Computerized assessment of linguistic indicators of lucidity in Alzheimer's Disease dementia

项目摘要 母项目（AG 069792）的重点是实现自动检测和分析 在患有晚期痴呆症的个体中意外的清醒发作，其中个体 长期以来，人们一直认为他或她患有痴呆症，失去了大部分认知能力 暂时恢复以清晰和连贯的方式进行沟通的能力。为实现这一 目标，该项目旨在开发一种）自动转换语音的技术， 和B）来自转录文本的语义连贯性的测量。 这项技术的发展依赖于最先进的人工智能（AI）和 机器学习（ML）方法，包括深度学习和时间序列分析。我们 目前正在开发这些方法，使用现有的数据集，包括卡罗莱纳 对话语料库（CCC）、威斯康星州纵向研究（WLS）和痴呆症银行（DB） 其中包含与痴呆症患者进行对话的音频和文本记录 不同严重程度和健康对照。目前，AI/ML社区的研究人员使用 这些资源通过以下单独开发的预处理程序， 在其工作产生的出版物中进行了描述，并在多个版本中提供了特设代码 储存库。不同研究人员获得的结果可能难以比较，因为 在如何处理数据和准备ML实验方面的个体差异。 对于为期一年的补充项目，我们建议创建一个开源平台， 这些工具将从DB、WLS和CCC数据集获取原始数据，并将其转换为 他们必须为AI/ML做好准备，以符合AI/ML社区当前的最佳实践。

Project Summary The focus of the parent project (AG069792) is to enable automated detection and analysis of episodes of unexpected lucidity in individuals with late-stage dementia in which the individual long thought to have succumbed to dementia and lost most of his or her cognitive abilities temporarily regains the ability to communicate in a clear and coherent fashion. Towards this goal, this project aims to develop a) technology for automatic conversion of speech produced by patients with dementia to text and b) measures of semantic coherence from the transcribed text. The development of this technology relies on state-of-the-art artificial intelligence (AI) and machine learning (ML) methods including deep learning and time series analysis. We are currently developing these approaches using existing datasets including the Carolina Conversations Corpus (CCC), Wisconsin Longitudinal Study (WLS) and Dementia Bank (DB) which contain audio and text transcripts of conversational interviews with patients with dementia of varying severity and healthy controls. Currently, the researchers in the AI/ML community use these resources by following individually developed pre-processing procedures that are partially described in publications resulting from their work and ad hoc code made available in multiple repositories. The results obtained by various investigators can be difficult to compare because of individual differences in how the data were processed and prepared for ML experimentation. For the one-year supplement project, we propose to create an open-source platform consisting of tools that will ingest original data available from DB, WLS, and CCC datasets and convert them to be AI/ML-ready in keeping with the current best practices in the AI/ML community.