Clustering Italian medical texts: a case study on referrals

In the medical domain, there is a large amount of valuable information that is stored in textual format. These unstructured data have long been ignored, due to the difficulties of introducing them in statistical models, but in the last years, the field of Natural Language Processing (NLP) has seen relevant improvements, with models capable of achieving relevant results in various tasks, including information extraction, classification and clustering. NLP models are typically language-specific and often domain-specific, but most of the work to date has been focused on the English language, especially in the medical domain. In this work, we propose a pipeline for clustering Italian medical texts, with a case study on clinical questions reported in referrals