Amazon Mechanical Turk

Measuring a hate speech spectrum with faceted Rasch item response theory and perspective-aware, explainable-by-design deep learning

arXiv:2009.10277v2 Announce Type: replace Abstract: We propose a system for measuring hate speech on a continuous, interval-valued spectrum ranging from genocidal to supportive speech by combining supervised deep learning with faceted Rasch item response theory (IRT). We decompose the theoretical construct of hate speech into constituent concepts operationalized as 10 ordinal labels. Those labels are reconstituted via IRT probabilistic latent modeling into an interval outcome measure while...