Alignment of narrative retellings for automated neuropsychological assessment

As neurological disorders such as dementia, autism, and language impairment become more common, there is growing need for simple, objective screening tools. Narrative retelling tasks—where subjects listen to a short story and retell it—offer promise because these abilities are often impaired in affected populations. This dissertation evaluates the reliability and diagnostic utility of automated narrative analysis. We introduce a novel word alignment method using random walks, achieving faster and more accurate alignment than traditional approaches. Automated scoring based on these alignments predicts impairment with accuracy comparable to manual scoring and generalizes to other language samples, supporting its potential as a flexible screening tool.