The classification of legal texts is usually carried out by domain experts in force at institutions. The classification process is very complex because the reference thesauri are very rich, both in terms of variety of concepts and in terms of numbers. In addition, they often contain very rarely used labels. In this paper we show how to implement a Machine Learning system that can support the domain experts of the Italian Senate, handling infrequently used labels (Zero/Few-shot classification) and making the output of the model explainable to humans.
De Angelis, A., di Cicco, V., Lalle, G., Marchetti, C., Merialdo, P. (2022). Multi-Label Classification of Bills from the Italian Senate. In CEUR Workshop Proceedings (pp.39-53). CEUR-WS.
Multi-Label Classification of Bills from the Italian Senate
De Angelis A.;di Cicco V.;Merialdo P.
2022-01-01
Abstract
The classification of legal texts is usually carried out by domain experts in force at institutions. The classification process is very complex because the reference thesauri are very rich, both in terms of variety of concepts and in terms of numbers. In addition, they often contain very rarely used labels. In this paper we show how to implement a Machine Learning system that can support the domain experts of the Italian Senate, handling infrequently used labels (Zero/Few-shot classification) and making the output of the model explainable to humans.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
