The audiovisual structure of onomatopoeias: An intrusion of real-world physics in lexical creation
Fig 4
Decoding the movement type of cross-linguistic onomatopoeias.
a. We train the LDA model with the AV onomatopoeias and used it as a movement-type classifier (hit, slide or ring) for the cross-linguistic onomatopoeias extracted from Wikipedia [26]. The list was restricted to non-human/animal actions with onomatopoeias in 10 languages for each action (balloon bursting, camera shutter, etc). The color code corresponds to the percentage of languages in which onomatopoeias are classified in each movement category. b. The same actions were classified by a group of 20 human raters, showing good agreement with the model-derived predictions.