DEXA Workshop LISA, Legal Information Systems and Application Munich, September Ontology driven Access to Legal Information

Transcription

1 DEXA Workshop LISA, Legal Information Systems and Application Munich, September 2001 Ontology driven Access to Legal Information Daniela Tiscornia Istituto per la Documentazione Giuridica del CNR Abstract From the analysis of a research project supported by government funding currently in progress in Italy, a few criteria of describing legal information on different levels are presented. The non-jurists users, who the public nature of the project regards as privileged, are not able to explain clearly the information they need, therefore particular attention should be paid to the description of the content. Examples of how techniques of finding relevant information turn out to be inefficient if not linked to a homogenous and conceptual organization based on theoretical models and formal criteria are given. Thus, the need of an ontology-grounded model, further useful in more complex applications (multilingual access, normative comparison and systems of legal drafting). 1. Introduction The Norme in rete (Law in the net) project (NIR), launched in 1999, has now reached its second phase 1 ; involving the most important Italian institutions (the Chamber of Deputies, the Senate, the Ministry of Justice its final goal is to create a portal which, through a single and simple user interface, allows research on all the documentation of normative interest published free on Internet, particularly by institutional sites. 2 The project aims to solve a double-sided problem. On the first place it should allow free access to normative data which, as today, is only partially available in fragments on Internet and 1 The project is mentioned in the 1999 E-Government Plan ( 2 Report on Il Progetto norme in rete, Roma, spread out in several sites; 3 the second purpose of the program is to rationalize information tools through standard methodology of editing and distributing data, which will lead to direct access to normative information by reaching the sites of law makers (and distributors) themselves; in fact, the project provides codification standards of the typologies of sources, identifiers (URN: uniform references notation), structure, links and metainformation. Based on the study of feasibility edited by IDG 4, the project went on to its operative phase in 2000: proposals on all aspects to be developed (Document Type Definition, metainformation, Urn, ) and an experimental portal was opened ( at present the project is focused on the experimentation and validation of the DTD to be adopted for the structuring of legal texts to be distributed in the net. This article describes the results obtained in the definition of the methodology for describing documents, from both formal and content aspects, only the solutions for the latter; such analysis will be the starting point for a discussion on the possible limits and extensions; examples of the current methodology as well as of that in prospect will be given. For a complete description of project, one should refer to the Study of feasibility. 2. Metainformation for NIR 2.1. What we mean by metainformation 3 EC legislation is available only for short periods from publication ( Italian legislation is free limited to last enacted statutes ( regional legislation is available in the region institutional sites, most of which charge fees; sector legislation is distributed by specific sites, professional organizations, etc.; the largest legal information systems (the Italgiure/Find System of the Court of Cassation, and Guritel, the on-line version of the Official Gazette edited by the State Press) as today, charge fees. 4 published in Informatica e Diritto, ESI, Naples, issue n. 1, 2000.

2 Metadata consists of all necessary elements to describe a resource, whereas by resource in NIR we mean the source of law (statutes, decisions, contracts, administrative acts, circular letters etc.). Compared to other information sources, legal documents show lower ambiguity in their structure and meet precise formal requirements, in many cases, foreseen by the norms themselves. A wider meaning of metadata therefore includes all the elements meant to identify the act, besides those meant to give information on the semantics. Regards this latter aspect a few choices have been made, conditioned by a number of factors: the need to adopt standard semantics (for example, classification schemes adopted by law makers, Dublin Core, etc.), impossibility of adopting descriptive models too deep and complex, unable to be applied to the large corpus of the existing legislation, Basically the following have been defined: two classes of information, a structural and a functional one; two levels of description granularity, one regarding the whole text, the other regarding the basic partitions (a section or paragraph of an article) and two locations of information: one inside the DTD model which structures the text (DTD norma) and the DTD Meta, linked to the text segments through the URNs. The DTD norma contains: Information on the legislative text such as space/time location, typology, partition based on criteria of hierarchy, indexing, etc.; Inline elements, that is, elements regarded as semantically relevant regardless of their position in the text (references, definitions, subjects, organizations, dates and places). The DTD Meta provides: 15 elements selected by Dublin Core; General information on the normative text (validity and period in force; preparation steps; notes; free information from the editor); Description of the functional aspects of