An OWL ontology library representing judicial interpretations

Paper Title: 
An OWL ontology library representing judicial interpretations
Marcello Ceci, Aldo Gangemi
The paper introduces a formal model of judgments that, starting from the text of decisions, produces an OWL ontology that represents the interpretations performed by a judge while conducting a discourse towards an adjudication. The final goal of this method is to design an ontology framework capable of detecting and modelling jurisprudence directly from the text, and performing some basic reasoning on the resulting ontologies.
Full PDF Version: 
Submission type: 
Ontology Description
Responsible editor: 
Major Revision

Submission in response to

Solicited review by Nick Bassiliades:

The paper is complete and it deals with building a legal ontology. The nice thing
about this paper is that it places the ontology development within the larger
framework of research of the authors and their team, so the reader can understand certain choices.

Some suggestions for improvement:

- Cite and compare relevant work on legal and/or public administration ontologies, such as
the following:

[1] Casanovas, P., Poblet, M., Casellas, N., Contreras, J., Benjamins, V. R., & Blazquez, M.
(2005). Supporting newly-appointed judges: A legal knowledge management
case study. Journal of Knowledge Management, 9(5), 7–27.
[2] I. Savvas, N. Bassiliades, "A Process-Oriented Ontology-Based Knowledge Management System
for Facilitating Operational Procedures in Public Administration", Expert Systems with Applications,
Vol. 36, Iss. 3, Part 1, 2009, pp. 4467-4478.
[3] Sylvie Despres and Sylvie Szulman, 'Construction of a Legal Ontology from a European Community Legislative Text' in T. Gordon (ed.),
Legal Knowledge and Information Systems. Jurix 2004: The Seventeenth Annual Conference. Amsterdam: IOS Press, 2004, pp. 79-88.
[4] Jaspreet Shaheed, Alexander Yip and Jim Cunningham, A Top-Level Language-Biased Legal Ontology, in Jos Lehmann, et al. (Eds),
LOAIT - Legal Ontologies and Artificial Intelligence Techniques, Bologna 2005, IAAIL Workshop Series No 4, Bologna 2005,
Wolf Legal Publishers ISBN ISBN 90-5850-504-89 pp 13-24.

- Concerning the use of defeasible logic for the legal argumentation system, maybe you should consider
deontic defeasible logic systems, as well, such as:

[5] Antoniou, G., Dimaresis, N., & Governatori, G. (2009). A Modal and Deontic Defeasible Reasoning System
for Modelling Policies and Multi-Agent Systems. Expert Systems With Applications, 36, 2, 4125-4134.
[6] SPINdle - this is already cited in the paper
[7] E. Kontopoulos, N. Bassiliades, G. Governatori, G. Antoniou, "A Modal Defeasible Reasoner of Deontic Logic
for the Semantic Web", International Journal on Semantic Web and Information Systems, Thomson-Scientific, 7(1),
pp. 18-43, 2011.
[8] Nute, D. (1998). Norms, Priorities, and Defeasible Logic. In McNamara, P. & H. Prakken (Eds.),
Norms, Logics and Information Systems, pp. 201-218, IOS Press, Amsterdam.
[9] Governatori, G., & Rotolo, A. (2008a). BIO Logical Agents: Norms, Beliefs, Intentions in Defeasible Logic.
Journal of Autonomous Agents and Multi Agent Systems, 17, 1, 36-69.

- Most figures of section 2.1 are not properly referenced from within the text.

- A figure for the judged_as Property Chain would help in reader comprehension.

- Figure 10 should have labels on the arcs.

- The last sentence of section 3 states that:
"A better exploitation of the OWL 2.0 property chains could lead to an ever more direct and complete solution,
mainly by removing the need for the anonymous subclass in order to identify the clause instances considered_by
the relevant law."
Since the authors consider that this is a better solution than the one implemented, why don't they try to implement
this solution in the current ontology version? The authors should clarify and justify their choice.

