Structural and Semantic Markup of Complaints: Case

0 downloads 0 Views 487KB Size Report
Abstract — In order to understand complaints more clearly, structural and semantic markup can be rather helpful. Akoma Ntoso, with its rich and flexible structure ...
Structural and Semantic Markup of Complaints: Case Study of Serbian Judiciary Marko Marković*,**, Stevan Gostojić*, Zora Konjović* *

Faculty of Technical Sciences, University of Novi Sad, Novi Sad, Serbia ** Appellate Court in Novi Sad, Novi Sad, Serbia [email protected], [email protected], [email protected]

Abstract — In order to understand complaints more clearly, structural and semantic markup can be rather helpful. Akoma Ntoso, with its rich and flexible structure, is a suitable format to markup complaints. To avoid different interpretations and to provide additional clarification of its content, complaints can be semantically enhanced using legal ontologies and legal knowledge interchange format argument graphs. That way, complaints can become more understandable for machines as well as for humans.

I. INTRODUCTION Complaints are an important type of judicial documents because they initiate court proceedings. Since any person can file a complaint with the court, and often those persons are not educated lawyers, there are no rigid constraints prescribing its form and content. Clearer, more precise and non-contradictory formulation of complaints can influence the result of the proceedings and help obtain satisfactory court decision. In [1], legal resources are presented as complex multilayer information architecture. From the lowest to the highest layer, text, structure, metadata, ontology and legal knowledge layers can be distinguish. Therefore, document structure needs to be established, appropriate set of metadata has to be chosen (e.g. document identifiers, keywords, normative references, etc.) and ontologies have to exist to properly represent semantic of complaints (i.e. to represent its meaning). In the next section will be reviewed several available formalisms that can be used for structural and semantic markup of complaints. As a case study, those formalisms will be used to represent corpus of anonymous complaints filed with courts under jurisdiction of the Appellate Court in Novi Sad. In the conclusion, benefits of the proposed solution will be considered and directions for further research will be given. II. RELATED WORK This section reviews mechanisms which can be used for structural and semantic markup of complaints. Structural document models that will be described are: Akoma Ntoso [2], CEN MetaLex [3], OASIS Electronic Court Filing [4] and National Information Exchange Model (NIEM) [5]. Akoma Ntoso semantic markup [2], LKIF-Core Ontology [6], Legal Case Ontology [7], Judicial Ontology Library (JudO) [8] and LKIF argument graphs [6] could be considered as formalisms for the semantic markup of complaints. Also, Ontology of Professional Judicial Knowledge (OPJK) [9] and CContology [10] will be described.

As shown in [11], Akoma Ntoso XML schema has been proven as suitable to describe the structure of judgments. Although complaints are not covered by the specification, Akoma Ntoso offers some methods for extending its original structure in order to support other document types. By using the proprietary element in the meta section of the document, it is possible to add new proprietary metadata. For more specific documents which use Akoma Ntoso vocabulary, generic elements doc and documentCollection can be used. It is also possible to redefine Akoma Ntoso schema, by creating custom schema, to meet specific requirements. CEN MetaLex provides an XML schema for the representation of generic legal documents [12]. It can be considered as the lowest common denominator of other standards in the legal field. CEN MetaLex defines mechanisms for schema extension, adding metadata, naming, cross-referencing and constructing compound documents. In [13], authors describe a rule-based reasoning mechanism which uses XML representation of contracts, lawsuits, affidavits and requests for summaries as an input to the system. Complaints described in this paper contain data describing actors (such as their personal data and their roles) and disputed monetary contract. They do not support claims which are not money related or data about facts and proofs. Another type of document, affidavit, is also defined, so plaintiff can further justify his claim by linking with the contract clauses. That way, it becomes clear to which clauses complaint relates to. The OASIS Electronic Court Filing technical committee [4] developed an XML schema for electronic filing with the courts of law. For this purpose, complaint documents are structured as messages which define metadata only. It provides elements used for specifying data about: case participants, case title, case type, confidentiality, etc. The content can be embedded or attached (it accepts complaint documents as binary documents). U.S. Department of Justice and U.S. Department of Homeland Security have developed NIEM data exchange model [5] capable of supporting complaints. In order to allow greater flexibility, ontological structure defined by Akoma Ntoso is not formalized [14]. There are 14 top level classes (10 classes for identifying entities in the document and 4 document related classes). A resolution layer with two available implementations are used to link URIs used in Akoma Ntoso documents with the entities. One implementation of this layer relies on redirection service, while the other is based on mappings between URIs (for example, using owl:SameAs property).

