Skip to main content

ProLeMAS: PROcessing LEgal language in normative Multi-Agent Systems

Periodic Reporting for period 1 - ProLeMAS (ProLeMAS: PROcessing LEgal language in normative Multi-Agent Systems)

Reporting period: 2015-06-01 to 2017-05-31

ProLeMAS aimed at devising a new logic for normative reasoning in multi-agent systems and implementing a pipeline from legal text to these formulae.
Current technologies in legal informatics, both at the academic and the industrial level, restrict their scope and capabilities to the terminological/syntactic level. Examples are the Eunomos [Boella et al., 2016] or the Eurocases system (eurocases.eu) distributed by APIS (http://www.apis.bg/en/) the Bulgarian company that hosted the secondment of ProLeMAS.
Although these techniques provide valid solutions to help navigate legislation and retrieve information, their overall usefulness and effectiveness are limited due to their focus on terminological issues while disregarding the specific semantic aspects of law.

This was the main issue/objective addressed in ProLeMAS, which was fully achieved by the design of a new logic, called reified Input/Ouput logic, whose definition has been published in the Journal of Logic and Computation [Robaldo and Sun, 2017].

Reified Input/Output logic is currently proposed in several ongoing/future projects as the underlying formalism to represent norms in natural language. The most important is DAPRECO (https://www.fnr.lu/projects/data-protection-regulation-compliance-2) which has been retained for funding by the Luxembourg National Research Fund in October 2016.

On the other hand, although experiments on several novel NLP procedures and techniques has been carried out (results have been published in several conference papers) to process legal texts and reason on them, automatic conversion from legal text to reified Input/Output logic was deemed to be too hard. In light of this, the tool that will be developed in DAPRECO will interface NPL procedures to allow for the SEMI-AUTOMATIC constructions of the formulae.

More details are provided in the two sections below.
The main result of ProLeMAS has been the definition of reified Input/Output logic, a novel formalisms for representing norms coming from existing legislation, which are available in natural language only [Robaldo and Sun, 2017].

Another relevant publication in ProLeMAS is [Sun and Robaldo, 2017], which has been published in the Journal of Applied Logic. This work investigates the complexity of Input/Output systems, in order to study the conditions under which they can be used in practical applications for the legal domain.

Furthermore, work has been done on legal document management systems and legal ontologies: Eunomos [Boella et al, 2016] and the Syllabus ontology [Ajani et al, 2017], which is used in Eunomos. In the future, Eunomos will be extended in order to broad its application scope. The possibility of initiating business activities in Luxembourg will be also considered, via Eunomos as well as via the development of applications in the data protection domain, which exploit the DAPRECO knowledge base.

Finally, together with Xin Sun, a former PhD student of the University of Luxembourg, deontic forms of reasoning in multi-agent systems has been investigated in several workshop/journal publications, e.g. [Xin and Robaldo, 2016].

Besides publications, dissemination of the results in ProLeMAS has been carried out via other two important channels:

- MIREL (http://www.mirelproject.eu): a Marie Skłodowska-Curie Research and Innovation Staff Exchange project aiming at creating an international and inter-sectorial network in legal informatics. MIREL involves 16 partners, among which APIS and Nomotika; Luxembourg university is the coordinator of the project. Several MIREL activities overlap with the ones of ProLeMAS, thus the result of ProLeMAS are currently disseminated through the MIREL network.

- LAST-JD (www.last-jd.eu): an interdisciplinary integrated doctorate in legal informatics. Luxembourg University is one of the partners and it hosted/hosts several PhD students enrolled in. the result of ProLeMAS are currently disseminated among them, in order to investigate possible overlapping between ProLeMAS results and their research activity.
"As it was planned in the original ProLeMAS proposal, reified Input/Output logic integrates the main insights of Input/Output logic, a well-known deontic framework defined in the past years by professor L. van der Torre (the supervisor of ProLeMAS) and the reification-based approach of prof. J. R. Hobbs designed for Natural Language Semantics (NLS).

I/O logic adopts norm-based semantics. It has been argued in past literature that norm-based approaches provide a better treatment of norms than possible world semantics. However, I/O logic was so far studied only from a theoretical point of view, in terms of propositional symbols only.

On the other hand, Hobbs's logic is a wide-coverage logic for NLS able to handle a fairly large set of linguistic phenomena into a simple first-order logical formalism. The choice of Hobbs's logic has been motivated by the need of keeping the architecture of the formulae as simple as possible, in order to increase readability and to help controlling the overall computational complexity. Since legislation is huge, simplicity is deemed to be a necessary feature for a logical formalism in legal informatics, in order to foster active collaboration of legal experts, usually having little expertise in logic, who can contribute to the building of large knowledge bases.

The first case study where reified I/O logic will be used is the data protection domain, specifically the translation of the provisions in the General Data Protection Regulation (GDPR, http://eur-lex.europa.eu/eli/reg/2016/679/oj) as well as some ISO security standards. The aim is to build a knowledge base of formalae in reified I/O logic representing the norms in the GDPR, and to devise future cutting edge applications in legal informatics using it.

As it is well-known, the GDPR is expected to have a significant impact on the EU society because it will revolutionize how enterprises have to protect individual's personal data records. The DAPRECO knowledge base will represent in a machine-readable format the provisions in the GDPR and some ISO security standard. The idea is to correlate the two kinds of provisions in order to measure the overlapping between them.

To build the DAPRECO knowledge base we need to overcome, on the one hand, the limitation of the manual translation of the provisions (this would be highly time-consuming and error-prone) and, on the other hand, to support current NLP technologies that, even at the best of their performances are however unable to automatically carry out the translation with a reasonable level of accuracy. Thus, DAPRECO will support the building of the knowledge base in a semi-automatic fashion.

The DAPRECO knowledge base is expected to have a central role in the next years in the data protection field. New projects will be prepared and proposed with the aim of exploiting the DAPRECO knowledge base. One of these project will be ""Privacy4Big"", which will be submitted to this forthcoming call:

https://ec.europa.eu/research/participants/portal/desktop/en/opportunities/h2020/topics/ds-08-2017.html

In Privacy4Big, it is proposed to interface the DAPRECO knowledge base with block-chain technologies, in order to devise privacy-compliant services."