Computational approaches to the study of systematicity in first language acquisition

Simple item page
dc.contributor.advisorLarson, M.A.
dc.contributor.advisorCassani, G.
dc.contributor.advisorFitz, H.
dc.contributor.authorVarst, van der, Willeke
dc.date.issued2023-02-20
dc.description.abstractIn this work we investigate the relationship between word from, word meaning, instantiated as word co-occurrence patterns, and lexical visual representations during language acquisition. Our study improves over previous approaches to the study of systematicity in language acquisition, which looked at correlations between form and meaning representations of known words and could not assess the degree of systematicity of a novel word with respect to a learner’s current vocabulary, which we did do. We used Form-to-Semantic Consistency (FSC) [7] and Linear Discriminative Learning (LDL) [5] to compute systematicity values for novel words based on the vocabulary learned up to that point to investigate whether systematicity between word form and co-occurrence patterns facilitates word learning. We found that systematicity computed using FSC, but not LDL, was a reliable predictor of when a word would be acquired, and this predictive effect depends on a small pool of very systematic words. Results from FSC support previous studies which have posited a beneficial role of non-arbitrariness in relation to language acquisition. Then, we used neural networks to investigate the relationship between word forms, co-occurrence patterns and corresponding visual representations during language acquisition. We found that word co-occurrence patterns are a good predictor of visual information, while at no stage during language acquisition word from information is a good predictor of visual information. We found that neither word co-occurrence nor word form information was a reliable predictor of when a word would be acquired. Thus, there is no indication from our neural network studies that either word co-occurrence or word form information is important to facilitate language acquisition. Results from our neural network studies fail to capture a relation between word form and corresponding image, which previous studies have, calling for future studies on how to implement a mechanistic account of how linguistic form and co-occurrence patterns map onto the visual world in a way that uncovers learning mechanisms present during acquisition.
dc.identifier.urihttps://theses.ubn.ru.nl/handle/123456789/16424
dc.language.isoen
dc.thesis.facultyFaculteit der Sociale Wetenschappen
dc.thesis.specialisationspecialisations::Faculteit der Sociale Wetenschappen::Artificial Intelligence::Master Artificial Intelligence
dc.thesis.studyprogrammestudyprogrammes::Faculteit der Sociale Wetenschappen::Artificial Intelligence
dc.thesis.typeMaster
dc.titleComputational approaches to the study of systematicity in first language acquisition
Files
Original bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
Varst, van der, W. s-1059500-MSc-MKI94-Thesis-2023.pdf
Size:
1.46 MB
Format:
Adobe Portable Document Format
Download
Collections
Faculteit der Sociale Wetenschappen