Биологически правдоподобная нейронная модель семантической деменции

In this study, we discuss modern theories of semantic system (Mahon & Caramazza, 2008; Mahon, 2015; Rogers et al., 2004; Patterson, Nestor, & Rogers, 2007; Gallese & Lakoff, 2005; Kiefer & Pulvermüller, 2012; Garagnani & Pulvermüller, 2016; Binder & Desai, 2011; Binder, 2016; Barsalou, 2016) and different approaches to model it (Huth et al., 2016; Anderson et al., 2016; Chen et al., 2017; Garagnani & Pulvermüller, 2016; Tomasello et al., 2019). We argue that hybrid theories — which postulate both substantial grounding in the sensory-motor system and higher-order convergence zones — are the most promising way to think about semantic representations (Kiefer & Pulvermüller, 2012; Garagnani & Pulvermüller, 2016; Binder & Desai, 2011; Binder, 2016; Barsalou, 2016). We use biologically- inspired neuronal model of the semantic system, built within this hybrid paradigm (Garagnani & Pulvermüller, 2016; Tomasello et al., 2018), as the basis, and augment it with the semantic dementia mechanism, as SD is the key disease which influences the semantic system (Hodges & Patterson, 2007; Montembeault et al., 2018; Spinelli et al., 2017). We demonstrate that our SD model’s predictions can help with a better understanding of the semantic system organization. We also show how our baseline model puts into question the most common arguments against hybrid theories. Finally, we highlight the necessity to improve such models if we want to be more productive in our research on semantic system understanding.