Emotions play a central role in how information spreads on social media. Traditional approachestoemotiondetectionoftenrelyonfixed,categoricallabels,butthesestruggletocapturethe nuanced, context-dependent affective expressions that are common in online discourse. In this study, wetakeadimensionalapproachtoemotion,usingtheCircumplexModelofAffecttomodelemotional statesalongcontinuousvalenceandarousalaxes.Wefine-tunedtwoRoBERTamodelstopredictthese dimensions from tweets and evaluated their performance using the TruthSeeker2023 dataset. Our results demonstrate that dimensional modeling facilitates the more expressive and accurate detection of affective content than conventional emotion classifiers. These findings highlight the advantages of dimensional emotion modelling for affective analysis and demonstrate its relevance to understanding emotional mechanisms.

1. Step Clone Repository
2. Download Truth Seeker Dataset 2023

https://www.unb.ca/cic/datasets/truthseeker-2023.html

4. Install requirements.txt