the text. The last class of elements codifies most of the information on the contents and presupposes the recognition of normative entities (the dispositions) inside the text, which are not necessarily the same as the structural entities. There is, therefore, a double view of the text, both as a document and as a collection of dispositions which makes it possible to understand several entities in a functional description, or assign more than one description to a single entity, establish relationships among the entities and so on. The regulative function of the dispositions is based on a theoretical model drawn up in[4], which allows an analytical classification Functional classification of dispositions In the model of Biagioli the segments of legislative texts are analyzed from the point of view of their illocutionary function in order to give the disposition, seen as meaning of linguistic expression, a well-defined ruling function. The connection between the structural and content aspects is based on the methodology of legislation science, which suggest textual standard structures corresponding to the categories of dispositions (prescriptive, constitutive, organizational, etc. In the DTD meta, the classes of dispositions are considered as a list of predefined values of the element disposition, where each predicate disp has as arguments marked text strings (pos) which instatiate classes of entities involved in the regulation (subject, object, action ): on a ontological level, they represent the entity categories involved in the relationship established by the disposition. The model also defines a number of elementary metarelationships among the classes of dispositions on which to make inference in order to improve descriptive skills, for instance,the correlation between the right/duty categories which makes it possible to describe the legal relationship from the double point of view of the actor (e.g. entitled) and the (bearer); other meta-relationships which can be used are those between prescriptions and procedures, between definitions and prescriptions, between obligations and sanctions.[23] The need of an ontology-based description model The documents XML resulting from the above described DTDs will therefore present two series of semantic information: inline elements inside the text and, outside the text, the regulative characteristic of each disposition. Between the two classes of elements there are some overlapping aspects, which involve the risk of a nonhomogeneous conceptualizing of the domain, carried out on two different levels. The boundaries between information and metainformation, between textual and conceptual level should therefore be defined by a knowledge model which: establishes the relationship between the inline terms (textual elements) and the concepts they express defines the relationship among the categories involved in normative types and the relevant elements in the text defines the entities and the relationships placing them in a formally homogeneous model, that is, ontology-based one. 3. The theoretical foundations of an ontology-based model 3.1. The formal ontology As can be seen in [24], it is difficult to define criteria to differentiate ontologies: is not so much a difference in

3 types of legal knowledge distinguished in the ontologies but it is a differences in priorities between these knowledge types. In this case, the ontology should allow the description of dispositions and the categories of the entities involved; the level of detail will be lower than that in the recently developed legal ontologies[3, 14, 21,22, 26] whose structures are more complex and organic as they aim to conceptualize not only knowledge, but also the relationships, tasks, thought dynamics In our case, the ontology should simply provide a formally correct specification to the conceptualization of the legal domain, already and deeply developed by jurisprudence. Domain ontology should be therefore the specialization of a Formal Top Ontology [7, 19] which provides criteria of formal correctness. The top ontology is built by defining the objects of the world and properties and the relations used to speak about such objects, that is, metaproperties 5, that allow to organize in a formally valid way the entities into types, categories, roles and attributes. Only the first two are relevant for the construction of taxonomies in which, classify concrete and abstract objects. 3.2 How to endow the top ontology with a law ontology: To specialize the top ontology into a domain ontology, one must define the ontological level on which the law lies. If we keep the distinction between actual and ideal world, we should consider a physical, biological, functional and intentional level on which, to represent the real world; and an ideal level on which to represent the world as the law sees/considers it, including the entities created by the law. There is a dependency relation between levels, in the way that the objects of the law lie on the physical and functional level, whereas the subjects of the law lie on the physical, functional, biological and intentional level (mental normative attitudes), as well as a social level, where the subjects seen as legal agents become roles discriminating subsets of the person entity (e.g. set affected by roles as under_age, mental illness ). In the top ontology a general agent role is introduced, valid for all entities, which depends on the possibility of carrying out an action; the legal agent role depends on the existence of a separate entity (legal system), which provides legal effect to its action. The role of legal agent is valid for 5 Such