Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Modelling Legislative Systems into Property Graphs to Enable Advanced Pattern Detection

Andrea Colombo, Anna Bernasconi, Stefano Ceri

TL;DR

This work proposes to model legislation into a property graph, where edges represent citations, modifications, and abrogations between laws and their articles or attachments, both represented as nodes and edges with properties, and shows how such a property graph enables an efficient answer to complex and relevant queries previously impractical on raw text.

Abstract

Legislative systems face growing complexity due to the ever-increasing number of laws and intricate interdependencies between them. Traditional methods of storing and analyzing legal systems, mainly based on RDF, struggle with this complexity, hindering efficient knowledge discovery, as required by domain experts. In this paper, we propose to model legislation into a property graph, where edges represent citations, modifications, and abrogations between laws and their articles or attachments, both represented as nodes and edges with properties. As a practical use case, we implement the model in the Italian legislative system. First, we describe our approach to extracting knowledge from legal texts. To this aim, we leverage the recently internationally adopted XML law standard, Akoma Ntoso, to parse and identify entities, relationships and properties. Next, we describe the model and the schema implemented using Neo4j, the market-leading graph database management system. The schema is designed to capture the structure and hierarchy of laws, together with their interdependencies. We show how such a property graph enables an efficient answer to complex and relevant queries previously impractical on raw text. By leveraging other implementations of the Akoma Ntoso standard and the proposed property graph approach, we are confident that this work will facilitate a comprehensive comparison of legislative systems and their complexities.

Modelling Legislative Systems into Property Graphs to Enable Advanced Pattern Detection

TL;DR

This work proposes to model legislation into a property graph, where edges represent citations, modifications, and abrogations between laws and their articles or attachments, both represented as nodes and edges with properties, and shows how such a property graph enables an efficient answer to complex and relevant queries previously impractical on raw text.

Abstract

Legislative systems face growing complexity due to the ever-increasing number of laws and intricate interdependencies between them. Traditional methods of storing and analyzing legal systems, mainly based on RDF, struggle with this complexity, hindering efficient knowledge discovery, as required by domain experts. In this paper, we propose to model legislation into a property graph, where edges represent citations, modifications, and abrogations between laws and their articles or attachments, both represented as nodes and edges with properties. As a practical use case, we implement the model in the Italian legislative system. First, we describe our approach to extracting knowledge from legal texts. To this aim, we leverage the recently internationally adopted XML law standard, Akoma Ntoso, to parse and identify entities, relationships and properties. Next, we describe the model and the schema implemented using Neo4j, the market-leading graph database management system. The schema is designed to capture the structure and hierarchy of laws, together with their interdependencies. We show how such a property graph enables an efficient answer to complex and relevant queries previously impractical on raw text. By leveraging other implementations of the Akoma Ntoso standard and the proposed property graph approach, we are confident that this work will facilitate a comprehensive comparison of legislative systems and their complexities.
Paper Structure (15 sections, 4 figures, 3 tables)

This paper contains 15 sections, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Legislative Property Graph Schema
  4. The XML Akoma Ntoso standard
  5. Integration of additional data sources
  6. Proposed Graph Schema
  7. The Italian Legislative Property Graph (ILPG)
  8. ETL Pipeline
  9. Main Features of the ILPG
  10. Queries and Graph Patterns
  11. Law-centric Queries
  12. Legislative Pattern Queries
  13. Legislative Chain Paths
  14. Discussion and Limitations
  15. Conclusion

Figures (4)

  • Figure 1: Graph schema visualization with the properties of nodes and edges; the related PG-Schema angles2023pg is provided on our repository ourrepo. Each node and edge can be enriched with additional properties, i.e., either country-specific features or other attributes that can be derived from the text.
  • Figure 2: ETL pipeline to build the Italian Legislative Property Graph. At each step, new components of the PG are added.
  • Figure 3: Panel (a) shows published laws resulting from \ref{['q:lawProduction']}, the initial peak is due to the shift from the monarchy to the republic. Panel (b) shows the the result of \ref{['q:neverCited']} (the fraction of laws published at a given time and never cited). Both \ref{['q:lawProduction']} and \ref{['q:neverCited']} display a change in drafting laws occurred around the '80s, with \ref{['fig:query2']} also showing that many recent laws are not yet cited. In panel (c), we find the result of \ref{['q:outdatedLaws']}; here, we considered all laws published before the cut-off date 1970, and then we used D=1992 (a significant year in Italian politics -- the start of the so-called second republic) in order to compute the fraction of laws not cited after D. Panel (d) responds to \ref{['q:stockLaws']}; here we highlighted the drops corresponding to simplification decrees of 2008 and 2010.
  • Figure 4: Direct and indirect legal basis of the law of interest 2024/8 (blue node on the right-end). Its preamble cites three laws (2010/66, 1988/400, 2022/71) and an article of the Italian Constitution (87). Law 2010/300 has its foundation in law 1999/300, which, in turn, is based on articles of other laws (1 of 1998/191, 9 of 1999/50, 12 of 1997/59, and 7 of 1997/127). Thus, the latter laws are also indirectly relevant to law 2024/8.