That way, simple Akoma Ntoso URIs can also refer to data about concrete entity. Akoma Ntoso provides several elements which can be used to markup semantic aspects of text fragments (normative references, people, organizations, locations, etc.). For instance, TLCPerson and TLCRole element in the metadata section and in-line element party in the document body are used for semantic annotation of parties. Classes from external ontologies are referenced using TLCReference element, and corresponding text fragments are marked up using entity element. LKIF-Core ontology [15] describes concepts from both: legal and commonsense domains. It contains 15 modules related to abstract concepts (top ontology, place, mereology and time), basic concepts (process, role, action and expression), legal concepts (legal action, legal role and norm), core and extended ontology. LKIF-Core ontology enables knowledge interchange between existing legal knowledge systems. It already contains classes appropriate for the judicial domain and represents a solid base for other legal ontologies. A legal case ontology, proposed in [7], focuses on main elements of legal cases, supports legal reasoning and can be used for text annotation. Key classes in the ontology are: Case, Hearing, Decision, Jurisdiction, Participant, Argument_Scheme and Element. Classes defined in this ontology cover many general legal concepts but more specific concepts in certain fields of law need to be defined in order to be fully applied to complaints. As described in [8, 16], to formalize legal concepts and argumentation patterns found in the process of adjudication, four different models should be used: a document metadata structure (creates a bridge between a text and the semantic annotations of legal concepts), a legal core ontology (models abstract legal concepts and institutions contained in a rule of law), a legal domain ontology (main legal concepts in the specific domain), and an argumentation system (models the structure of argumentation using arguments, counterarguments, premises, conclusions, rebuttals, proof standards, argument schemes, etc.). Since requirement for document metadata structure is already supported by standards such as Akoma Ntoso, legal core ontology and legal domain ontology could be provided, introducing Judicial Ontology Library (JudO) [8]. JudO is designed in two modules: Core ontology (as extension of the LKIF-Core legal ontology) and a Domain ontology (representing the concepts and the rules expressed by the Italian Consumer Code, Italian Civil Code and some Italian judgments). JudO core ontology module is based on three main classes: Qualifying_Legal_Expression (legal expressions which give legal status to a person or an object), Qualification (includes legal acts which produce qualifying legal expressions), and Qualified (includes objects of a qualifications). Those classes represents nucleus of JudO Core Ontology. LKIF argument graphs explained in [6] are useful for representing proofs which helps people to understand complex chains of reasoning and arguments. The content model of a LKIF argument graph consists of statement elements followed by argument elements. An argument element links premise elements to conclusion element while both, premise and conclusion elements, are referenced to corresponding statement element using statement attribute. As explained in [17], XML format

