Establishing the DementiaBank Delaware Corpus: an Online Multimedia Database for the Study of Language and Cognition in Dementia

Background To better understand the progressive decline of language abilities in aging and dementia, we expanded the quality and quantity of resources in DementiaBank — an open‐access database of multimedia spoken language interactions for the study of speech and language abilities across the progression of dementia. This work builds from the success of the TalkBank Project with regard to data sharing, transcription, analysis, and web delivery. Specifically, this work collected connected speech and language data to develop the new “Delaware corpus” and to share resources for future analyses. Methods Cognitively unimpaired adults (n = 20) and adults with mild cognitive impairment (MCI) (n = 35) completed the DementiaBank discourse protocol and a cognitive‐linguistic battery in a 90‐minute session by Zoom with an administrator. The discourse protocol elicited four types of discourse: picture description, story narrative, procedural discourse, and personal narrative. Results The new Delaware corpus is available on the DementiaBank website. Data include spoken language CHAT transcripts linked to audio, test results, and medical/demographic information for 55 participants. CHAT is a standardized format for transcription and coding that is compatible with the CLAN language analysis software. CLAN provides methods for linking transcripts to media and automatic computation of a wide range of morphosyntactic features and other complex linguistic analyses. Additional resources include an automatic speech recognition pipeline code with tutorials to help streamline the transcription process and a collaborative commentary system that allows consortium members to insert comments or codes directly into online transcripts linked to media files. The collaborative commentary system is beneficial for training research assistants in coding techniques and for collaborative analysis of discourse samples. Conclusions Detailed digital measurement of speech and language abilities of older adults with and without mild cognitive impairment may help advance research and clinical practice by expanding the subtyping of MCI to include speech/language profiles. Open‐access databases, like DementiaBank, facilitate research collaborations, support advancements in computational analyses, and improve research dissemination to advance dementia science. The next steps for this project involve recruiting more researchers, clinicians, and students to use the new protocol and contribute their data to grow the database.