An Event Based Design Pattern for Intertextual Legal Links

This paper presents a model for representing links between (para)legal texts that centers around the different types of events that motivate the existence of the link. For instance, a codification event, where a proposed law is turned into a law proper (though it may not have entered into force). The model is used in the LegiLocal project, a project that aims to improve the accessibility of local regulations in France.

There are a number of things that can be remarked about this work:

The paper is not really convincing in the argumentation for the need of this work: how does it improve on existing models? For instance, the main benefit over the MetaLex ontology seems to be in "reducing instantiation efforts" (p.5) because the reference and the referencing event are represented as the same object. Why this is a benefit is left to the imagination of the reader. From an ontological perspective, the confounding of references (which persist through time) and events (which do not) is a rather uncommon decision, that needs to be much better motivated. Secondly, the paper does not cite work on legal modification events by e.g. Palmirani and others. Also, the distinction between 'links' to textual and non-textual objects are already present in MetaLex (cites and refersTo).

For an ontology pattern, it is crucial to show how this pattern applies in multiple situations, or at least: how it can be extended to new situations. This is not done at all. The authors very briefly mention the UK and Dutch linked open legal data work, but do not give any example how their pattern would apply here.

Smaller remarks:
* The footnotes are not listed (3,4,5), are not numbered correctly, and footnotes 10, 11, 12 appear inside the References section. Footnote 10 and 11 appear twice with different content.
* Examples in Turtle are extremely verbose. It's better to use e.g. manchester syntax or just the formal DL notation.

The paper describes a rich ODP for legal references. The choice of representing the reference as an event allows for a detailed description of the reference itself in terms of documents and agents involved, and should (as the authors claim) result in a reduction of instantiation efforts, which is always good since computability is a common issue in legal knowledge representation. Moreover, the paper pays much attention on the multiplicity of references that are possible in the legal domain.

The main problem with the presented approach is that it seems to ignore the distinction between a legal text (structure) and its semantics (see M. Palmirani, G. Contissa and R. Rubino, Fill the Gap in the legal Knowledge Modelling, in Proceedings of RuleML 2009, LNCS 5858, Berlin, Springer, 2009, pp. 305-314). Even the sole related work cited in the paper (the Metalex initiative) identifies 'non-textual objects' in order to leave the door open for any reference operating beyond mere textual modifications.

The lack of distinction between the textual and semantic layers negatively affects the accuracy of representation, which in the legal setting is of paramount importance. The main two aspects of legal semantics that would be missed in this approach are legal validity and temporal parameters.

Knowing the date of a legal document modifying another one, in fact, tells us nothing about the date when the changes will take effect (there is a fixed period of vacatio legis, which the law itself may or may not specify), and thus the answer to a query such as "what is the version of the law in force on december 1st, 2013" would be inaccurate. In order to properly model the versioning of valid laws over time, it is necessary to give account for the different dimensions of legal force (i.e. validity, efficacy, enforceability). See M. Palmirani and F. Benigni, Norma-system: A legal information system for managing time, in Proceedings of the V Legislative XML Workshop, 2007, pp. 205-224; M. Palmirani and R. Brighi, Model Regularity of Legal Language in Active Modifications, LNCS, pp. 54-73, 2010; M. Palmirani, T. Ognibene and L. Cervone, Legal rules, text, and ontologies over time, in Proceedings of the RuleML@ ECAI 6th International Rule Challenge, Montpellier, 2012.

The paper claims to have specified the Metalex model, but the list of possible references included in the new model is far from accurate, e.g. what about the suspension of norms? In laws containing suspension no textual modification occurs, yet a reference it is, and one that greatly affects the panorama of applicable laws (see M. Palmirani, M. Ceci, D. Radicioni and A. Mazzei, FrameNet model of the suspension of norms, in The 13th International Conference on Artificial Intelligence and Law, Proceedings of the Conference, Pittsburgh, 2011).

To summarize: the model proposed in the paper is an interesting approach as a model for mere documental references (i.e. to answer simple questions such as the three questions indicated in section 5.), but the introduction of the paper suggests otherwise (it cites "topics" and "concepts", which go beyond mere textual information and therefore require full accounting for legal interpretation - see G. Boella, G. Governatori, A. Rotolo and L. van der Torre, A Logical Understanding of Legal Interpretation, KR, 2010). To capture such elements a more complex model is required, both for temporal paramenters and for legal validity. My suggestion is to reduce the scope of the paper accordingly, modifying the introduction, and highlighting that this design pattern can work for mere documental references, but in order to retrieve applicable law without incurring in major errors a further layer of representation is required.