from the ESTRELLA project is replaced by new Carneades Argument Format (CAF). In CAF version 1.3, statement element contains an optional metadata element whose attributes correspond to Dublin Core Metadata Element Set. Argument graphs are suitable to provide visual overview of complaints and could easily become a basis for the rest of court proceedings. The ontology of Professional Judicial Knowledge (OPJK), presented in [9] and [18], is developed on the basis of surveys results, with the focus on professional judicial knowledge. During the development of OPJK some upper-level ontologies were considered and after complex process of revisions and comparisons, some conceptual ideas were adopted. Due to various contexts in which particular concepts of upper-level ontologies were described, some difficulties in formalization of such concepts were found. Thus, concepts in OPJK are only inspired by those upperlevel ontologies, but are not reusing them. The customer complaint ontology (CCOntology), introduced in [10], consists of seven modules (Complaint, Complainant, Recipient, Address, Complaint Problems, Complaint Resolutions, and Contract) and enables semantic description of complaints between legal persons. In addition, a methodology for multilingual lexicalization is proposed which, using English as the native language, enables ontology translation into other languages. III. CASE STUDY OF SERBIAN JUDICIARY This section gives an overview of a case study in which formalisms reviewed in the previous section were applied to complaints. It describes how complaints written in the natural language can be structured in Akoma Ntoso and annotated with concepts derived from LKIF-Core, Legal Case Ontology and JudO ontologies and visually represented by argument graph formulated in CAF. Due to space constraints, this paper shows document fragments only. Complete documents are available online at [19]. As described in [20], Akoma Ntoso can be used to markup structured textual documents used in different judicial information systems. Furthermore, it is the base of the work conducted by OASIS LegalDocML Technical Committee [21] as the future legal document standard. Out of particular importance is that Akoma Ntoso separates content, presentation and metadata layers and has found practical applications in several jurisdictions. Because of that, and the fact that it already supports many structural elements found in complaints, Akoma Ntoso appears to be an appropriate choice for complaints. According to the Article 192 of the Civil Procedure Code [22], “complaint must contain a claim concerning the subject matter of litigation, the facts on which the plaintiff bases the claim, proof of such facts, the value of subject matter as well as other information necessary for any filing”. In addition, the Article 98 prescribes that every filing, including complaints, shall be comprehensible and shall specify name of the court, first and last name of the plaintiff (or registered name if the plaintiff is a legal person), permanent or temporary residence, legal representatives and attorneys, the subject matter of the dispute, the contents of the statement and the signature of the submitter.

Therefore, considering flexible nature of Akoma Ntoso schema, an extended schema is developed. For XML schema development Altova XMLSpy 2011 [23] is used. New Akoma Ntoso document type and its structure are shown in Fig. 1.

Figure 1. Structure of complaint document

Similar to the structure of judgments, mandatory elements are meta (for identification, classification, referencing to legal documents, individuals of ontology classes, etc.), header (usually contains court type, parties, lawyer, subject matter, value of subject matter, etc.) and complaintBody (arguments supported with facts and evidence). Optional elements are coverPage (for various identification numbers, editorial metadata and typographic elements), conclusions (contains signatures, date and location) and attachments (to support additional documents related to the main text).

Osnovni sud u Novom Sadu tužilac: Marko Marković iz Novog Sada tuženi: Stevan Gostojić iz Novog Sada TUŽBA radi utvrđivanja prava svojine Vrednost spora: 100.000,00 dinara u dva primerka s dva priloga Listing 1. Header element

Corresponding concepts for subject matter and subject matter value, in references section of meta block, are shown in Listing 2. Listing 2. Concept declaration of subject matter and its value

Providing facts and evidence in consecutive order is shown in Listing 3 where, after specifying a fact, corresponding evidence is given.

Figure 2. Structure of complaintBody element

The complaintBody element, through choice based element complaintBlock (shown in Fig. 2), offers following elements: introduction (explanations that are not part of facts, evidence or claims), fact (assertions that confirm claims), evidence (usually specifies material and/or testimonial proofs), claim (requests for determination of some legal status or fulfillment of some obligation, declared in [22] in terms of article 194 and 196). All these elements are of Akoma Ntoso maincontent type, enabling them to support another Akoma Ntoso structures. This is especially useful in the case of claim element since claims are often declared in the form of proposed decision as Akoma Ntoso decision element. Complaint template for determination of ownership of real estate property, published in [24], will be used as an example of complaint. Element header, containing data about court, plaintiff and defendant along with marked up subject matter and its value, is shown in Listing 1.

Tuženi je zemljišnoknjižni vlasnik nepokretnosti - parcele br. 1234/5 LN 1234 KO Novi Sad 2 površine 510 m2 s ekonomskom zgradom površine 70 m2. Dokaz: izvod iz zemljišne knjige Listing 3. Plaintiff's argument

The fact and the evidence shown in the previous XML snippet both reference an object representing extract from cadastre. The element corresponding to this object is declared in references section, as shown in Listing 4. Listing 4. References block

Plaintiff's claim, in the form of judgment, is shown in Listing 5.

