I delivered a live session at SomosNLP, the largest Spanish-speaking NLP community, exploring media bias through an NLP research lens. The presentation covered how media bias can be defined and contextualized, how it can be addressed during dataset creation, and how NLP techniques can be applied to detect it in practice.
Key topics included: definitions and manifestations of media bias in journalism; bias mitigation strategies during dataset development using Counterfactual Data Augmentation (CDA); a perspectivist NLP approach incorporating Learning with Disagreements (LeWiDi); and cascade-based text classification for hierarchical datasets.
The session was moderated by María Grandury, SomosNLP founder, and connected to the SomosNLP Hackathon 2024.