Mental state detection and tagging in nursing records

Staff at geriatric care facilities compile nursing records, containing information from patients' vital signs or treatments suggested by doctors, to comments about patients interactions with the nursing staff, their families and other patients. Especially the latter type of entries often seems to include clues to patients' emotional well-being. Following the assumption that physical and mental health exert a mutual influence on each other, the authors believe that explicitly monitoring patients' emotions and moods can enhance the understanding of changes in physical health. It may also assist nurses in, e.g., preventing negative emotional states like persistent depression affecting patients' overall health for the worse. This paper proposes a strategy to use machine learning techniques to detect and classify emotion in nursing records. Since a first annotation step revealed that entries containing direct speech seem to be especially “emotionally salient”, special focus of our future work will be on those entries.