Natural Language Processing is increasingly adopted within media studies for large scale analysis of discourse and representational patterns. While techniques such as sentiment analysis and topic modelling can reveal how social issues are framed across time and platforms, their application to marginalised subjects such as transgender representation, risk reinforcing existing discursive biases. This study analyses a corpus of 10,629 English language legacy media articles from the UK and US (2014–2025), using RoBERTa based sentiment analysis and BERTopic topic modelling to examine how trans athletes are represented within mainstream sports reporting. The findings demonstrate significant tensions between computational outputs and established theories of media discourse. For example, RoBERTa frequently misclassifies exclusionary or anti trans positions as “positive” due to the model’s reliance on surface level affective cues, such as descriptions of bans on trans athletes as “fair”, “balanced”, or “protecting women’s sport”, revealing a mismatch between emotional valence and ideological stance. Similarly, topic modelling systems tend to marginalise or treat community-centred narratives as statistical outliers, reinforcing patterns of erasure long noted in critical media scholarship. Together, these limitations highlight how NLP can inadvertently reproduce rather than illuminate dominant ideological frames. The paper argues for two interventions: (1) the retraining and fine tuning of NLP models using inclusive, community-informed datasets derived from activist and independent media sources such as Trans Media Watch and Pink News, and (2) the integration of qualitative, multimodal, and discourse-analytic methods to ensure computational insights remain anchored in media studies’ long-standing commitments to interrogating power, representation and inequality.

Authors:
Yen Nee Wong, University of Leeds, United Kingdom

About the Presenter(s)
Dr Yen Nee Wong is currently a Lecturer in Sociology and Social Policy in the University of Leeds, United Kingdom. Their recent projects are in the areas of dance, media and culture, queer theory, and algorithmic-driven technologies.

See this presentation on the full schedule – Monday Schedule

Bookmark