Code consistency, of norms inventory and typing, of administrative case law, of road accidents and victims compensation.

The recent history of the Legal Knowledge Systems[xi] reveals relationships between conceptualization in the legal field and ontology building. However formalization (even moremathematization) of law is an old quest with philosophical roots. The study of Leibniz’s system of philosophy and legal philosophy (a reconstructed form given by the secondary literature[xii]) remains a prerequisite, such as Bentham[xiii] and Aristotle[xiv] for legal fundamentals (see section 3.4).

Several points of convergence can be highlighted from these contributions:

- the hierarchical layered structure, from metaphysics (ontology, logic) to elements (legal norm, principles) through the concept of law and its structure (legal system);

- the dichotomy between a static description and a polemic description (lawyers’ reasoning, fact finding process);

- the incomplete mathematization missing the really primitive axioms or rules in a less complex environment.

As a lawyer, a philosopher and a mathematician, Leibniz worked on uncertain events. One example of uncertainty and proportionality found in the French highway regulation[xv] aims at improving the compensation for the victims. This regulation study reveals an interesting opposition between the liability founded on the principle of primacy (pedestrian-driver) and the liability founded on the principle of proportionality (driver-driver).

The road signs, non-textual representations of the Highway Code (code de la route) introduce a legal icon language. The drivers' actions rely on three interwoven components: road contexts, icons and texts, highlighting complementarity and substitutability of words and pictures to define categories, objects and actions.

Another type of convergence between philosophy and computer science has been detected through the analysis of the logical consistency of the French Civil code. Such a methodology of questioning must be the constant rule during legal modeling: which goal, which tool, which corpus, which method, which computing.

First, it seems impossible to produce either a taxonomy of norms or a taxonomy of definitions (ontology by differentiation) because they are overlapping. Second, vagueness is essential to the legal definition functioning. On the one hand, vagueness deserves to be an object of study apart from the usual deductive inference investigations. On the other hand, the existence of categories waiting to be instantiated is taken for granted. Unfortunately, when attempting at modelling a concrete case of dispute, one discovers that the distinction between categories and instances is blurred.

The final convergence point is that the syntactical ontological approach (coding what the things are) is opposed to the institutional legitimacy of what the legal things are.

3.6 Guidelines for legal ontology building

If law is handed over to knowledge engineers in order to achieve ontologies, and if these ontologies are handed over to lawyers in order to implement law, neither the law nor the ontology will be enriched. At this level, we distinguish two types of engineering: legal operational engineering and legal cognitive engineering.

- Legaloperational engineering can be considered as closed and static (see fig 3.6.a). It proceeds from a bottom-up extraction from textual corpus after a linguistics parsing. The representation process smoothes and rubs out the evolutionary and open characteristics of law.

- Legal cognitive engineering tends to include a dynamic representation (fig 3.6.b), adding three dimensions to the previous figure: time, context and goals. First, a legal ontology is able to act upon the standard evolution, reinforcement or questioning. Second, context may specify the available coded information while potential circumstances can be recorded. Third, a legal ontology is mirroring the legislator teleological aspects: its use has to take the implicit goals into account while the explicit goals must be coded into the information.

3.7 Properly designed ontologies for an overall benefit Conclusion

The previous analyses, propositions, suggestions and recommendations may lead to a better understanding of the ontology building challenges and an increasing interest for operational and properly designed ontologies.

From a methodological point of view, properly designed ontologies will help to achieve operations on legal matter in order to reduce its constituent indetermination. From a practical point of view, computer and law research as a whole will benefit from properly designed ontologies. More precisely, when one considers the decision-making process during the legal activity, the main technical supports come from legal databases (keywords mapping), from academic and doctrinal literature (prototypical use cases) and key precedents (hard cases).

A closer examination (see fig 3.7) reveals that to store, to model, to integrate and to differentiate are the key processes for Information Technology in legal domain. Ontologies building and use are an asset for better indexing, integration for use cases modeling, hard cases differentiation.

