Konuşmacılar
Açıklama
Social media serves as a forum where individuals, organizations, and businesses can communicate information and opinions on any topic. The apparent anonymity has fostered a lack of moral or social filters for posting content, regardless of whether it's false, discriminatory, or defamatory. This resulted in posts and comments becoming fertile ground for research into representations and discourses. Likewise, a topic that triggers passionate and extreme reactions is gender, particularly concerning women and their struggles for equality, equity, and against physical and sexual violence.
In 2019, the artistic and feminist collective, Las Tesis, performed “A rapist on your way”. Its audiovisual recording went viral on social media, leading to women worldwide replicating it and sparking discussions among users.
This, coupled with advancements in natural language processing, motivated us to conduct a discourse analysis about women in the comments of a YouTube video discussing the representation of this performance across different parts of the world. From a corpus of 14,884 comments on that video, we set two objectives: to determine the prevalent discourses regarding women's roles and to implement and compare two methodological approaches for analysis.
The first involved applying automated methods, employing topic modelling (i.e., Latent Dirichlet Allocation (Blei et al., 2003) and sentiment analysis with Large Language Models (i.e., RoBERTuito (Pérez et al., 2022) and BETO (Cañete et al., 2020)). The second used textual analysis through the Appraisal Theory (Martin and White, 2005). The results indicate that computational models do not provide clarity regarding content and evaluations on social media. However, they are useful for text filtering. The Appraisal Theory allowed us to systematize the areas, actions, and characteristics of women being evaluated, highlighting discriminatory discourse and a literal interpretation of the artistic piece.
Keywords: Sentiment Analysis, Appraisal Theory, Spanish, Large Language Models, Topic Modelling
Institution / Affiliation / Kurum
Potsdam University
| Presentation language / Sunum Dili | EN (English) |
|---|---|
| Disciplines / Disiplinler | Linguistics / Dilbilim |
| E-mail / E-posta | valentina.tretti@uni-potsdam.de |
| ORCID ID | 0000-0001-9132-3224 |
| Country / Ülke | Alemania |