In general, the participation of a legal expert would help dealing with legal theory (see e. g. G. Sartor, Legal Concepts as Inferential Nodes and Ontological Categories, Artificial Intelligence and Law, pp. 217-251, 2009) thus ensuring the correct tackling of legal-specific issues. The need for legal expertise beyond mere IT and logic proficiency is the main reason why legal knowledge representation is such a complex task.

The paper lacks a proper conclusion, and fails at citing relevant work; following are some references that probably need proper consideration:

- S. Despres and S. Szulman, Construction of a Legal Ontology from a European Community Legislative text, in Legal Knowledge and Information Systems. Jurix 2004: The Seventeenth Annual Conference, Amsterdam, 2004.
- A. Gangemi, M. T. Sagri and D. Tiscornia, A Constructive Framework for Legal Ontologies, in R. Benjamins, J. Breuker, P. Casanovas and A. Gangemi, eds., Legal Ontologies and the Semantic Web, Springer, 2005.
- R. Hoekstra, J. Breuker, M. Di Bello and A. Boer, LKIF Core: Principled Ontology Development for the Legal Domain, Law, Ontology and the Semantic Web, 2009, pp. 21-52.
- L. Mommers, Ontologies in the Legal Domain, in Theory and Applications of Ontology: Philosophical Perspectives, Springer, 2010, pp. 265-276.
- I. Savvas and N. Bassiliades, A Process-Oriented Ontology-Based Knowledge Management System for Facilitating Operational Procedures in Public Administration, in Expert Systems with Applications, 2009, pp. 4467-4478.
- J. Shaheed, A. Yip and J. Cunningham, A Top-Level Language-Biased Legal Ontology, in LOAIT, Bologna, 2005.

some specific corrections in-text:

- page 1, section 1, third paragraph, 3rd line: documents->document;
- page 1, section 1, third paragraph, 7th-9th line: watch out when saying "in other words, all case law": case law identifies only judicial decision, but earlier on the sentence it is written "all the texts";
- page 2, 12th line: then->than;
- page 6, Listing 2: ref:art-46-quarter-1991->ref:art-46-quater-1991 (2 occurrences)

The paper (An Event Based Design Pattern for Intertextual Legal Links) is a pretty direct piece of work and seems like a contribution to using CODPs to formalize legal documents concepts and their relations for the purposes of design but also retrieval, although the legal field is not in my areas of expertize.
The presented models framed around Events seems reasonable and the one use case example followed by its use for search seems clear enough.

I give the document a minor revision category although there seem to me a number of things that should be addressed.

1. There is inadequate review of the extant literature and discussion of the relation of this work to previous work. One example is the work of HOEKSTRA and associates. There seem to me to be several pieces of that work worth mentioning. An example is "Wyner, Adam, and Rinke Hoekstra. "A Legal Case OWL Ontology with an Instantiation of Popov v. Hayashi." While this is mainly about legal processes it does have a model of legal documents with a few sub-classes noted.

Other, later work by Boer could also be cited and discussed e.g. Lauritsen, Marc, and Thomas F. Gordon. "Toward a general theory of document modeling." Proceedings of the 12th International Conference on Artificial Intelligence and Law. ACM, 2009.

The authors might want to consider a section called Related Work to cover the discussion of the work in such papers.

2. There seems to be no real conclusion section to discuss & summarize issues or to describe next steps. This seems essential. Some commentary on how well the ODP worked for design might be covered along with more discussion on making the products usable for search.

3. Of minor note are some small errors such as goe:SpatialThing 5 being used in Table 1 - it should be geo.
As far as i can tell footnotes 3-6 are not shown at the bottom of the relevant page.
Some of the English expression can be improved. For example in Section 5 (Use Case) I would change
"Does this legal text has been modified? "
to read "Has this legal text been modified?" and change 2 to read
"How has the legal text of article 46 been codified?"

The abbreviation CODeP is used when I have seen CODP for a conceptual ODP.
Why is section 4 called "Position"? This involves the choice of ODPS for formalization. Perhaps another header for this section could be crafted.

The paper addresses the issue of providing an ontology for intratextual links in legal documents. It builds upon the event ontology. It refers to the Event ontology and considers the links as concerning a source and a matter and a result.
I found a bit confusing the presentation of the elements of such an event, being unclear the distinction between the ideas used of of target, matter and a result.
I also wonder whether the temporal aspects are taken case adequately. An aspect that is not clear to me is also how temporal aspects are dealt with. So when a an article in a piece of legislation entering in force at a certain time changes another article from a certain time, maybe retroactively, how are this temporal connections expressed.
I found it strage also that the Akomantoso document and reference model has not been considered since they also address the links between legal documents.