Dipartimento di Ingegneria informatica, automatica e gestionale

This paper presents the problem statement and the research approach on an Italian project in the field of e-justice. We present the motivation and methodology for the application of an automatic writing assistant pipeline to Italian civil cases. The proposed solution is based on fine-tuning a transformer on a pre-processed corpus of Italian civil judgments. The resulting language model may be deployed as a writing assistant for legal users, in order to improve the efficiency of text writing, or further fine-tuned to be deployed in other law-related NLP tasks.