Data-driven Learning Meets Generative AI: Introducing the Framework of Metacognitive Resource Use

Mizumoto, Atsushi

doi:10.1016/j.acorp.2023.100074

Cited by 9 publications

(1 citation statement)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This would enable direct exploration of which other words occur with each emotion-body word pair. Additionally, the recent advances in generative language models may extend corpus-based methods in the future 21,[34][35][36][37] . This is particularly important considering the current uses of generative language models in Akkadian material, with a particular note that such methods have been used to explore metaphors that use words relating to parts of the body 21 .…”

Section: Methodological Considerationsmentioning

confidence: 99%

Embodied Emotions in Ancient Neo-Assyrian Texts Revealed by Bodily Mapping of Emotional Semantics

Lahnakoski,

Bennett,

Nummenmaa

et al. 2024

Preprint

View full text Add to dashboard Cite

Autonomic activity during emotions is subjectively felt in the body via interoceptive and somatosensory pathways resulting in emotion specific bodily sensations. Here we utilized this relationship and computational linguistic methods to map a representation of emotions in ancient texts. We analysed Neo-Assyrian texts from 934–612 BCE to discern consistent relationships between linguistic expressions related to both emotions and bodily sensations. We then computed statistical regularities between emotion terms and words referring to body parts, and back-projected the resulting emotion body part relationships on a standard body template, yielding bodily sensation maps for the emotions. We found consistent embodied patterns for eighteen distinct emotions. Hierarchical clustering revealed four main clusters of bodily emotion categories, two clusters of mainly positive emotions, one large cluster of mainly negative emotions and one of empathy and schadenfreude. These results reveal the historical use of embodied language pertaining to human emotions. Our data-driven tool projecting emotional lexical spaces on a human body provides a novel method for comparing textual embodiment patterns across different languages and cultures across time.

show abstract