Emerald: Legal Knowledge Engineering using OWL and Rules
András Förhécz, Gábor Kőrösi, András Millinghoffer and György Strausz

Emerald is a knowledge engineering environment developed by the people at the Budapest University of Technology and Economics together with Multilogic. The system is an extensive redesign of the Allex Gold system, which performed backward-chaining on Prolog like rules. Emerald is based on the OWL 2 language, and adds expressiveness in the form of aggregate functions (translatable to DL queries), and reasoning with incomplete knowledge (i.e. a local closed-world interpretation of property values). The system also uses a simple rule language to specify the flow of user interaction: rules indicate what property values must be filled before a class restriction is satisfied.

I think it’s unfortunate that this work was only accepted as a short paper, since this work may turn out to become very influential in enterprise-class legal assessment systems (e.g. it is a big step from the propositional calculus of ORACLE’s Haley Office Rules to OWL 2 DL). On the other hand, this work should really be presented at a venue such as OWLED or the KR workshop series as the system is essentially domain independent. From a more legal perspective, the system is described at a too technical level, leaving many issues unmentioned that are particular to (normative) legal knowledge, such as exceptions. Also, I feel the Hungarian language should drop some of its accents.

MetaLex Naming Conventions and the Semantic Web
Alexander Boer

As an XML standard for legal sources (laws, court proceedings etc.) MetaLex has been around for some time now (developed by the University of Amsterdam in 2001), but the relatively new CEN MetaLex brings a significant overhaul of the original design. The new mechanism for specifying naming conventions is but one aspect of this.  MetaLex names are used in self-identification of documents, citation of other documents, and inclusion of document components according to the FRBR levels of item, manifestation, expression and work. A naming mechanism is important for interchange between systems because local implementations may depend on different ways to combine property-value pairs in constructing the IRI’s used to uniquely identify bibliographic entities. In other words, insight in the way in which IRI’s are constructed helps in the discovery of owl:sameAs relations between entities. Arguably, this is a topic very much of interest to the linked open data community (e.g. considering the submissions to the Web of Data track at ESWC 2010, and European projects such as OKKAM).

This paper introduces the naming mechanism (which is quite intricate), and describes how the uniqueness of IRIs can be guaranteed by using a GRDDL transform for translating the property-value pairs encoded in the IRI to OWL class axioms (using nominals, and proper relations between the different FRBR levels). A DL classifier can then infer owl:sameAs relations between entities (individuals) described using the appropriate property value pairs.

Again, given the issues at hand in linked open data, this approach has potential well beyond the AI and Law community.

Rule-based versus Principle-based Regulatory Compliance
Brigitte Burgemeestre, Joris Hulstijn and Yao-Hua Tan

The paper discusses the problems surrounding regulatory compliance, and the influence of different legal systems in specifying regulations on the type of legal reasoning required for systems. The authors focus on rule-based versus principle-based regulations, and identify seven dimensions along which the two approaches differ:

• temporal (ex ante vs. ex post),
• conceptual (specific vs. general),
• functional (little vs. large discretionary power),
• representation (procedural vs. declarative),
• background knowledge needed (little vs. a lot),
• exception handling (strict vs. defeasible),
• conflict resolution (no conflicts possible vs. tradeoff between weights)

They study these dimensions in the context of European (AEO) and US customs regulations (C-TPAT). It turns out that although AEO is mostly principle-based, and C-TPAT is mostly rule-based, both can indeed be positioned at a continuum on which strictly rule- or principle-based regimes form the extremes (as argued earlier by Cunningham and Sadiq et al.). The paper represents a case using Reason Based Logic (Verheij et al.) which illustrates the difficulty in deciding what to represent as a principle, and what to represent as a rule.

I think it is a good thing that the distinction between the two regimes is again brought to the attention of the AI and Law community (especially by authors not especially active in the field): it emphasises the bigger picture of one of the real challenges in AI and Law, namely the development of technology that can be broadly applied, rather than only to interesting, but highly specific problems.

… that’s it for now, more to come (I hope)

Related articles by Zemanta

• Ontology sharing and copyright considerations (sciencecommons.org)
• Semantics in Financial Services -David Newman (slideshare.net)

No related posts.