English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Alzheimer's disease (AD) is a progressive neurodegenerative disease andrecently attracts extensive attention worldwide. Speech technology isconsidered a promising solution for the early diagnosis of AD and has beenenthusiastically studied. Most recent works concentrate on the use of advancedBERT-like classifiers for AD detection. Input to these classifiers are speechtranscripts produced by automatic speech recognition (ASR) models. The majorchallenge is that the quality of transcription could degrade significantlyunder complex acoustic conditions in the real world. The detection performance,in consequence, is largely limited. This paper tackles the problem viatailoring and adapting pre-trained neural-network based ASR model for thedownstream AD recognition task. Only bottom layers of the ASR model areretained. A simple fully-connected neural network is added on top of thetailored ASR model for classification. The heavy BERT classifier is discarded.The resulting model is light-weight and can be fine-tuned in an end-to-endmanner for AD recognition. Our proposed approach takes only raw speech asinput, and no extra transcription process is required. The linguisticinformation of speech is implicitly encoded in the tailored ASR model andcontributes to boosting the performance. Experiments show that our proposedapproach outperforms the best manual transcript-based RoBERTa by an absolutemargin of 4.6% in terms of accuracy. Our best-performing models achieve theaccuracy of 83.2% and 78.0% in the long-audio and short-audio competitiontracks of the 2021 NCMMSC Alzheimer's Disease Recognition Challenge,respectively.

Exploiting Pre-Trained ASR Models for Alzheimer's Disease Recognition Through Spontaneous Speech