Transfer Learning via Lexical Relatedness: A Sarcasm and Hate Speech Case Study
This work addresses the problem of detecting non-direct hate speech for social networks, but it is incremental as it builds on existing transfer learning methods with specific datasets.
The paper tackled the challenge of detecting implicit hate speech, such as irony and sarcasm, by exploring whether sarcasm pre-training improves detection performance. Results showed that sarcasm pre-training increased recall by 9.7%, AUC by 7.8%, and F1-score by 6% on the ETHOS dataset, and precision by 7.8% on the Implicit Hate Corpus.
Detecting hate speech in non-direct forms, such as irony, sarcasm, and innuendos, remains a persistent challenge for social networks. Although sarcasm and hate speech are regarded as distinct expressions, our work explores whether integrating sarcasm as a pre-training step improves implicit hate speech detection and, by extension, explicit hate speech detection. Incorporating samples from ETHOS, Sarcasm on Reddit, and Implicit Hate Corpus, we devised two training strategies to compare the effectiveness of sarcasm pre-training on a CNN+LSTM and BERT+BiLSTM model. The first strategy is a single-step training approach, where a model trained only on sarcasm is then tested on hate speech. The second strategy uses sequential transfer learning to fine-tune models for sarcasm, implicit hate, and explicit hate. Our results show that sarcasm pre-training improved the BERT+BiLSTM's recall by 9.7%, AUC by 7.8%, and F1-score by 6% on ETHOS. On the Implicit Hate Corpus, precision increased by 7.8% when tested only on implicit samples. By incorporating sarcasm into the training process, we show that models can more effectively detect both implicit and explicit hate.