Repositorio Digital de la UPF
Publicación
Artículo científico (JournalArticle)
How well do hate speech, toxicity, abusive and offensive language classification models generalize across datasets?
A considerable body of research deals with the automatic identification of hate speech and
related phenomena. However, cross-dataset model generalization remains a challenge. In this
context, we address two still open central questions: (i) to what extent does the generalization
depend on the model and the composition and annotation of the training data in terms of
different categories?, and (ii) do specific features of the datasets or models influence the
generalization potential? To answer (i), we experiment with BERT, ALBERT, fastText, and SVM
models trained on nine common public English datasets, whose class (or category) labels are
standardized (and thus made comparable), in intra- and cross-dataset setups. The experiments
show that indeed the generalization varies from model to model and that some of the categories
(e.g., ‘toxic’, ‘abusive’, or ‘offensive’) serve better as cross-dataset training categories than others
(e.g., ‘hate speech’). To answer (ii), we use a Random Forest model for assessing the relevance
of different model and dataset features during the prediction of the performance of 450 BERT,
450 ALBERT, 450 fastText, and 348 SVM binary abusive language classifiers (1698 in total). We
find that in order to generalize well, a model already needs to perform well in an intra-dataset
scenario. Furthermore, we find that some other parameters are equally decisive for the success
of the generalization, including, e.g., the training and target categories and the percentage of
the out-of-domain vocabulary., The first author is supported by the research grant SFRH/BD/143623/2019, provided by the Portuguese national funding agency
for science, research and technology, Fundação para a Ciência e a Tecnologia (FCT), within the scope of Operational Program Human
Capital (POCH), supported by the European Social Fund and by national funds from MCTES. The work of the second and third authors
has been supported by the European Commission in the context of the H2020 Research Program under the contract numbers 700024
and 786731.
Proyecto:
EC / H2020 700024
,
EC / H2020 786731