BU-NEmo: news and emotions dataset
Date
2022-06-20
DOI
Authors
Reardon, Carley
Paik, Sejin
Gao, Ge
Parekh, Meet
Zhao, Yanling
Guo, Lei
Betke, Margrit
Wijaya, D.
Version
OA Version
Citation
C. Reardon, S. Paik, G. Gao, M. Parekh, Y. Zhao, L. Guo, M. Betke, D. Wijaya. 2022. "BU-NEmo: News and Emotions Dataset"
Abstract
BU-NEmo is a multimodal affective dataset of gun violence news content. BU-NEmo extends the Gun Violence Framing Corpus (GVFC) proposed by Liu et. al (2019) and Tourni et. al (2021), which contains pairs of news headlines and lead images and their "frames" (view points) from gun violence-related articles. The extension concerns the results of an annotation experiment that evaluates the effect of the news content on the emotions of news consumers.
The data in BU-NEmo are annotated with three types of affective annotations:
(1) The emotion the annotator feels from looking at the content, out of the following 8 classes: Amusement, Awe, Contentment, Excitement, Fear, Sadness, Anger, Disgust.
(2) The intensity of the annotator's emotional response, on a scale from 1-5 (5 being the most intense).
(3) A free-text written response explaining their emotional response, structured as "I feel because."
These annotations were collected in three experimental conditions: only the headline text was presented, only the image, and text and image together. By comparing the annotations across these three conditions, the relationship between news modality, frames, and emotional response can be studied.