19. Constructing a Taxonomy of Implicit Hate Speech Grounded in Social Theory with Diyi Yang and David Muchlinski

19. Constructing a Taxonomy of Implicit Hate Speech Grounded in Social Theory with Diyi Yang and David Muchlinski

Author: Katherine A. Keith, Naitian Zhou, & Lucy Li July 9, 2022 Duration: 56:02

Our guests on this episode are Diyi Yang, assistant professor at the School of Interactive Computing, and David Muchlinski, assistant professor in the Sam Nunn School of International Affairs, both at Georgia Tech. We discuss their EMNLP 2021 paper, "Latent Hatred: A Benchmark for Understanding Implicit Hate Speech." This paper is co-authored with Mai ElSherief, Caleb Ziems, Vaishnavi Anupindi, Jordyn Seybolt, and Munmun De Choudhury.

Diyi and David reveal that the annotation process behind this paper took two years and incorporated domain expertise on the broader context around hateful language. That is, an understanding of the social groups who produce this language allowed for better categorization and interpretation of implicit hate. We also discuss the cross-discipline connections they’ve forged in the past and present, and the ongoing challenges this type of work poses for computational methods.


Behind every published paper or headline-grabbing finding using social data, there's a hidden story of collaboration, dead ends, and problem-solving. Diaries of Social Data Research pulls back the curtain on that process. Hosted by researchers Katherine A. Keith, Naitian Zhou, and Lucy Li, this series sits down with scholars working at the intersection of computational methods and social science to explore the real, human effort behind the datasets. Each conversation functions like an open research diary, detailing how interdisciplinary teams actually come together, navigate differing academic cultures, and tackle the practical hurdles of working with massive, often messy, information about human behavior. You'll hear about the stalled projects, the unexpected breakthroughs, and the meticulous work that turns a raw idea into a credible contribution. This isn't a podcast about polished results, but about the fascinating and often untold journey of modern research. For anyone curious about how we actually study society through data-the alliances built, the ethics debated, and the code debugged late into the night-this series offers a rare and authentic look inside the lab.
Author: Language: English Episodes: 20

Diaries of Social Data Research
Podcast Episodes