Towards an Italian Healthcare Knowledge Graph

Electronic Health Records (EHRs), Big Data, Knowledge Graphs (KGs) and machine learning can potentially be a great step towards the technological shift from the one-size-fit-all medicine, where treatments are based on an equal protocol for all the patients, to the precision medicine, which takes count of all their individual information: lifestyle, preferences, health history, genomics, and so on. However, the lack of data which characterizes low-resource languages is a huge limitation for the application of the above-mentioned technologies. In this work, we will try to fill this gap by means of transformer language models and few-shot approaches and we will apply similarity-based deep learning techniques on the constructed KG for downstream applications. The proposed architecture is general and thus applicable to any low-resource language.