Exploring neural models for predicting dementia from language. (July 2021)

• Record Type: Journal Article
• Authors: Kong, Weirui; Jang, Hyeju; Carenini, Giuseppe; Field, Thalia S.
• Is Part Of: Computer speech & language. Volume 68(2021)
• Publication Date: 2021-07
• Subjects: Automatic dementia prediction -- Multimodal embedding -- Coherence model -- Hierarchical attention networks; Speech processing systems -- Periodicals; Automatic speech recognition -- Periodicals; Computers -- Periodicals; Linguistics -- Periodicals; Speech-Language Pathology -- Periodicals; Traitement automatique de la parole -- Périodiques; Reconnaissance automatique de la parole -- Périodiques; Automatic speech recognition; Speech processing systems; Electronic journals; Periodicals; 006.454
• DOI: 10.1016/j.csl.2020.101181
• Languages: English
• ISSNs: 0885-2308

Abstract

Highlights: We use neural network models for dementia prediction to avoid task-specific features. Our model for combining multimodal features give performances comparable to baselines. Coherence scores are not different between dementia patients and healthy controls. Our HAN-AGE model achieves state-of-the-art performance for dementia detection. Abstract: Early prediction of neurodegenerative disorders such as Alzheimer's disease (AD) and related dementias may facilitate earlier access to medical and social supports. Further, detection of individuals with preclinical disease may help to enrich clinical trial populations for studies examining disease-modifying interventions. Changes in speech and language patterns may occur in the early stages of neurodegenerative diseases such as AD and frontotemporal dementia, with worsening as the disease progresses. This has led to recent attempts to create automatic methods that predict cognitive impairment and dementia through language analysis. Previous works have improved the prediction accuracy by introducing some task-specific features in addition to task-agnostic linguistic and acoustic features. However, task-specific features prevent the model from generalizing to other tests and languages. In this paper, we focus on exploring the effectiveness of neural network models that require no task-specific feature for dementia prediction in three different ways. First, we use a multimodal neural model to fuse linguistic features and acoustic features, and investigate the performance change compared to simply concatenating these features. Second, we propose a novel coherence feature generated by a neural coherence model, and investigate the predictiveness of this new feature for dementia prediction. Finally, we apply an end-to-end neural method which is free from feature engineering and achieves state-of-the-art classification result on a widely used dementia dataset.