meta-properties [7] are: identity (I), rigidity, dependence (D), the relationship part/whole and that of Unity (U). According to the metaproperty one may distinguish the ontology classes, the most important being the Types, provided with the properties which carry identity (+I) and are rigid (+R). The categories are rigid (+R), but do not carry identity(-i) (there are a few and general, the most important being the spatial/temporal location). The roles are anti-rigids and dependent (on properties which carry Identity) and may be formal and material. The attributes are neither rigids nor dependent and do not carry identity either. three types of entities: persons, organizations, and countries. Besides the persons, the corporate bodies are legal subjects, which are a part of the social organizations type. The events are occurrent disjoint objects and depend on the continuant objects, even though they are part of the concrete objects; the natural facts may be distinguished from the legal effects role as natural facts to which the law gives relevance (e.g. birth, death), involuntary facts (e.g. unpremeditated illicit facts), voluntary facts, which the large category of the acts consists of : a relationship with the traditional partition of doctrine into three: facts, actions, legal transactions may be established; the latter may be ontologically included in the category of actions, but the distinctive criterion introduced by the pragmatics doctrine may be easily represented.an alternative solution is to consider the categories of legal transactions, such as contracts, wills, as abstract objects, or distinguish the legal transaction as a source of legal effect from the legal transaction as an object (specially those legal transactions which require the ad substantiam form) The relationships between classes of dispositions and ontological classes The illocutionary acts [16, 17] are functions which turn a proposition into an institutional act: they express powers which create meaning, such as definitions; deontic powers which create rights and obligations; constitutive powers which create status, such as citizenship; procedural powers which create conditions under which the previously mentioned powers may be exercised. Where do such attributive powers to the illocutionary functions come from? Where the conditions on which the illocutionary act does not simply describe, but asserts, orders, prescribes and regulates? Here law philosophy, as in the field of AI & LAW, is rich of background on which the methodological choices can be justified: the realistic theories of law15, 25] meet the criteria adopted by the top ontology on a philosophical/epistemological level 6 [. The norm is a constitutive rule of an institutional fact. In fact, it establishes links among concepts (institutional types), the conditions under which a brute fact becomes a token of the institutional category (concept) and defines the juridical consequences. The norm itself is an institutional fact being a token of other types (meta norms ). 6 It implies an isomorphism between the structure of legal concepts and the structure of their representation, based on a realistic vision, which considers ontological types belonging to social, legal, psychological and scientific sciences as objective entities. It states that such Types belong to the category of institutional types. A token of an institutional type is an institutional fact.(x implies Y in C means the brute (or irrational) fact X, e.g., the deliver of a property is an institutional fact Y, e.g., a sale in the code/system C)

4 On the ontological level the classes of dispositions are relationships, one could even consider an ontology of universals typical of the social/legal world since they offer criteria of organizing the world s elements and the real world s entities through the ideal vision of the law. 4. The applied model: examples from the Privacy Law In the following example we shall compare the results we would obtain in retrieving information from an information system based on the NIR methodology, with information provided by a system driven by an ontological model. Experimentation with the DTDs is still in progress at IDG on a restricted domain about personal data protection, whose core is the Italian Statute n. 675 enacted in The text has been structured in a XML document whose atomic unit is the paragraph. To simulate its use in searching information, let s consider a practical case: the following is one of the FAQ published in a private Italian site, which provides online advisory services on the right of privacy. Question: I d like some explanation on the Chapter VIII regarding sanctions. In fact, in each article there is no reference to a specific subject (controller, processor or person in charge) but the general whoever diction is used.. Could you please point out the possible subjects of sanction in each article? Art. 34 Failure to notify or incorrect notification Whoever fails to comply with the notifications as per articles 7 and 28, or provides incomplete or incorrect information through such notifications, shall be punished by imprisonment for between three months and two years. If the act concerns the notification as per paragraph 1 of article 16, the punishment shall be imprisonment for up to one year. Answer: the controller or the responsible for processing. Article 35 Unlawful processing of personal data Any person who, with a view to gain for himself or another or with intent to cause loss to another, processes personal data in breach of articles 11, 20 and 27, shall be punished by imprisonment for up to two years or, if the fact consists in data