Text mining and network analysis to support improvements in legislative action. The case of the earthquake in Emilia-Romagna

http://jadt2016.sciencesconf.org/ | http://jadt2016.sciencesconf.org/submission/submit
Abstract submitted to
JADT2016 Journées internationales d’Analyse statistique des Données Textuelles
7-10 Juin 2016 Nice (France)
Text mining and network analysis to support improvements in legislative action. The case of the earthquake in Emilia-Romagna
By Pasquale Pavone, Simone Righi, Margherita Russo
^Dipartimento di Economia Marco Biagi, Università di Modena e Reggio Emilia, Italy, pasquale.pavone@unimore.it, margherita.russo@unimore.it, simone.righi@unimore.it
15 November 2015
Key words: automatic classification; legislative corpus; clustering and network analysis
Abstract
In the three years after the 2012 earthquake in Emilia-Romagna, through the enactment of more than 350 ordinances, the Commissioner has structured interventions to cope with emergency and reconstruction. The intense law-making, essential to fill a legal vacuum, has enabled to overcome the uncertainties of the difficult phase of recovery. There is agreement among experts that a large number of those ordinances was due to the absence of national rules governing the urgent intervention in case of natural disasters. On the push of actions taken in Emilia-Romagna, the Italian Parliament has reopened the debate on a national law on emergency after natural disasters.
Through a systematic content analysis of the corpus of ordinances issued in Emilia-Romagna, in this paper we propose a contribution in drafting a law on emergency. Two main strands of analysis have been developed.
In the first one, an automatic text analysis, supported by Taltac2, has provided inputs for a factor analysis and a cluster analysis of the thematic areas covered by the ordinances. Four main topics have been singled out: grant criteria and contributions; management of allocation of resources; urgent works for municipalities, schools and churches buildings; interventions to support population. Having associated each ordinance to one of the four topics, a temporal analysis of the issues addressed during the emergency and reconstruction phase highlights the sequence of actions that were undertaken in Emilia.
In a second step, the set of terms characterizing each cluster it is used to obtain a redefinition of disjunctive classification towards a fuzzy multi-class.
Furthermore, by defining a specific textual lexicon model, it was possible to extract the legislation (and other ordinances) cited in the ordinances and identify clusters of regulatory areas of reference to meet the emergency and reconstruction following natural disasters. In this second strand, clusters of citations were detected with an algorithm of network analysis (Infomap) based on information theory (Rosvall & Bergstrom, 2008; De Domenico et al, 2015). This analysis highlights overlapping subsets of nodes intensely connected and allows to outline which are the most relevant issues involved in the areas of intervention after a natural disaster.
Clustering produced with automatic text analysis has also provided an input for a more detailed tagging of each ordinance, according the categories and sub topics detected in our analysis, developed in the Energie Sisma Emilia research project. This tagging has been applied in the complementary paper developed, within the research project, by Palmirani et al. that – among other things supports also a check for the robustness of the cluster analysis produced through automatic text analysis, and for the network analysis produced with Infomap.
Network analysis combined to text analysis support some policy recommendations to improve the legislation of emergence under scrutiny that could not have been reached by reading the individual ordinances and legislation.