Legal Knowledge Extraction for Knowledge Graph Based Question-Answering

Abstract

This paper presents the Open Knowledge Extraction (OKE) tools combined with natural language analysis of the sentence in order to enrich the semantic of the legal knowledge extracted from legal text. In particular the use case is on international private law with specific regard to the Rome I Regulation EC 593/2008, Rome II Regulation EC 864/2007, and Brussels I bis Regulation EU 1215/2012. A Knowledge Graph (KG) is built using OKE and Natural Language Processing (NLP) methods jointly with the main ontology design patterns defined for the legal domain (e.g., event, time, role, agent, right, obligations, jurisdiction). Using critical questions, underlined by legal experts in the domain, we have built a question answering tool capable to support the information retrieval and to answer to these queries. The system should help the legal expert to retrieve the relevant legal information connected with topics, concepts, entities, normative references in order to integrate his/her searching activities.

From Philosophy to Interfaces: an Explanatory Method and a Tool Inspired by Achinstein’s Theory of Explanation

“…This way we can easily make them inter-operate with deep-learning based QA algorithms and existing language models. More in detail, as in [27], we perform KG extraction by:…”

A dataset for evaluating legal question answering on private international law

“…In this Section, we explain how we expanded the dataset presented in [13], doubling its size. We improved over [13],…”

“…It is important to highlight the fact that, for the construction of the new dataset, we decided to inherit some methodological choices from [13], considering PIL as a subject simply from the point of view of these three EU Regulations, as a self-contained environment, i.e., excluding references to other international conventions and general principles. So that it is possible to evaluate Q&A techniques with respect to their ability to handle the general principles in the recitals, the scope of application in the initial articles, and the specific cases (e.g.…”

“…While, in 2015 Kim et al [10] presented one of the very first algorithms based on DNNs for Legal Question Answering (reasoning) applied to a dataset of Boolean questions from Japanese legal bar exams, then followed up by [7] and others [9,12]. In 2020, Sovrano et al [13] proposed a novel and hybrid approach for legal question answering on PIL, using a legal ontology based on Ontology Design Patterns (like agent, role, event, temporal parameter, action) in order to mirror the legal significance of the relationships within and among the provisions. More generally, automated reasoning over legal texts (not just the PIL's ones) is not a trivial task, due to the fact that the legal jargon (legalese) is less frequent and more ambiguous than commonly-used natural language.…”

“…Our goal is to be able to properly evaluate canonical question answering techniques for PIL. This is why we try to expand the work presented by Sovrano et al in [13], publishing a larger and more curated dataset In Section 2 we describe our dataset, and the methodology we followed to design it. While in Section 3 we analyse the results obtained by re-running the experiment of [13] on the new dataset, pointing to future work in Section 4.…”