communication or dissemination, by imprisonment for between three months and two years, unless the offence is more serious. Paragraph 2. Any person who, with a view to gain for himself or another or with intent to cause loss to another, communicates or disseminates personal data in breach of articles 21, 22, 23 or 24, or of the prohibition as per article 28(3), shall be punished by imprisonment for between three months and two years, unless the offence is more serious. Answer: the controller, the responsible for processing as well as the persons in charge. Art. 36 Failure to adopt measures required for data security. Whoever fails to adopt the measures he is required to adopt in order to ensure security of personal data, in breach of the regulations as per paragraphs 2 and 3 of article 15, shall be punished by imprisonment for up to one year. If the fact causes damage to another, the punishment shall be imprisonment for between two months and two years. Answer: the controller, the responsible for processing. Art. 37- Failure to comply with measures taken by the Authority. Whoever fails to comply with measures taken by the Authority, pursuant to article 22 (2), or to paragraphs 4 and 5 of article 29, shall be punished by imprisonment for between three months and two years. Answer: the controller, the responsible for processing. According to the semantic notation provided by the DTD, we would have obtained:, as inline terms only the references i;- as functional markup of the dispositions: art.34 sanction (whoever, one-year imprisonment) art.35 paragraph 1, sanction (whoever, two-year imprisonment) art.35 paragraph 2, sanction (whoever, three-month to two-year imprisonment).etc., are obtained. A more expressive result could be obtained through the elementary meta-relations defined between normative categories, in this case assuming that the sanction depends on the violation of an obligation and that the subject of the obligation is (save exceptions) the subject whom the sanction is comminated. For example, the first paragraph of article 7 (A controller intending to process personal data falling within the scope of application of this Act shall have to notify the Authority thereof.) defines: Art.7 obligation (controller, notification), from which we deduce: sanction (controller). It is not possible, though, to draw the information that the sanction provided by the art. 35, is extended to the person in charge. A deeper description of the dispositions requires the formalisation of the two key concepts of controller 7 and responsible, which keep a rather stable meaning in law. They seem to have a different ontological nature: the state of being controller is a role: it is not rigid, carries no identity criteria and depends on a institutional fact (in this case the power to take decision and to control, deriving from the exercise of a profession or public function, etc); responsible is a legal relation..both are defined in the art. 1 of the Statute, containing the definitions of all entities, which are regulated: controller shall mean any natural or legal person, public administration, body, association or other agency that is competent to determine purposes and methods of the processing of personal data, as also related to security; The role of controller applies therefore to the natural persons and social organizations as the legal agent role; the actions (processing personal data) are legal acts. Moreover, the role of controller implies the responsibility of processing who may affect the controller or a different 7 Actually the italian term is titolare, that has a wider meaning of person entitled to exercise legal actions ; controller is in the EC Directive from which the Italian Statute follows(directive 95/46/EC of the European Parliament and of the Council on the protection of individuals with regard to the processing of personal data and on the free movement of such data: controller shall mean the natural or legal person, public authority, agency or any other body which alone or jointly with others determines the purposes and means of the processing of personal data ).

5 person. Again from the definition given by the article 1: responsible for processing shall mean any natural or legal person, public administration, body, association or any other agency that is processes data on behalf of the controller; The ontological collocation of responsible for processing presents more problems: on a theoretical level, one should look into the formal theories of rights, but responsibility is neither clearly defined in the Fundamental Legal Conceptions [8] nor in the Normative Positions [12,13]. The recently developed formal and computational languages offer a greater expressive dimension by adding the action operators to those of obligation/prohibition. In the A-Hofheld language developed by [1], the closest concept to responsibility is that of liability, correlated to the concept of power and opposed to immunity: legal relation r has Liability of being created by pl means pl has Power to create legal relation r. In Kanger[12] a is responsible for F means a is blameworthy for F. Kanger s further formalization in terms of Duty x ought to bring it about that F is criticized in [18] as there are some senses of obligation -as when we say e.g. there is an obligation on x to see to it that F, or x is responsible for, or held accountable for, ensuring that F is the case, which are not adequately represented by this construction. In [9] the concept of responsibility is linked to the controller s obligation of controlling and is expressed as x brings it about that b brings it about that p. On the