Vis enkel innførsel

dc.contributor.authorStave, Daniel Årrestad
dc.contributor.authorKorneliussen, Hanne
dc.contributor.authorHjellup, H. Nøkkelby
dc.contributor.authorShrestha, Raju
dc.date.accessioned2023-11-08T09:19:56Z
dc.date.available2023-11-08T09:19:56Z
dc.date.created2023-11-06T11:32:10Z
dc.date.issued2023
dc.identifier.isbn979-8-3503-4823-1
dc.identifier.isbn979-8-3503-4824-8
dc.identifier.urihttps://hdl.handle.net/11250/3101314
dc.description.abstractToday, many artificial or virtual influencers roam social media platforms to maximise followers and offer commercial options for companies. This work focuses on developing artificial influencers using state-of-the-art techniques within deep learning. Specifically, an autonomous theoretical framework for generating social media content that maximises user engagement is proposed. Deep learning models for generating realistic images and hashtags are trained on a dataset from a social media platform, and content is optimised for user engagement using an evolutionary algorithm. The generated images were evaluated by participants from existing social media users through two separate surveys. The complete framework is built, trained, and tested, and functionality is confirmed. The framework, which appears to be the first of its kind, produces content that matches the users' preferences well.en_US
dc.language.isoengen_US
dc.publisherInstitute of Electrical and Electronics Engineers (IEEE)en_US
dc.relation.ispartofThe 4th International Conference on Artificial Intelligence, Robotics and Control (AIRC 2023)
dc.titleSoMeMax - A Novel AI-driven Approach to Generate Artificial Social Media Content That Maximises User Engagementen_US
dc.typeChapteren_US
dc.typePeer revieweden_US
dc.typeConference objecten_US
dc.description.versionacceptedVersionen_US
cristin.ispublishedtrue
cristin.fulltextpostprint
dc.identifier.doihttps://doi.org/10.1109/AIRC57904.2023.10303076
dc.identifier.cristin2192526
dc.source.pagenumber7en_US


Tilhørende fil(er)

Thumbnail

Denne innførselen finnes i følgende samling(er)

Vis enkel innførsel