Electronic health records can vary in length and complexity, making it challenging for clinicians to quickly access important information about their patients. To address this issue, we aim to fine-tune Gatortron to perform three critical downstream tasks: Named Entity Recognition (NER), Relation Extraction (RE), and Question Answering (QA).

NER can identify key entities such as drugs, drug attributes (frequency, dosage, etc.), reasons for taking the drug, and adverse events related to drug usage. RE can identify relationships between these entities. And QA can answer questions about drugs taken by a patient, for example, "How long has John been taking Cefepime?" With these models in place, clinicians can easily identify drug allergies and ask relevant questions about drugs taken by their patients, leading to more informed and effective medical decisions.