基于深度学习的老年认知障碍与语言特征研究  

Deep-Learning-Based Research on Cognitive Impairment and Linguistic Features in Older Adults

作　　者：黄立鹤[1] 叶子 HUANG Lihe;YE Zi

机构地区：[1]同济大学老龄语言与看护研究中心/外国语学院,上海200092

出　　处：《外语与外语教学》2024年第4期81-90,149,共11页Foreign Languages and Their Teaching

基　　金：国家社科基金项目“基于机器学习的认知障碍人群语言特征及自动筛查研究”(项目编号:24BYY120)的阶段性成果。

摘　　要：近年来,以深度学习为代表的人工智能技术是老年语言学研究中值得关注的新方法。利用深度学习模型,基于语言标志物的老年人认知筛查手段正在成为国际研究热点,可弥补基于神经影像学、体液标志物的传统方法存在的采样成本高、具有侵入性的不足。本文从深度学习与神经网络的背景出发,梳理了神经网络方法在基于语言标志物对老年人开展认知障碍检测研究中各环节的应用情况,介绍了国内外学界在提升模型性能的新技术、嵌入特征的可解释性,以及医疗领域大模型三方面的前沿进展,并对未来以汉语母语老年人为对象的相关研究方向作出展望。利用深度学习技术赋能认知筛查是语言学服务人类卫生健康的新兴路径,具有广阔前景。In recent years,artificial intelligence technologies,particularly deep learning,have emerged as noteworthy new approaches in gerontolinguistics.Utilizing deep learning models for cognitive screening of the older adults based on linguistic markers has become an important international research topic.This method addresses the high sampling costs and invasiveness associated with traditional techniques based on neuroimaging and fluid biomarkers.This paper,starting from the background of deep learning and neural networks,reviews the application of neural network methods in various stages of cognitive impairment detection in older adults using linguistic markers.It introduces the latest advancements in enhancing model performance,the interpretability of embedding features,and large language models for medicine at home and abroad.Furthermore,it proposes future research directions focusing on native Chinese-speaking older adults.In sum,utilizing deep learning technology to empower cognitive screening is an emerging pathway for linguistics to serve human health,which has broad prospects.

关 键 词：老年语言学 认知障碍检测 语言标志物 深度学习 神经网络 

分 类 号：H319[语言文字—英语]