- The issues described at the end of section 5 have to do with non-monotonic reasoning, combined with DL-reasoning.
Maybe the authors would like to consider approaches, such as Conjunctive Query Programs (CQ-Programs) [10]
or Closed World Reasoning [11].

[10] Minh, D.T.; Eiter, T.; Krennwallner, T. Realizing Default Logic over Description Logic
Knowledge Bases. Proc. of the ECSQARU 2009. Springer, 2009.
[11] Yuan Ren, Jeff Z. Pan, Yuting Zhao, Closed World Reasoning for OWL2 with NBox, Tsinghua Science & Technology,
Volume 15, Issue 6, December 2010, Pages 692-701, ISSN 1007-0214, 10.1016/S1007-0214(10)70117-6.

- The conclusion section is section 7, not 6.

Solicited review by Andrew Koster:

The paper describes a couple of ontologies that, together, allow for precedents in Italian contract law to be represented and reasoned about.

Unfortunately, there are a number of design choices that are not described properly, and I am not sure the methodology is applicable beyond cases similar to the example described in the paper.

Firstly, the authors do not adequately describe why OWL2 was chosen. They highlight the need for some form of defeasible reasoning in law, in order to deal with the many exceptions. This is, in fact, one of the major issues, with the chosen model, as the authors point out in 6.2. The question is therefore why OWL2, rather than a model in which defeasible reasoning is naturally included. For instance, Prakken's dialogue model of adjudication:

Prakken, Henry. Formalising Ordinary Legal Disputes: a Case Study. AI and Law 16: 333-359. 2008.

I, of course, realize that OWL2 has other strengths, and in particular, modeling relations in a computationally efficient (comparatively, anyway) manner is a strong reason to use OWL2 over an argumentation-based approach. However, the choice for OWL2 needs to be motivated in the paper. Preferrably in comparison to other approaches.

A number of requirements for the modeling are mentioned on pages 2 and 3. It would be useful to show how the modeling choices achieve these requirements. Most of these are left implicit, or are already achieved by previous work (such as Akoma Ntoso).

Secondly, some of the choices in the modeling process are unclear. For instance, a Legal Expression is, in figure 3, only a direct subclass of the top level concept, however in the textual description it seems further nested and is a subclass of Expression, which makes more sense (and is also in figure 6). There are a number of such inconsistencies throughout the text and the figures, making it hard to understand some of the design choices in extending the core, and domain ontologies.

Finally, the figures 9, 10 and 17 are chaotic and do not seem to add anything to the understanding of the model. Furthermore, the colours are not explained and if the images are printed in black and white will be largely indistinguishable. Figures 9 and 17 need to be clarified significantly. Figure 10 seems redundant.

Solicited review by anonymous reviewer:

I am not qualified to judge the finer points of the modeling techniques involved, and so will not do so. The overall approach seems to me to be an eminently practical one. The attempt to get away from disconnected abstractions and to realistically consider what is needed to create the domain ontology is commendable -- I am a big fan of the middle-out approach.

Expansion of the work to a significant body of (eg.) American caselaw might run into problems with the structuring of judicial opinions themselves, which (it seems to me, at least) is a little less "templatized" than it is in jurisdictions where judges are part of the career civil service. This is another aspect of the knowledge acquisition problem that the authors point out in 6.1. In the spirit of reviewers who usually comment from the perspective of the paper that they themselves might like to write (as opposed to the one they are reviewing), I'll say that it would be interesting to tie the authors' comments in that section to prior work and practical discoveries regarding acquisition in the legal knowledge-management systems of the early to middle 1990s; this is not a new problem.

I've asked for minor revisions, but they truly are trivial -- I find figures 9 and 17 difficult to read and connect to the textual narrative of the case. One might use color coding in the text to relate passages to the diagram, or (after presenting the entire diagram) present sub-portions alongside relevant areas of the text. It is also possible that each presents rather more information than is needed for understanding.

There are also some minor bumps and bangs in the English that a friendly edit would quickly fix.

A commendably practical approach.