design and development of a customized language model for medicine: permed

This paper introduces permed, an enhanced version of the large medical language model meditron, which provides support for the persian language. meditron, as a specialized open-source model in the medical field, has demonstrated remarkable performance in english language processing and medical text analysis. however, the lack of persian language support has been identified as a key limitation of this model, hindering access to accurate medical information for persian-speaking professionals and patients. in this research, using translation tools and developed scripts, the permed model was designed with the capability to translate medical texts from english to persian, answer questions in persian, and facilitate full interaction in persian. permed's performance was evaluated on a set of medical tasks, including question answering, text analysis, and translation. the results showed that this model has significantly increased the accuracy and quality of its services in persian. by expanding access to medical knowledge for persian speakers, permed represents an effective step towards developing multilingual medical models and has high potential for use in multilingual medical environments.