Publications by authors named "Fumimaro Doi"

Article Synopsis
  • Recent developments in natural language processing aim to aid psychiatric diagnosis by analyzing the history of present illness (HPI), overcoming limitations of prior studies that often focused on a few major disorders or small sample sizes.
  • The research involved extracting HPIs and diagnostic information from 2,642 cases at a Japanese hospital and utilizing a specialized model (UTH-BERT) to predict diagnoses, comparing it against actual psychiatrist evaluations.
  • The model achieved a diagnosis match rate of 74.3%, outperforming semi-experienced and novice psychiatrists, suggesting its potential use as a diagnostic tool in clinical settings.
View Article and Find Full Text PDF