Identifying cross-lingual temporal speech-markers as early cognitive decline indicators
Dementia represents a considerable global health challenge and is often diagnosed too late for effective intervention. Even in the early stages of dementia or in conditions preceding it, such as mild cognitive impairment (MCI), observable signs are already present, including measurable alterations in speech. For example, voice assistants could systematically collect and analyze speech data to identify "digital biomarkers" that accurately indicate underlying conditions. Research suggests that changes in speech timing—such as prolonged pauses or increased silences—can signal the presence of MCI or dementia. Notably, these temporal features are language-independent, enabling their universal application in analyzing speech from individuals with dementia across languages.
This thesis aims to investigate the significance of temporal speech features across different languages and to identify which features serve as universal indicators of MCI or dementia. Initially, relevant temporal features will be selected based on clinical research and existing studies. The Delaware dataset from DementiaBank, which includes English-language recordings from both MCI patients and healthy controls, will be used to evaluate the importance of the selected features. Subsequently, datasets from DementiaBank in other languages, for example, German, Korean, and Mandarin, will be analyzed using the same set of features. Finally, the findings regarding the significance of these features across the different language datasets will be compared, while accounting for variations in the tasks
| Attribute | Value |
|---|---|
| Title (de) | Identifying cross-lingual temporal speech-markers as early cognitive decline indicators |
| Title (en) | Identifying cross-lingual temporal speech-markers as early cognitive decline indicators |
| Project | |
| Type | Master's Thesis |
| Status | started |
| Student | Sofiia Danylchenko |
| Advisor | Alexandre Mercier |
| Supervisor | Prof. Dr. Florian Matthes |
| Start Date | 31.05.2026 |
| Sebis Contributor Agreement signed | yes |
| Checklist filled | yes |
| Submission date | 30.11.2026 |