Utvrđuje se da je tužilac po osnovu kupoprodajnog ugovora, overenog kod Osnovnog suda u Novom Sadu dana 12.12.2010. godine, pod brojem Ov 12345/10, stekao pravo svojine na kat. parceli br. 1234/5 LN 1234 KO Novi Sad 2 površine 510 m2 na kojoj se nalazi ekonomska zgrada površine 70 m2, što je tuženi dužan da prizna i trpi da tužilac uknjiži u zemljišnim knjigama pravo svojine na opisanim nepokretnostima, a u protivnom će mu ova presuda služiti kao osnov za uknjižbu. Listing 5. Plaintiff's claim in the form of judgment

In order to properly annotate the complaint, an ontology is derived from LKIF-Core, Legal Case Ontology and JudO. Protégé Desktop [25], an open source ontology editing environment, was used for ontology development. Declarations of namespaces needed for previous ontologies, with corresponding prefixes, are shown in Listing 6. xmlns:lkif= ”http://www.estrellaproject.org/lkif-core/lkif-core.owl#” xmlns:legal-case-ontology= "http://www.best-project.nl/owl/legal-case-ontology.owl#" xmlns:judo= "http://judo.googlecode.com/svn/tunk/Judging_Contracts_Core.owl" Listing 6. Namespace declarations

“http://informatika.ftn.uns.ac.rs/legal/judiciary.owl” namespace was chosen for ontology described in this paper. The Complaint class, which represents some complaint document, is defined as subclass of Legal_Document class from ‘norm’ module of LKIF-Core ontology (as shown in Fig. 3).

Figure 5. Plaintiff and Defendant classes from Legal Case Ontology

Since this paper considers complaint related to some real estate purchase contract, class Contract (defined in LKIF-Core ontology and shown in Fig. 6) is found to be suitable to represent this contract.

Figure 6. Contract class from LKIF-Core

All classes and individuals mentioned in the complaint are described using RDF triplets which reference Akoma Ntoso elements. Since all significant elements in Akoma Ntoso documents have their own identifiers, those identifiers are used as subjects or objects of RDF statements. That way, facts about the contract become machine-readable. For example, plaintiff and defendant are persons declared in the references section of Akoma Ntoso document as shown in Listing 7. Listing 7. Declaration of persons

Optional RDFa syntax in combination with Akoma Ntoso in-line elements could also be used to describe resources. Namespace prefix which references individuals of Akoma Ntoso top-level classes is declared as in Listing 8. xmlns:an="http://www.akomantoso.org/" Listing 8. Akoma Ntoso namespace declaration

Figure 3. Complaint Class

On the basis of Expression class, documented in ‘expression’ module of LKIF-Core ontology as proposition which is carried by some medium, subclass Claim is introduced (Fig. 4).

Figure 4. Claim Class

Class Medium is a domain and class Expression is a range of object property bears defined in expression module. Thus, it is suitable for expressing the relation between the complaint and its claims since Complaint and Claim classes are subclasses of domain and range respectively. The parties (i.e. classes Plaintiff and Defendant), which belong to Legal Case Ontology, are shown in Fig. 5.

RDF statements describing the complaint could be introduced as shown in the Listing 9. 50000 utvr đivanje prava svojine Listing 9. RDF statement describing a complaint

Akoma Ntoso markup of complaint documents provides unique identifiers to some elements. Those identifiers could be used to reference corresponding parts of the document in statement elements of Carneades

Argument Format (CAF) through source attribute of metadata element. That way, parts of complaint body could be reused as textual elements of argument graph appearing in premises or conclusions. An example is shown in Listing 10. Listing 10. Using URI as a source for CAF statement

Support for proposed mechanism of referencing statement sources could be provided through Clojure language since, as described in [26] since Carneades argumentation system is based on Clojure from version 2. Another method, independent of technologies used by Carneades argumentation system, to achieve the same goal would be to invoke a simple XSL transformation which fills textual content of statements with content of corresponding parts of complaint body. Representation of an argument, using Carneades web application, connecting the premise to the conclusion, is shown in Fig. 7. It also shows how the fact is supported by the corresponding evidence.

