Current methods for diagnosis of Alzheimer’s disease rely on structured interviews and cognitive tests performed by trained personnel in clinical settings; there is therefore much interest in automatic methods to increase availability and access. One possibility is via language: cognitive impairments are associated with characteristic effects on patients’ speech. Many of these effects can be detected automatically by computational methods, thus potentially leading to future assistive diagnosis and monitoring technology for clinicians.
Vast majority of this work has been done in English, and (as far as we are aware) none in Slovene. It is not clear how well current methods and models can transfer to other languages, particularly less-resourced languages like Slovene, and to what extent they are dependent on a given language’s vocabulary, syntax, prosody or interaction patterns.
The overall contribution of the project is to discover how models for cognitive impairment detection can be built that transfer well across languages, while giving interpretable output that could be used by clinicians.