FConclusion and further research

In this work, we tried to define some common language between disciplines constraints and ontologies building methods. Thanks to interdisciplinary networking methodology, two main concepts have emerged from our experiences and debates: the search of metadata and means of alignment as operational paths between generic and applicative ontologies and between legal practice and knowledge management.

The working group activities will continue epistemological insight and legal practical developments within a collaborative workspace environment. The recently created French Legal Information Institute ( will host this environment to facilitate access to private and public legal databases.

Acknowledgments

The authors are grateful to all the fellow members of the Specific Action and to the invited speakers of the closing symposium who accepted to evaluate the group’s results: Giovanni Sartor, Aldo Gangemi and Joost Breuker, whilst taking full responsibility for the views and ideas presented in this paper.

This research has been supported by the Pluridisciplinary Network on Law and Information Systems of French National Center for Scientific Research (CNRS).

i Proceedings of the Conference on Legal Ontologies, 2004

December 3rd, Paris.

ii An ontology around the concept of worker has been

described in Sylvie Després and Sylvie Szulman,

"Construction of a Legal Ontology from a European

Community Legislative Text", Jurix 2004 Proceedings.

iii Aldo Gangemi, Alessandra Prisco, Maria-Teresa Sagri, Geri

Steve, Daniela Tiscornia, “Some ontological tools to support

legal regulatory compliance, with a case study” in Lecture

Notes in Computer Science : OTM 2003 Workshops, Volume

2889, 2003, Springer Verlag, Heidelberg.

iv Bruno Bachimont, Antoine Isaac, Raphaël Troncy,

“Semantic Commitment for Designing Ontologies: A

Proposal”, in Asuncion Gomez-Pérez and V. Richard

Benjamins, editors, 13th International Conference on

Knowledge Engineering and Knowledge Management,

EKAW’2002, volume LNAI 2473, pages 114-121, Sigüenza,

Spain, October, 1-4 2002. Springer Verlag.

v Danièle Bourcier, “Institutional pragmatics and legal

ontology - Limits of the descriptive approach of texts”, in:

Pompeu Casanova and alii (eds), Semantic Web and Legal

Ontologies, Springer Verlag, 2005, (to be published).

[i]Proceedings of the Conference on Legal Ontologies, 2004 December 3rd, Paris.

[ii]Alexander Boer, Tom van Engers, Radboud Winkels, “Using Ontologies for Comparing and Harmonizing Legislation”, ” in Proceedings of the 9th International Conference on Artificial Intelligence and Law (ICAIL 2003), ACM Press, N.Y. 2003, pp. 60-69.

[iii]Aldo Gangemi, Alessandra Prisco, Maria-Teresa Sagri, Geri Steve, Daniela Tiscornia, “Some ontological tools to support legal regulatory compliance, with a case study” in Lecture Notes in Computer Science: OTM 2003 Workshops, Volume 2889, 2003, Springer Verlag, Heidelberg.

[iv]L. Mommers, “Application of a knowledge-based ontology of the legal domain in collaborative workspaces” in Proceedings of the 9th International Conference on Artificial Intelligence and Law (ICAIL 2003), ACM Press, N.Y. 2003, pp. 70-76.

[v]An ontology around the concept of worker has been described in Sylvie Després and Sylvie Szulman, "Construction of a Legal Ontology from a European Community Legislative Text", Jurix 2004 Proceedings.

[vi]Acknowledgments to Christophe Roche, University of Savoie,

[vii]Melanie Dulong de Rosnay, “Cognitive Interfaces for Legal Expression Description – Application to Copyrighted Works Online Sharing and Transaction”, in D. Bourcier (ed.), Legal Knowledge and Information Systems. Jurix 2003: The Sixteenth Annual Conference. Amsterdam: IOS Press, 2003, pp. 121-130.

[viii]Bruno Bachimont, Antoine Isaac, Raphaël Troncy, “Semantic Commitment for Designing Ontologies: A Proposal”, in Asuncion Gomez-Pérez and V. Richard Benjamins, editors, 13th International Conference on Knowledge Engineering and Knowledge Management, EKAW’2002, volume LNAI 2473, pages 114-121, Sigüenza, Spain, October, 1-4 2002. Springer Verlag.