Figure 7. Argument with single premise

This argument (built on a single premise) is shown in Listing 11 in XML form. Listing 11. CAF argument with single premise

Similarly, the argument with two premises is shown in Listing 12. Listing 12. CAF argument with two premises

Graphical representation of this argument is also shown in Fig. 8.

Figure 8. Argument with two premises

Fig. 9 shows how conclusions of two arguments become premises in another argument. In the context of complaint which was considered, it also depicts how exposed facts should confirm a claim.

Figure 9. Chaining of arguments

This argument is also represented in CAF syntax as shown in Listing 13. Listing 13. Claim as conclusion of an argument

Listing 14 shows how statement representing claim, which contains essence of plaintiff's intentions, is marked according to the CAF syntax using main attribute. Listing 14. Main CAF statement

IV. CONCLUSIONS This paper proposes how complaints could be annotated before filing them with the court, in order to augment its content written in the natural language with more precise semantics. After analysis, Akoma Ntoso XML schema, LKIF-Core, Legal Case Ontology, JudO and Carneades Argument Format are shown to be a promising solution to this problem. There are several benefits of the proposed representation. Preliminary examination of the complaint would be easier to fulfill since argumentation would be organized in graph structure and claims contained in the complaint would be linked with concepts defined in an ontology. Complaints in this format could also be submitted to the defendant. The defendant could more easily understand what exactly plaintiff’s claims are, submit a response after studying argument graph and plan the defense which points directly to disputed claims. Going even further, the response to the complaint prepared by the defendant and filed with the court, could be in a similar form. That way, it would be obvious to the court what should be the subject of future hearings. In rare cases, if there are no disputable claims and the defendant agrees with the plaintiff's claims, the judgment could be automatically generated (on the grounds of admission). Complaints described in this paper are written in accordance with the Civil Proceedings Code. Same formalisms, with no significant modifications, could be used to represent complaints written according to the Criminal Proceedings Code [27]. Since complaints formulated as suggested in this paper still contain narrative part along with metadata, it should be fully backward compatible with existing e-filing

services. Those services could simply ignore additional data or complaints could be transformed to desired form. It should be noted that no changes in legislation are necessary since all requirements, in terms of considered rules of court proceedings, remain satisfied. Further research should be directed towards designing and developing an editor capable of supporting complaints (argumentation modeling and text markup). Apart from the editor, a viewer should be provided to the court and other interested parties, to enable presentation of the complaints written in the natural language together with the corresponding argument graph. Such viewer could help the defendant to better understand the claim and its arguments and also could help in training the law students. REFERENCES [1]

M. Palmirani, L. Cervone, and F. Vitali, "A legal document ontology: The missing layer", in legal document modelling", in Approaches to Legal Ontologies , Springer, 2011, pp. 167-178 [2] United Nations Department of Economic and Social A ffairs (UNDESA), "AKOMA NTOSO - XML for parliamentary, legislative and judiciary documents", http://www.akomantoso.org/ [3] A. Boer, E. Hupkes, F. Vitali, M. Palmirani, and B. Ratai, "Metalex CEN workshop proposal, Technical report", CEN Workshop on an Open XML Interchange Format for Legal and Legislative Resources (Metalex), 2008 [4] A. Angione, J. Cabral, "Electronic Court Filing Version 4.01", The Organization for the Advancement of Structured Information Standards (OASIS), 2013. [5] "National information exchange model (NIEM)", U.S. Department of Justice and Department of Homeland Security [6] T. F. Gordon, "The legal knowledge interchange format (LKIF)", Estrella deliverable d4.1 , Fraunhofer FOKUS Germany, 2008 [7] A. Wyner, and R. Hoekstra, "A legal case OWL ontology with an instantiation of Popov v. Hayashi", Artificial Intelligence and Law, vol. 20.1, 2012, pp. 83-107 [8] M. Ceci, A. Gangemi, "An OWL ontology library representing judicial interpretations", unpublished [9] P. Casanovas, N. Casellas, J. J. Vallbé, M. Poblet, J. Contreras, M. Blázquez, and V. R. Benjamins, “Semantic Technology for Professional Judicial Knowledge”, Semantic knowledge management: integrating ontology management, knowledge discovery, and human language technologies , Springer Berlin Heidelberg, 2009, pp. 199-214. [10] M. Jarrar, R. Verlinden, and R. Meersman. “Ontology-based customer complaint management”, On the move to meaningful internet systems 2003: OTM 2003 workshops , Springer Berlin Heidelberg, 2003, pp. 594-606 [11] M. Marković, S. Gostojić, Z. Konjović, and M. Laanpere, "Machine-Readable Identification and Representation of Judgments in Serbian Judiciary", Novi Sad Journal of Mathematics, in press.

