Alzheimer's Disease Detection from Spontaneous Speech through Combining Linguistic Complexity and (Dis)Fluency Features with Pretrained Language Models

arXiv (Cornell University)

Abstract

In this paper, we combined linguistic complexity and (dis)fluency features with pretrained language models for the task of Alzheimer's disease detection of the 2021 ADReSSo (Alzheimer's Dementia Recognition through Spontaneous Speech) challenge. An accuracy of 83.1% was achieved on the test set, which amounts to an improvement of 4.23% over the baseline model. Our best-performing model that integrated component models using a stacking ensemble technique performed equally well on cross-validation and test data, indicating that it is robust against overfitting.

Authors 4

1. Yu Qiao Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

2. Xuefeng Yin Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University

3. Daniel Wiechmann

   University of Amsterdam

   Affiliation as printed

   University of Amsterdam

4. Elma Kerz Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University