The project ProLeMAS aims at (1) using reification to fill the gap between the current formalizations in deontic logics and the richness of natural language semantics and (2) implementing tools for (semi-)automatically building machine-readable representations from legal texts via NLP. The supervisor of the project is Leon van der Torre.
The following two main limitations of current approaches in deontic logic were identified and are addressed in ProLeMAS:
- Proposals in deontic logic are typically propositional, i.e. their basic components are whole propositions. A proposition basically refers to a whole sentence. On the other hand, NLS includes a wide range of fine-grained intra-sentence linguistic phenomena: named entities, anaphora, quantifiers, etc. It is then necessary to move beyond the propositional level, i.e. to enhance the expressivity of deontic logic to formalize the meaning of the phrases constituting the sentences.
- Few proposals in deontic logic have been implemented and tested on real legal text. Most of them are only promising methodologies, which overcome the limits of other approaches on the theoretical side. It is time to see whether the proposed formalizations are expressive enough to handle the fine-grained linguistic phenomena mentioned above.
ProLeMAS aims at designing neo-Davidsonian logical representations of legal text on which enabling legal reasoning. Specifically, a new logical framework, called reified Input/Output logic, has been devised. Reified Input/Output logic integrates the reification-based logic defined by Jerry R. Hobbs, on which I gained expertise in the context of the PDTB (cf. (Robaldo and Miltsakaki, 2014)), within the Input/Output logic proposed and extensively studied by prof. van der Torre (for an introduction, cf. [Makinson and van der Torre, 2000])
Reified Input/Output logic is used in the project DAPRECO to model norms from the upcoming General Data Protection Regulation (GDPR) and several ISO standards, whose norms are represented as reified Input/Output logic formulae.
Furthermore, by exploiting NLP techniques such as Dependency Parsing, Named Entity Recognition, and statistical methods, a suite of NLP tools have been developed in order to convert legal documents in AkomaNtoso , identify named entities (proper nouns of lawyers, judges, companies, institutions, etc., geographical entities, time expressions, quantities of money, etc.), obligations, rights, etc. as well as to perform reasoning in multi-agents systems.
Publications in ProLeMAS I have authored are: (Robaldo and Sun, 2017), (Sun and Robaldo, 2017), (Ajani et al., 2017), (Boella et al., 2016), (Robaldo et al., 2016), (Adebayo et al., 2016), (Bartolini et al., 2016), (Robaldo and Sun, 2016), (Sun and Robaldo (a), 2016) (Sun and Robaldo (b), 2016), (Sun and Robaldo (c), 2016), (Humphreys et al. (a), 2015) .