[12] A. Boer, and R.G.F. Winkels, "CEN MetaLex: Facilitating interchange in e-Government", Government and Electronic Participation: Joint Proceedings of Ongoing Research and Projects of IFIP EGOV and ePart 2011, Schriftenreihe Informatik 37 - Trauner Verlag, 2011, pp. 240-248. [13] L. L. Leff, “Automated reasoning with legal XML documents”, In Proceedings of the 8th international conference on Artificial intelligence and law, ACM, 2001, pp. 215-216 [14] G. Barabucci, A. Di Iorio, and F. Poggi, “Bridging legal documents, external entities and heterogeneous KBs: from metamodel to implementation”, Semantic Web Journal, IOS Press, 2012 [15] "The LKIF Core Ontology - A core ontology of basic legal concepts", The European project for standardized transparent representations in order to extend legal accessibility (ESTRELLA), http://www.estrellaproject.org/lkif-core/ [16] M. Ceci, and M. Palmirani, "Ontology framework for judgment modelling", In AI Approaches to the Complexity of Legal Systems. Models and Ethical Challenges for Legal Systems, Legal Language and Legal Ontologies, Argumentation and Software Agents, Springer Berlin Heidelberg, 2012, pp. 116-130 [17] T. F. Gordon, “Introducing the Carneades web application”. In Proceedings of the Fourteenth International Conference on Artificial Intelligence and Law , ACM, 2013, pp. 243-244 [18] N. Casellas, “Modelling Judicial Professional Knowledge: A Case Study”, Legal Ontology Engineering: Methodologies, Modelling Trends, and the Ontology of Professional Judicial Knowledge , Springer Netherlands, 2011, pp. 171-240 [19] “Structural and Semantic Markup of Complaints”, Faculty of Technical Sciences, Novi Sad, available at https://informatika.ftn.uns.ac.rs/legal/svn/complaints/trunk/ [20] G. Sladić, S. Gostojić, B. Milosavljević, Z. Konjović, "Automatic Anonymization of Judgments in Electronic Format", In 3rd International Conference on Information Society Technology and Management (ICIST), Kopaonik, Serbia, 2013, pp.206-211 [21] OASIS (Organization for the Advancement of Structured Information Standards) LegalDocumentML (LegalDocML) TC, https://www.oasis-open.org/committees/tc_home.php? wg_abbrev=legaldocml [22] "Civil Procedure Code", Official Gazette of the Republic of Serbia, No. 74/2013 [23] “XMLSpy XML Editor”, Altova, http://www.altova.com/xmlspy.html [24] "Complaint template for determination of ownership of real property", Ing-Pro Package of Enactments 10+ , Ing-Pro, Belgrade [25] “Protégé Desktop”, Stanford Center for Biomedical Informatics Research, Stanford University School of Medicine, http://protege.stanford.edu/ [26] M. Ceci, and T. Gordon, “Browsing Case-law: an Application of the Carneades Argumentation System”, RuleML (2), 2012 [27] "Criminal Procedure Code", Official Gazette of the Republic of Serbia, No. 45/2013

Suggest Documents