2 octobre 2018
info:eu-repo/semantics/openAccess
Erik Haegert et al., « Feature Explorations for Hate Speech Classification », Elektronisches Publikationsportal der Österreichischen Akademie der Wissenschafte, ID : 10670/1.xg9bqm
In this work, we present a hate speech classifier for German tweets for the GermEval2018 Shared Task. Our best models are Linear SVM classifiers using character ngrams as well as additional textual features. We achieve a macro F1-score of 0.77 (95% confidence interval: ±0.04) in cross validation. We also present an ensemble classifier based on majority voting of the three component models.