Contextual Affective Analysis: A Case Study of People Portrayals in Online #MeToo Stories

Abstract

In October 2017, numerous women accused producer Harvey Weinstein of sexual harassment. Their stories encouraged other women to voice allegations of sexual harassment against many high profile men, including politicians, actors, and producers. These events are broadly referred to as the #MeToo movement, named for the use of the hashtag “#metoo” on social media platforms like Twitter and Facebook. The movement has widely been referred to as “empowering” because it has amplified the voices of previously unheard women over those of traditionally powerful men. In this work, we investigate dynamics of sentiment, power and agency in online media coverage of these events. Using a corpus of online media articles about the #MeToo movement, we present a contextual affective analysis—an entity-centric approach that uses contextualized lexicons to examine how people are portrayed in media articles. We show that while these articles are sympathetic towards women who have experienced sexual harassment, they consistently present men as most powerful, even after sexual assault allegations. While we focus on media coverage of the #MeToo movement, our method for contextual affective analysis readily generalizes to other domains.1

Multilingual Contextual Affective Analysis of LGBT People Portrayals in Wikipedia

“…Recent advances in NLP have analyzed stereotypes and biases in narratives in English (Bamman, O'Connor, and Smith 2013;Wagner et al 2015;Sap et al 2017). Field, Bhat, and Tsvetkov (2019) establish a framework called Contextual Affective Analysis (CAA) that focuses on affective dimensions of power (strength/weakness), agency (active-ness/passiveness), and sentiment (goodness/badness). Analyzing portrayals of people along these dimensions (e.g., are men or women portrayed as more powerful?)…”

“…Analyzing portrayals of people along these dimensions (e.g., are men or women portrayed as more powerful?) has revealed stereotypes and bias in various domains, including movie scripts and newspaper articles (Rashkin, Singh, and Choi 2016;Sap et al 2017;Field, Bhat, and Tsvetkov 2019;Antoniak, Mimno, and Levy 2019).…”

“…Measuring these affective dimensions relies on connotation frames-lexicons of verbs annotated to elicit implications (Rashkin, Singh, and Choi 2016;Sap et al 2017). These connotations can be subtle, and the same verb often has different connotations in different contexts (Field, Bhat, and Tsvetkov 2019;. Until now, these manually-annotated affective lexicons have existed only for English.…”

“…We follow prior work in developing a supervised classifier trained on our contextualized multilingual annotations that can predict a connotation frame label y v for any unseen (in-context) verb v (Rashkin, Singh, and Choi 2016;Field, Bhat, and Tsvetkov 2019). Unlike prior models, ours is trained on contextualized annotations, and it leverages pretrained cross-lingual language models (CLMs).…”

#MeTooMA: Multi-Aspect Annotations of Tweets Related to the MeToo Movement

“…Our dataset could be utilized for extracting actionable insights and virtual dynamics to identify gender roles for analyzing sexual abuse revelations similar to (Yuce et al 2014). • The dataset could be utilized by psycholinguistics for extracting contextualized lexicons to examine how influential people are portrayed on public platforms in events of mass social media movements (Field, Bhat, and Tsvetkov 2019). Interestingly, such analysis may help linguists determine the power dynamics of authoritative people in terms of perspective and sentiment through campaign modeling.…”