Publikation: Comparison of Image Generation Models for Abstract and Concrete Event Descriptions
Dateien
Datum
Autor:innen
Herausgeber:innen
ISSN der Zeitschrift
Electronic ISSN
ISBN
Bibliografische Daten
Verlag
Schriftenreihe
Auflagebezeichnung
DOI (zitierfähiger Link)
Internationale Patentnummer
Angaben zur Forschungsförderung
Projekt
Open Access-Veröffentlichung
Sammlungen
Core Facility der Universität Konstanz
Titel in einer weiteren Sprache
Publikationstyp
Publikationsstatus
Erschienen in
Zusammenfassung
With the advent of diffusion-based image generation models such as DALL-E, Stable Diffusion and Midjourney, high quality images can be easily generated using textual inputs. It is unclear, however, to what extent the generated images resemble human mental representations, especially regarding abstract event knowledge. We analyse the capability of four state-of-the-art models in generating images of verb-object event pairs when we systematically manipulate the degrees of abstractness of both the verbs and the object nouns. Human judgements assess the generated images and demonstrate that DALL-E is strongest for event pairs with concrete nouns (e.g., “pour water”; “believe person”), while Midjourney is preferred for event pairs with abstract nouns (e.g., “raise awareness”; “remain mystery”), irrespective of the concreteness of the verb. Across models, humans were most unsatisfied with images of events pairs that combined concrete verbs with abstract direct-object nouns (e.g., “speak truth”), and an additional ad-hoc annotation contributes this to its potential for figurative language.
Zusammenfassung in einer weiteren Sprache
Fachgebiet (DDC)
Schlagwörter
Konferenz
Rezension
Zitieren
ISO 690
KHALIQ, Mohammed, Diego FRASSINELLI, Sabine SCHULTE IM WALDE, 2024. Comparison of Image Generation Models for Abstract and Concrete Event Descriptions. 4th Workshop on Figurative Language Processing (FigLang 2024). Mexico City, Mexico (Hybrid), 21. Juni 2024 - 21. Juni 2024. In: GHOSH, Debanjan, Hrsg., Smaranda MURESAN, Hrsg., Anna FELDMAN, Hrsg. und andere. Proceedings of the 4th Workshop on Figurative Language Processing (FigLang 2024). Kerrville, TX: Association for Computational Linguistics, 2024, S. 15-21. ISBN 979-8-89176-110-0. Verfügbar unter: doi: 10.18653/v1/2024.figlang-1.3BibTex
@inproceedings{Khaliq2024Compa-71192, year={2024}, doi={10.18653/v1/2024.figlang-1.3}, title={Comparison of Image Generation Models for Abstract and Concrete Event Descriptions}, isbn={979-8-89176-110-0}, publisher={Association for Computational Linguistics}, address={Kerrville, TX}, booktitle={Proceedings of the 4th Workshop on Figurative Language Processing (FigLang 2024)}, pages={15--21}, editor={Ghosh, Debanjan and Muresan, Smaranda and Feldman, Anna}, author={Khaliq, Mohammed and Frassinelli, Diego and Schulte Im Walde, Sabine} }
RDF
<rdf:RDF xmlns:dcterms="http://purl.org/dc/terms/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:bibo="http://purl.org/ontology/bibo/" xmlns:dspace="http://digital-repositories.org/ontologies/dspace/0.1.0#" xmlns:foaf="http://xmlns.com/foaf/0.1/" xmlns:void="http://rdfs.org/ns/void#" xmlns:xsd="http://www.w3.org/2001/XMLSchema#" > <rdf:Description rdf:about="https://kops.uni-konstanz.de/server/rdf/resource/123456789/71192"> <dc:creator>Schulte Im Walde, Sabine</dc:creator> <dc:creator>Khaliq, Mohammed</dc:creator> <dc:contributor>Khaliq, Mohammed</dc:contributor> <dcterms:abstract>With the advent of diffusion-based image generation models such as DALL-E, Stable Diffusion and Midjourney, high quality images can be easily generated using textual inputs. It is unclear, however, to what extent the generated images resemble human mental representations, especially regarding abstract event knowledge. We analyse the capability of four state-of-the-art models in generating images of verb-object event pairs when we systematically manipulate the degrees of abstractness of both the verbs and the object nouns. Human judgements assess the generated images and demonstrate that DALL-E is strongest for event pairs with concrete nouns (e.g., “pour water”; “believe person”), while Midjourney is preferred for event pairs with abstract nouns (e.g., “raise awareness”; “remain mystery”), irrespective of the concreteness of the verb. Across models, humans were most unsatisfied with images of events pairs that combined concrete verbs with abstract direct-object nouns (e.g., “speak truth”), and an additional ad-hoc annotation contributes this to its potential for figurative language.</dcterms:abstract> <dc:contributor>Frassinelli, Diego</dc:contributor> <dcterms:issued>2024</dcterms:issued> <dc:contributor>Schulte Im Walde, Sabine</dc:contributor> <dspace:isPartOfCollection rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/45"/> <dcterms:isPartOf rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/45"/> <foaf:homepage rdf:resource="http://localhost:8080/"/> <dc:date rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2024-11-11T11:42:02Z</dc:date> <bibo:uri rdf:resource="https://kops.uni-konstanz.de/handle/123456789/71192"/> <dcterms:title>Comparison of Image Generation Models for Abstract and Concrete Event Descriptions</dcterms:title> <dcterms:available rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2024-11-11T11:42:02Z</dcterms:available> <dc:creator>Frassinelli, Diego</dc:creator> <dc:language>eng</dc:language> <void:sparqlEndpoint rdf:resource="http://localhost/fuseki/dspace/sparql"/> </rdf:Description> </rdf:RDF>