Redundancy, isomorphism, and propagative mechanisms between emotional and amodal representations of words: A computational study

Emotional Valence Precedes Semantic Maturation of Words: A Longitudinal Computational Study of Early Verbal Emotional Anchoring

“…This mechanism could propagate the modal properties of some amodal representations to other amodal representations and generate inferences (like eliciting emotional responses) based on their relations with other words (see a graphical illustration in Figure 2). Different studies have shown that this is a plausible mechanism for the generation of emotional (Hollis et al., 2017; Martínez‐Huertas et al., 2021) and perceptual (Günther et al., 2020; Günther et al., 2020) responses. Psychobiological evidence has been found for this proposal (Binder, 2016; Nastase & Haxby, 2016) and neural network architectures have been successfully used to computationally model it (Günther et al., 2020; Hoffman et al., 2018; Howell et al., 2005; Martínez‐Huertas et al., 2020, 2021).…”

“…Modal and amodal word representations can be reached even with no direct experience with emotional or sensorimotor responses. Some studies have shown this phenomenon in human development (Field & Schorah, 2007;García-Palacios et al, 2018;Grégoire & Greening, 2020) and also in computational modeling (Hofmann et al, 2018;Hollis et al, 2017;Howell, Jankowicz, & Becker, 2005;Martínez-Huertas et al, 2021). Thus, anchored words can propagate the emotional or sensorimotor information to words without direct exposure (Slousky & Deng, 2019).…”

“…Theoretical proposals like the language and situated simulation theory (Barsalou, Santos, Simmons, & Wilson, 2008) or the symbol interdependency hypothesis (Louwerse, 2011, 2018) have been developed to explain how these amodal representations can generate emotional and other sensorimotor word representations. Additionally, some mapping mechanisms have also been proposed in vector space models to act as a link between amodal and modal representations (Günther, Petilli, & Marelli, 2020; Günther, Petilli, Vergallito, & Marelli, 2020; Hollis, Westbury, & Lefsrud, 2017; Martínez‐Huertas, Jorge‐Botana, Luzón, & Olmos, 2021). In some contexts, this perspective has been termed the specific dimensionality hypothesis , as only some parts of the amodal representation seem to be mapping both formats of representation.…”

“…Thus, it would not be possible to predict all the emotional valence of a word from its amodal representation. Although the amodal distances of vector space models are effective to predict the emotional valence of words by means of semantic neighbors (Bestgen & Vincze, 2012; Hofmann et al., 2018; Kuhlmann et al., 2017; Recchia & Louwerse, 2015), only some amodal features are relevant for this purpose and thus there is scant isomorphism between modal and amodal representations (Hollis et al., 2017; Martínez‐Huertas et al., 2021). This means that some (but not all) features act as a bridge (mapping function) between modal and amodal emotional representations. Modal and amodal word representations can be reached even with no direct experience with emotional or sensorimotor responses.…”

“…Specifically, this study computationally models the amodal word representations using vector space models (see a brief introduction below), the modal word representations using emotional feature‐based models, and the mapping function with neural network models. Previous research has found that combining amodal word representations derived from vector space models and neural network models can be an efficient alternative to study such mapping function for emotional properties (see Hoffman, McClelland, & Lambon Ralph, 2018; Martínez‐Huertas, Jorge‐Botana, Olmos, & Martínez‐Mingo, 2020, 2021). Then, based on the Word Maturity methodology (Kireyev & Landauer, 2011; Landauer, Kireyev, & Panaccione, 2011; see the next section), this study will explore the mapping function between emotional and amodal word representations in two different developmental stages of semantic representation.…”