Mapping semantic networks to Dutch word embeddings as a diagnostic tool for cognitive decline

This study explores the relationship between semantic networks and cognitive decline, specifically in the Dutch language, with the objective of expanding the existing literature in this area to a different linguistic context. Semantic networks serve as abstract representations of the semantic memory system. Semantic networks can be estimated through verbal fluency data, enabling valuable insights into higher cognitive concepts, including cognitive decline. While the number of studies supporting the connection between semantic networks and cognitive decline continues to increase, most evidence has been obtained primarily for the English language. For less widely studied languages, like Dutch, this relation has not yet been investigated. To bridge this gap, we employ an experimental paradigm to examine the connection between semantic networks and cognitive decline, utilizing Dutch verbal fluency data. Leveraging a comprehensive dataset containing semantic verbal fluency data from mildly impaired Alzheimer's patients (Konijnenberg et al., 2018), we employ the Dutch language model BERTje to construct individual semantic networks. These networks are built based on cosine similarity, representing the semantic relatedness between words in the verbal fluency data. Furthermore, we discuss the predictive capability of these network scores in relation to participants' Mini-Mental State Examination (MMSE) results. By exploring this relationship in the Dutch language, we shed light on the unique characteristics of semantic networks and their potential role in assessing cognitive decline in populations speaking less commonly studied languages. Overall, this study expands our understanding of the connection between semantic networks and cognitive decline by venturing into the realm of the Dutch language, thus broadening the scope of existing literature and paving the way for future research in similar linguistic contexts.