Comparing pre-trained language models for Spanish hate speech detection