Skip to main content

MIREL - MIning and REasoning with Legal texts

Periodic Reporting for period 2 - MIREL (MIREL - MIning and REasoning with Legal texts)

Reporting period: 2018-01-01 to 2019-12-31

The project MIREL aimed at bridging the gap between the communities working in Mining and Reasoning on legal texts, which previously have mostly worked in isolation.
Although current technologies provide valid solutions to help navigate legislation, their overall usefulness is limited due to their focus on terminological issues while disregarding the specific semantic aspects of law, which allows legal reasoning. MIREL was proposed to promote mobility and staff exchange in order to create an inter-continental inter-disciplinary consortium in Law and AI areas including Natural Language Processing, Computational Ontologies, Argumentation, and Logic & Reasoning. Legal Informatics is growing also at the commercial level, as witnessed that MIREL included three industrial partners in the legal tech.
The project was structured in four main work packages and lasted four years, from 1.1.2016 until 31.12.2019. Travels were mostly devoted to import competences traditionally lacking in Europe. In particular, researchers from UNLP who visited UNIBO brought their expertise in legal theory to strengthen theoretical foundations of machine-readable formats to represent legal texts, researchers from UNS who visited European universities, as well as the researchers from Europe who visited ROIS, Data61 (NICTA), and ZJU jointly developed novel frameworks and solutions for argumentation, both at the theoretical level (WP1), in argumentation mining (WP2), and as an tool for reasoning on legal texts (WP3); researchers from UL, UNIBO, and UNITO who visited Stanford University, had the possibility to acquire competences needed to design and implement novel instruments such as the Privacy Ontology, the DAPRECO knowledge base, and various NLP software modules to extract data from legal texts (WP2). Finally, UCT contributed to the research in description logic for reasoning about legal ontologies, with HUD and UL in particular (WP3).
Concerning use cases (WP4), INRIA and Cordoba extended former tools for processing licenses, contracts, technical documents, and multi-lingual corpora of norms, while evaluating them on data made available by Nomotika SRL in particular. Furthermore, Nomotika SRL developed a demo in Semantic Media Wiki (SMW) of the legal document management system MenslegiS, looking for a new version of MenslegiS that may reduce manual work; however, the final evaluation revealed that the SMW demo still requires lot of manual work to update the knowledge bases. On the other hand, APIS used its travels to UNITO in order to train their programmers on novel NLP techniques that were later used within their services. Finally, DLVSystem worked closely with HUD, UNS, and Data61 (NICTA) on Answer Set Programming, while extending and refining their software suite to process legal documents.
The work in WP1 was oriented towards: carrying out transdisciplinary research and devising conceptual models for legal knowledge representation and reasoning, devising formal languages for representing norms, policies, and values in the law, and Logics for modelling the interpretation of legal provisions. The activities done have been carried out by UNIBO, Stanford Univ., UNITO, Data61, UNLP, UNS, UL, INRIA, and ZJU.
The mentioned partners: (1) designed and built an OWL ontology for the privacy and data protection domain with a special focus on the GDPR, (2) devised a novel formalization of a belief revision operator applied to an epistemic model that considers rules and time, (3) investigated new formal methods for deontic reasoning in legal interpretation, in particular those concerned with deontic argumentation, (4) carried out model-theoretic investigations of monotonic modal logics, and (5) started novel theoretical research in legislative dialogues.
The work in WP2 was oriented towards: studying natural language semantics for normative knowledge, developing and extending existing NLP systems for mining both concept to be linked to the T-BOX and named entities in order to populate the A-BOX, building GOLD standard corpora for syntactic and semantic analysis. The activities done have been carried out by UNITO, UL, UNIBO, INRIA, Cordoba Univ., Nomotika SRL, ZJU, Stanford Univ, and ROIS.
The mentioned partners: (1) developed a new knowledge base of reified formulae for the GDPR built on the PrOnto ontology from WP1 and integrating insights from the literature in natural language semantics, (2) implemented cutting-edge procedures for named entity recognition and classification, legal textual entailment, open information extraction from legal corpora, and argument mining, (3) implemented a demo of the Eunomos legal document management system in Semantic Media Wiki, (4) started a collaborative work on argumentative reasoning and persuasion, which will strengthen models for judges reasoning aiming at providing explanations for their conclusions, (5) implemented a three-dimensional classification of legal ontologies and tools to support their exploration, (6) carried out basic research in rules of conflict of laws.
The work in WP3 was oriented towards: ontology-based access to normative knowledge, computational solutions for decision making and compliance, massive parallelization. The activities done have been carried out by HUD, DLVSystem, Data61, UL, UCT, UNS, ZJU, and UNIBO.
The mentioned partners: (1) designed and implemented a new platform for formalizing legal interpretations and their compilation into legal knowledge bases, (2) devised novel procedures for argumentation-based reasoning for the legal domain, including structured argumentation in DeLP, (3) conducted basic research in ontology-based abstract normative systems, (4) extensively investigated and compared reasoning mechanisms for the legal domain, with a special focus on Defeasible Deontic Logic, Answer Set Programming and Argumentation-based normative reasoning, (5) and investigated the feasibility of using AI techniques in order to support the evaluation of legal aspects for the PESTLE analysis.
The work in WP4 was oriented towards: user needs study, licenses and contracts case study, technical documents case stud, and the multilingual corpora of norms case study. The activities done have been carried out by INRIA, Cordoba Univ., UNITO, Nomotika SRL, Data61, and UL.
The mentioned partners: (1) evaluating NLP techniques on Nomotika SRL datasets, (2) evaluated novel Argument Mining techniques on multi-lingua corpora of norms, (3) and modelled dialogues for optimal legislation, following the theoretical results in legislative dialogues obtained in WP1.
The work done in MIREL led to 114 publications, among which 33 journals, in top venues in legal informatics. Research results greatly contributing to the progress beyond the state of the art.
Overall, we believe the MIREL project succeeded in addressing its objectives, first of all in bridging the gap in legal informatics between the community working on mining and the one working on reasoning. This has been achieved via the organization of several conferences, workshops, and events, as well as with new projects retained for funding, which will allow the MIREL partners to continue the dissemination, communication, and networking strategy of the project in the forthcoming years.
MIREL partners - Meeting
MIREL logos