BU-NEmo: an Affective Dataset of Gun Violence News

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 2507–2516 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 2507 BU-NEmo: an Affective Dataset of Gun Violence News Carley Reardon, Sejin Paik, Ge Gao, Meet Parekh, Yanling Zhao, Lei Guo, Margrit Betke, Derry Wijaya Boston University {reardonc, sejin, ggao02, meet09, lingzhao, guolei, betke, wijaya}@bu.edu Abstract Given our society’s increased exposure to multimedia formats on social media platforms, efforts to understand how digital content impacts people’s emotions are burgeoning. As such, we introduce a U.S. gun violence news dataset that contains news headline and image pairings from 840 news articles with 15K high-quality, crowdsourced annotations on emotional responses to the news pairings. We created three experimental conditions for the annotation process: two with a single modality (headline or image only), and one multimodal (headline and image together). In contrast to prior works on affectively-annotated data, our dataset includes annotations on the dominant emotion experienced with the content, the intensity of the selected emotion and an open-ended, written component. By collecting annotations on different modalities of the same news content pairings, we explore the relationship between image and text influence on human emotional response. We offer initial analysis on our dataset, showing the nuanced affective differences that appear due to modality and individual factors such as political leaning and media consumption habits. Our dataset is made publicly available to facilitate future research in affective computing. Keywords: affective computing, multimodal news data, crowdsourcing, language resources 1. Introduction Through the rise in popularity of online media con- sumption in the last decade (2011-2021), digital con- tent types are more varied than ever with increased multimedia experiences (Westcott et al., 2021; Shearer, 2021). In particular, a wide-range of research exists on the emotional impact of online, multimodal content. Examples of such work include measuring the emo- tional impact of social media post shares (Gupta and Yang, 2019), analyzing the influence of partisanship in affective political news (Hasell, 2020), and uncov- ering high-frequency emotions in fake news (Zhang et al., 2021b). Within machine learning, the task of affective emotion recognition aims to predict the emotional response that a given piece of media will elicit. While much work exists to benchmark the basic performance of this task, there is a lack of news media data for this task, with most notable datasets collecting abstract emotional re- sponses through art, such as Achlioptas et al. (2021). Some use social and/or news media, but they neglect to fully annotate the data with detailed emotional re- sponses, as in You et al. (2016), Yang et al. (2017), and Guo et al. (2021). This study focuses on gun-violence news as it contin- ues to significantly impact the lives of everyday peo- ple, with an average of 108 people in the U.S. dying daily from gun-related deaths (Lin, 2021) and federal data showing the rise of gun sales during the COVID- 19 pandemic (Schaeffer, 2021). Despite this, gun vi- olence and gun control continue …

Content truncated to 500 words