My personal small project on word vector representation study for drug data. I temporally named drug2vec simply but it's not exactly 'drug' properties as i used prescription codes from EHR data (MIMIC) by now. To my disappointment i couldn't discovered such a meaningful clusters for vectors i think it's because mimic is Critical Care data and doesn't include long-term prescriptions or diagnosis information. But interestingly, a few prescriptions on chronic diseases like the statins, antibiotics, vitamin with mineral and IV solutions showed clusters.