other hand, both the controller and the responsible for processing may delegate the practical execution of the actions to the person in charge: art. 8, paragraph 5: The persons in charge of the processing shall have to process the personal data to which they have access by complying with the instructions given by the controller or processor. In order to be able to describe this last situation, we need to explicit a complex relationship characterized by at least: a power to create other powers and to generate obligations, that is, transferring responsibility to someone other than the controller a power to see that the responsible fulfills the obligation a practical power (can) and consequent obligation of the persons in charge to execute the actions. An adequate formalization of the three relations would allow the recognizing of correspondent typologies of the bearers of directed obligations [11] and of the passive subjects of sanctions in case of violation. Research on the field of normative positions has recently been stimulated by the connection with the MAS (multi-agents systems) area [2, 5, 6], further refinement of the present project will therefore provide a detailed description of normative positions, consistently with the criteria of formal ontology. Conclusion Through the analysis of a public project now in progress in Italy, we have examined the need of methodologies to improve access to legal information by non jurist users. The non-expert user has no precise idea of what he is looking for, and uses general terms of common language rather than specific legal concepts; thus the need to explicit the semantic aspects carefully so that the search is driven by a meta-description which keeps univocal references to the text. It has been shown how the solutions chosen for the project have indeed taken an intermediate course, able to reconcile efficient instruments with sensible criteria for practical use. The example shows that the a further refinement of the proposed methodology is possible, defining an external conceptualization of the domain, linked to general formal ontologies; such approach, may be of other use such as normative comparison, conceptual links to multilingual texts, re-utilization in domains of similar application, system of legal advising[26]. References [1]L.E. Allen & C.S. Saxon, A-Hofheld: A language for Robust Structural Representation of Knowledge in The Legal Domain to Build Interpretation-assistance Expert Systems, Deontic Logic and Computer Science, (Meyer and Wiering eds), Wiley Ltd., 1993; [2]Conte, R.. and C. Castelfranchi, 1995, Cognitive and Social Action, UCL Press. [3]Bench Capon T.J., & Visser P.R.S., 1997, Ontologies in legal information systems; the need for explicit specifications of domain conceptualisation, Proceedings of ICAIL 97, ACM Press. [4]Biagioli C., 1997, Towards a legal rules functional microontology, Proceedings of the Workshop LEGONT 97, Melbourne. [5] Boella,G. and L. Lesmo, Deliberate Normative Agents, in (Conte and Dellarocas eds.), Social Order in multi-agent systems, Kluwer (to appear).; [6]Boella G., Favali L. e L. Lesmo, 2001, An Action-based Ontology of Legal relation, ICAIL 2001, ACM Press. [7]Guarino N., Some Ontological Principles for Designing Upper Level Lexical Resources, Proceedings of the First Int. Conf. on Language resources and Evaluation, Granada. SP. [8]Hofheld, W., N., Fundamental Legal Conceptions as Applied in Judicial Reasoning, 23 Yale Law Journal, 16-59, [9]Jones A. & M. Sergot, On the Characterization of Law and Computer Systems: The Normative Systems Perspective, in Deontic Logic and Computer Science, (Meyer and Wiering eds), Wiley Ltd., 1993; [10]A Jones & A, Sergot, 1995, Formal Specification of Security Requirements using the Theory of Normative Positions. [11]Herrestad,H., Formal Theories of Rights, Oslo, JuristForbundets forlag, 1995; [12]S.Kanger, Law and Logic, Theoria, 38, [13]L.Lindhal, Position and Change- A study in Law and Logic, D. Reidel, Dordrecht, [14]L.T. McCarty, A language for legal discourse. I Basic feature, in Proceedings of the second international Conference on artificial intelligence and Law, 1989, ACM press, New York.

6 [15]Ruiter D., W., 1993, Institutional Legal Facts:Legal Powers and Their Effects, Kluwer. [16]Searle, J.R.: 1969, Speech Acts: An Essay in the Philosophy of Language, London: Cambridge University Press. [17]Searle, J.R. and D. Vanderweken: 1985, Foundation of Illocutory logic, Cambridge: Cambridge University Press. [18]M. Sergot, A Method for Automating the Analyze of Normative positions, in Proceedings of Deon 98; [19]Smith B., 1998 Basic Tools of Formal Ontology, in Guarino N. (ed), Formal Ontology in Information systems,ios Press. [20]Sowa J.F., Top-Level Ontological Categories, in Guarino N. e Poli R. (eds), Formal Ontology in Conceptual Analysis and Knowledge Representation, Dordrect: Kluwer.. [21]Valente A. 1995, Legal Knowledge Engineering, A modelling Approach, Amsterdam: IOS Press, [22]van den Haan, N., 1999, Word and normative knowledge in automated legal reasoning, in Artificial Intelligence and Law, Kluwer [23]Tiscornia D. and F. Turchi, 1997, Formalization of legislative Documents Based on a functional Model, in Proceedings of ICAIL 1997, ACM Press. [24]Visser, P:R:S:, and Bench-Capon, T:J:M:, (1996), The formal Specification of a Legal Ontology, in van Kralingen te al(eds), Proceedings of Juriz 96, Tilburg University Press. [25]Weimberger, O.:1991, Law, Institution, and Legal Politics. Dordrect: Kluwer Academic Publishers. [26]Winkels, R., 1998 Clime: Intelligent Legal Information Serving, in Droit et Intelligence Artificielle, Nice, 1998.