“…Compared to hate speech, the detection of sexually explicit content has received less attention from the NLP community, with existing ML approaches focusing mainly on the detection of explicit images (Wehrmann et al, 2018;Rowley et al, 2006) and URLs (Matic et al, 2020), whereas n-grambased approaches remain predominantly used in practice by web providers (Hammami et al, 2003;Polpinij et al, 2006;Ho and Watters, 2004). In our analysis, we used a list of n-grams extracted from adult websites in order to establish the percentage of websites from our sample that contained sexually explicit content; however, we found no available statistical or ML-based approach that we could use to compare our count-based approach with.…”