AUTHOR=Yuan Jiahong , Cai Xingyu , Bian Yuchen , Ye Zheng , Church Kenneth 

TITLE=Pauses for Detection of Alzheimer’s Disease

JOURNAL=Frontiers in Computer Science

VOLUME=2

YEAR=2021

URL=https://www.frontiersin.org/journals/computer-science/articles/10.3389/fcomp.2020.624488

DOI=10.3389/fcomp.2020.624488

ISSN=2624-9898

ABSTRACT=<p>Pauses, disfluencies and language problems in Alzheimer’s disease can be naturally modeled by fine-tuning Transformer-based pre-trained language models such as BERT and ERNIE. Using this method with pause-encoded transcripts, we achieved 89.6% accuracy on the test set of the ADReSS (<underline>A</underline>lzheimer’s <underline>D</underline>ementia <underline>Re</underline>cognition through <underline>S</underline>pontaneous <underline>S</underline>peech) Challenge. The best accuracy was obtained with ERNIE, plus an encoding of pauses. Robustness is a challenge for large models and small training sets. Ensemble over many runs of BERT/ERNIE fine-tuning reduced variance and improved accuracy. We found that <italic>um</italic> was used much less frequently in Alzheimer’s speech, compared to <italic>uh</italic>. We discussed this interesting finding from linguistic and cognitive perspectives.</p>