| language: en | |
| tags: | |
| - transfer-learning | |
| - bert | |
| - hatespeech | |
| - covid19 | |
| license: "mit" | |
| datasets: | |
| - COVID-HATE | |
| metrics: | |
| - f1-score | |
| Since the start of the COVID-19 pandemic, there has been a widespread increase in the amount of hate-speech being propagated online against the Asian community. This project builds upon and explores the work of He et al. Their COVID-HATE dataset contains 206 million tweets focused around anti-Asian hate speech. Using tweet data from before the COVID-19 pandemic, as well as the COVID-HATE dataset from He et al, we performed transfer learning. We tested several different models, including BERT, RoBERTa, LSTM, and BERT-CNN. | |
| Some of these models hindered the performance of He et al’s model, while others improved it. | |