ADHD

LLM-Assisted Discovery of ADHD Signals in Turkish Teacher Narratives Beyond Rating Scales

This study analyzes de-identified Turkish teacher evaluation forms from clinical ADHD assessments, comparing predictive signals from structured rating scales (CTRS-R:S) and open-ended teacher narratives. The authors find that structured and narrative information encode complementary signals, with minimal overlap between cases missed by each modality. An LLM-assisted theme discovery pipeline reveals distinct attention, behavioral, and family-related patterns in narratives that structured scales miss, demonstrating NLP's potential to augment traditional ADHD screening.